Download Full Version ebook - Visit ebookmeta.

com

Training Data for Machine Learning: Human

Supervision from Annotation to Data Science 1st
Edition Anthony Sarkis

https://ebookmeta.com/product/training-data-for-machine-
learning-human-supervision-from-annotation-to-data-
science-1st-edition-anthony-sarkis/

OR CLICK HERE

DOWLOAD NOW

Discover More Ebook - Explore Now at ebookmeta.com

 Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Start reading on any device today!

Training Data for Machine Learning: Human Supervision from

Annotation to Data Science 1st Edition Anthony Sarkis

https://ebookmeta.com/product/training-data-for-machine-learning-
human-supervision-from-annotation-to-data-science-1st-edition-anthony-
sarkis/
ebookmeta.com

Training Data for Machine Learning Models 1st Edition

Anthony Sarkis

https://ebookmeta.com/product/training-data-for-machine-learning-
models-1st-edition-anthony-sarkis/

ebookmeta.com

Scaling Python with Dask: From Data Science to Machine

Learning 1st Edition Holden Karau

https://ebookmeta.com/product/scaling-python-with-dask-from-data-
science-to-machine-learning-1st-edition-holden-karau/

ebookmeta.com

The Mini Book of Pies Sophie Conran

https://ebookmeta.com/product/the-mini-book-of-pies-sophie-conran/

ebookmeta.com
Sacrificial Princess and the King of Beasts Volume 5 1st
Edition Yu Tomofuji

https://ebookmeta.com/product/sacrificial-princess-and-the-king-of-
beasts-volume-5-1st-edition-yu-tomofuji/

ebookmeta.com

MXenes Fundamentals and Applications 1st Edition Inamuddin

https://ebookmeta.com/product/mxenes-fundamentals-and-
applications-1st-edition-inamuddin/

ebookmeta.com

Wine Tasting: A Professional Handbook Ronald S. Jackson

https://ebookmeta.com/product/wine-tasting-a-professional-handbook-
ronald-s-jackson/

ebookmeta.com

Cotton Candy 1st Edition Stuart Drake Bray

https://ebookmeta.com/product/cotton-candy-1st-edition-stuart-drake-
bray/

ebookmeta.com

Peasant Protest and Social Change in Colonial Korea Korean

Studies of the Henry M Jackson School of International
Studies Gi-Wook Shin
https://ebookmeta.com/product/peasant-protest-and-social-change-in-
colonial-korea-korean-studies-of-the-henry-m-jackson-school-of-
international-studies-gi-wook-shin/
ebookmeta.com
Sir Isaac Newton Overlord of Gravity 1st Edition Angela
Royston

https://ebookmeta.com/product/sir-isaac-newton-overlord-of-
gravity-1st-edition-angela-royston/

ebookmeta.com
 Training Data for Machine
Learning Models
With Early Release ebooks, you get books in their earliest form—
the author’s raw and unedited content as they write—so you can
take advantage of these technologies long before the official
release of these titles.

Anthony Sarkis
Training Data for Machine Learning Models
by Anthony Sarkis
Copyright © 2021 Anthony Sarkis. All rights reserved.
Printed in the United States of America.
Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North,
Sebastopol, CA 95472.
O’Reilly books may be purchased for educational, business, or sales
promotional use. Online editions are also available for most titles
(http://oreilly.com). For more information, contact our
corporate/institutional sales department: 800-998-9938 or
[email protected].

Acquisitions Editor: Rebecca Novack

Development Editor: Jill Leonard

Production Editor: Caitlin Ghegan

Copyeditor: TK

Proofreader: TK

Indexer: TK

Interior Designer: David Futato

Cover Designer: Karen Montgomery

Illustrator: Kate Dullea

January 2022: First Edition

Revision History for the Early Release
2021-05-21: First release
2021-10-13: Second release

See http://oreilly.com/catalog/errata.csp?isbn=9781492094524 for

release details.
The O’Reilly logo is a registered trademark of O’Reilly Media, Inc.
Training Data for Machine Learning Models, the cover image, and
related trade dress are trademarks of O’Reilly Media, Inc.
The views expressed in this work are those of the author, and do not
represent the publisher’s views. While the publisher and the author
have used good faith efforts to ensure that the information and
instructions contained in this work are accurate, the publisher and
the author disclaim all responsibility for errors or omissions,
including without limitation responsibility for damages resulting from
the use of or reliance on this work. Use of the information and
instructions contained in this work is at your own risk. If any code
samples or other technology this work contains or describes is
subject to open source licenses or the intellectual property rights of
others, it is your responsibility to ensure that your use thereof
complies with such licenses and/or rights.
978-1-492-09452-4
[LSI]
Chapter 1. Training Data Introduction

A NOTE FOR EARLY RELEASE READERS

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited
content as they write—so you can take advantage of these technologies long before the official
release of these titles.
This will be the 1st chapter of the final book. Please note that the GitHub repo will be made active
later on.
If you have comments about how we might improve the content and/or examples in this book, or if
you notice missing material within this chapter, please reach out to the author at
[email protected].

What is Training Data?

Training Data is the control of a Supervised System.
Training Data controls the system by defining the ground truth goals for the creation of Machine
Learning models. This involves technical representations, people decisions, processes, tooling, system
design, and a variety of new concepts specific to Training Data. In a sense, a Training Data mindset is a
paradigm upon which a growing list of theories, research and standards are emerging.

Model
A Machine Learning (ML) Model that is created as the end result of a ML Training Process.
 Figure 1-1. Diagram or screenshot of common supervision interface.

Training Data is not an algorithm, nor is it tied to a specific machine learning approach. Rather it’s the
definition of what we want to achieve. A fundamental challenge is effectively identifying and mapping
the desired human meaning into a machine readable form.
The effectiveness of training data depends primarily on how well it relates to the human defined
meaning and how reasonably it represents real model usage. Practically, choices around Training Data
have a huge impact on the ability to train a model effectively.
Training Data makes sense when a set of conditions are true. For example, training data for a parking
lot detection system may have very different views. If we create a training data set based on a top
down view (see left side of Figure 1.2) and then attempt to use the image on the right we will get
unexpected results. That’s why it’s important that the data we use to train a system closely matches the
data our trained system would see in production.
 Figure 1-2. If the left is your training data and right is your use case, you are in trouble!

See figure 1-2—A machine learning system trained only on images from a top-down view as in the left
image has a hard time running in an environment where the images are from a front view front as in
the right image. Our system would not understand the concept of a car and parking lot from a front
view if it has never seen such an image during training.

Good Robot, Bad Robot

Training a machine to understand and intelligently interpret the world may feel like a monumental task.
But there’s good news! The algorithms behind the scenes do a lot of the heavy lifting. Our primary
concern with training data can be summed up as defining what’s good, what should be ignored, and
what’s bad.
Of course real Training Data requires a lot more than a head nod or head shake. We must find a way to
transform our rather ambiguous human terminologies into something the machine can understand.
Thinking of Training Data as Code
One way to think of Training Data is as a higher level programming language. In the same way we use
languages like Python to be more expressive than Assembly code1, we can use Training Data to be more
expressive than Python. Instead of trying to define a Strawberry as hundreds of lines of if statements,
and manually identifying key features, we can simply draw and label the Strawberry as such. Similar to
how we just tell a child “that’s a strawberry”.
This is distinct from using an application, such as a Low Code system, because it is directly
programming meaning to raw data. It’s real programming, just at a higher level of abstraction. Like
code, if the Training Data is changed, then the results of the system are changed. From a data science
perspective, training data is integral to the machine learning process because it’s the input for Training
and the ground truth upon which to measure results.
As a brief comparison, traditional software programs often have Business Logic, or logic that only makes
sense in the business context exterior to the program. Similarly training data relates to a context
exterior to itself, often the real world. While business logic is inferred, Training Data is literally directly
mapped to that exterior context of raw data and related assumptions. It’s human meaning encoded in a
form ready for consumption by a machine learning algorithm. Training Data is Code.

Concepts Introduction
There are two general categories of training data: Classic and Supervised. The general focus of this
book is on Supervised. We will contrast Supervised to Classic later in more detail. The following
presentation of concepts is intended to be introductory to provide a baseline understanding around
definitions and assumptions. These themes will be explored in greater detail throughout the book.

Representations
Let’s imagine we are working on a Strawberry picking system. We need to know what a strawberry is,
where it is, and how ripe it is. Let’s introduce a few new terms to help us work more efficiently:
Label
A Label, also called a class2 Other names include: Label template, represents the highest level of
meaning. For example, a label can be “Strawberry” or “Leaf”. For those technical folks, you can think
of it as a Table in a database. This is usually attached3 to a specific Instance.

Instance
A single example. It is connected to a label to define what it is. And usually contains positional or
spatial information defining where it is. Continuing the technical example, this is like a row in a
database. An Instance may have many Attributes.
 Figure 1-3. Diagram showing labeled and not labeled instances.

Attributes
Attributes are things unique to a specific Instance. Imagine you only want the system to pick
Strawberries of a certain ripeness. You may represent the ripeness as a slider or you may also have
a multiple choice question answer about ripeness. From the database perspective this is somewhat
akin to a column. This choice will affect the speed of supervision. A single instance may have many
unique attributes, for example in addition to ripeness there may be disease identification, produce
quality grading, etc.
Choices
Your choices here are as much part of Training Data as doing the literal supervision. You also make
choices on what type of Spatial representation to use.

Spatial
We can represent the Strawberries as boxes, polygons, and other options. In later chapters I will discuss these
trade-offs.
There are other choices which will be detailed in later chapters. From a system design perspective there
can be choices around what type of raw data to use, such as image, audio, text, or video. And
sometimes even combining multiple modalities together. Angle, size, and other attributes may also apply
here. Unlike attributes generally, spatial locations are singular in nature. An object is usually only in one
place at a given moment in time.4

Who Supervises the Data

For our strawberry system, who supervises the data may matter depending on the context. For
example, if the system is to be installed in a grocery store, perhaps a grocery store employee is best
able to identify if the strawberries are “OK” to sell or not. Whereas an automated farm installation may
require greater precision.
This leads us to considerations around who supervises the data. Their backgrounds, biases, incentives
and more. What incentives are the people who supervise the data given? How much time is being spent
per sample? When is the person doing it? Are they doing 100 samples at once?
As a supervisor, you may work for a firm that specializes in data supervision, or you may be hired to one
company, or you may be a subject matter expert. Finally, the supervision may come directly from an
end user, who is potentially unaware they are even doing supervision.
Generally if an end user is doing the supervision the volume and depth of supervision will be lower.
However it may be more timely, and more specific. Consider that both can be used together. For
example an end user suggesting something was “bad” may be used as a flag to initiate further direct
supervision.

This is all in the context of explicit (or direct) supervision of the data. Someone is directly viewing the
data and taking action. This contrasts with classic training data where the data is implicitly observed “in
the wild” and not editable by humans.
We will cover this in more detail in Chapter 3.

Sets of Assumptions
Imagine you are an Olympic runner training to run in the set of conditions that are expected to be
present at the Olympics. It’s likely if you are training for a specific event, say the 100 Meter, then you
will train only for the 100 Meter, and not for the 400, 800, or High Jump. Because while similar, those
events are outside of the scope of what you expect to encounter—the 100 Meter.
Training Data is very similar. We define a set of assumptions and expect the Training Data to be useful
in that context - and only in that context. Similar to the above we can start with high level assumptions.
Our strawberry picker is assumed to be on a commercial strawberry field. Then, like the 100M
specificity, we can get into more specific assumptions. For example perhaps we assume the camera will
be mounted to a drone, or a ground based robot, or the system will be used during the day.
Assumptions are in any system, however, they take on a different context with Training Data because of
the inherent randomness involved. Somewhat surprisingly, human analogies around Training (for sports,
work etc.) are actually very applicable to Training Data.

Randomness
Let’s zoom in on this human centric example of training for the Olympics. I can train how to do
something for all my life, such as beating an Olympic record - and still not be 100% certain that I will
be able to do it. In fact for many things, I can probably only be certain I won’t be able to do it. The
intuition that I am not guaranteed to be an olympian is clear. Getting a similar intuition around AI
training is part of the challenge.
This is especially hard because we typically think of computer systems as being deterministic, meaning
if I run the same operation twice I will get the same result. The way AI models get trained is not
deterministic. The world in which AIs operate is not deterministic. The processes around creation of
training data involve humans and are not deterministic. Therefore at the heart of training data is an
inherent randomness. Much of the work with training data, especially around the abstractions, is
defining what is and is not possible in the system. Essentially trying to reign in and control the
randomness into more reasonable bounds.
We create training data to cover the expected cases. What we expect to happen. And as we will cover
in more depth later, we primarily use rapid retraining of the data to handle the expected randomness of
the world.

Processes and Process Automation

Defining a process is one of the most fundamental ways to set up guardrails around randomness. Even
the most basic supervision programs require some form of process. This defines where the data is, who
is responsible for what, the status of tasks, etc. Quality assurance in this context is a mini artform with
many competing approaches and opinions. We will discuss multiple levels of sophistication here, going
from manual to fully automated, self healing, multi-stage pipelines.

Figure 1-4. Process visual

Supervision Automation and Tooling
As soon as you understand the basics of training data you will quickly realize there is an obvious
bottleneck: Doing the literal annotation. There have been many approaches to speed up this core part
of the process. We will explore the processes and trade-offs. Choices here can quickly become very
complex. It’s one of the most often misunderstood parts but is an important part of Training Data.

Dataset Construction & Maintenance

When creating a model, a common practice is to split an original set into three subsets, it’s called
Train/Val/Test. The idea is to have a set that’s trained on, a second set that’s withheld from training to
Validate on, and a 3rd set that’s held in reserve until work is complete, to do a single use final test on.
But where did the “Original” set come from?! This is the concern of the Training Data context,
constructing the Original set.
Usually there is more raw data then can be annotated. So part of the basic selection process is choosing
which of the raw samples to be annotated. There are normally multiple Original sets. In fact many
projects quickly grow to have hundreds of these Original sets. More generally, this is the concern of
overall organization and structure of the Datasets, including selecting which samples are to be included
in which sets.

Relevancy
Continuing the theme of validation of existing data, how do we know if our data is actually relevant to
the real world? For example, we could get a perfect 100% on a test set, but if the test set isn’t relevant
to the real world data then we are in trouble! There is currently no known direct test that can define if
the data is relevant to the production data - only time will truly tell. This is similar to the way a
traditional system can be loaded tested for x number of users, but the true usage pattern will always be
different. Of course we can do our best to plan for the expected use and design the system.

Integrated System Design

There are usually many choices of how to design the data collection system in supervised learning. For
example, for a grocery store, a system could be positioned over an existing checkout register, such as to
prevent theft, or aid in faster checkout. Or, a system could be designed to replace the checkout entirely,
such as placing many cameras throughout and tracking shopper actions. There is no right or wrong
answer here - the primary thing to be aware of is that the Training Data is tied to the context it’s
created in. In later sections we will cover system design here in much more depth.
 Figure 1-5. Goal of showing left an existing check out process, vs on right a novel “checkout free” process

Raw Data Collection

Beyond the high level system design perspective, the collection and storage of raw data is generally
beyond the scope of this book. This is part because of the vast array of options for raw data. Raw data
can be real life sensors, it can be screenshots of webpages, pdf scans, etc. Virtually anything that can
be represented as an image, video, or text can be used for raw data.

What-To-Label
As part of Dataset Construction we know we need to create a smaller set from a larger set of raw data -
but how? This is the concern of What-To-Label. Generally, the goal of these approaches is to find
“interesting” data. For example if I have thousands of images that are similar, but 3 are obviously
different, I may get the most bang for my buck if I start with the 3 most different ones.

Iterations
In traditional programming we iterate on the design of functions, features, and systems. In Training
Data there is a similar form of iteration on all of the concepts we are here in discussing. The models
themselves are also iterative, for example they may be retrained on a predetermined frequency, such as
daily. Two of the biggest competing approaches here are the “Fire and forget” and the “Continual
Retrain”. In some cases it may be impractical to retrain, and so a single, final model is created.
Transfer Learning
The idea of transfer learning is to start off from an existing knowledge base before training a new
model.
Transfer learning is used to dramatically speed up training new models. From a Training Data view,
transfer learning introduces challenges around bias. Because we are indirectly using training data from
that prior model training. If there was undesirable bias in that model, it may carry over to our new case.
Essentially it creates a dependency on that prior training data set. Dependencies are an unavoidable
reality of software, but it’s important to be aware of them and surface the trade-offs clearly.

Per Sample Judgement Calls

Ultimately a human will supervise each sample, generally one sample at a time. In all of this we must
not forget the decisions each person makes has a real impact on the final result. There are no easy
solutions here. There are tools available such as taking averages of multiple opinions, requiring
examinations etc.
Often people, including experts, simply have different opinions. To some extent these unique
judgements can be thought of as a new form of intellectual property. Imagine an oven with a camera. A
chef who has a signature dish could supervise a training dataset that in a sense reflects that chef’s
unique taste. This is a light intro to the concept that the line between system and user content becomes
blurred with training data in a way that’s still developing.

Ethical & Privacy Considerations

First, it’s worth considering that some forms of supervised data are actually relatively free of bias. For
example, it’s hard for me to imagine any immediate ethical or privacy concerns from our strawberry
picking dataset. There are however very real and very serious ethical concerns in certain contexts. This
is not an ethics book and there are already some relatively extensive books on the effects of automation
more generally. However, I will touch on some of the most immediate and practical concerns.

Technical Specifics
There are a variety of technical specifics, such as formats and representations that I will cover in some
detail. While generally these representations have a “flavor of the month” feel, I will cover some of the
currently popular ones and speak to the general goals the formats are aiming to achieve.

Why Training Data Matters for Supervised Learning

Now that we have a high level understanding of what Training Data is, let’s consider why it matters.
[Visual showing Training Data as foundation for AI system]
Training Data is the foundation for successful supervised learning. Machine Learning is about learning
from data. Historically, this meant datasets in the form of logs, or similar tabular data such as “Anthony
viewed a video.”

Dataset
A dataset is like a folder. It usually has the special meaning that there are both “raw” data (such as images) and
annotations in the same place. For example a folder of 100 images plus a text file that lists the annotations.

These systems continue to have significant value - however - they have some limits. They won’t help us
build systems to interpret a computerized tomography (CT) Scan, understand football tactics, or drive a
car. As models require less and less data to be effective, it puts more emphasis on creating application
specific training data.
The idea behind Supervised Learning is generally a human expressly saying “here’s an example of what
a player passing a ball” looks like. “Here’s what a tumor looks like”. “This section of the apple is rotten”.

Control
The question in any system: control.
How Training Data Controls the Model
Where is the control? In normal computer code this is human written logic in the form of loops, if
statements, etc. This logic defines the system.
In Machine Learning I define features of interest and a dataset. The algorithm generates a model which
is effectively the control logic. I exercise control by choosing features.
In a Deep Learning system, the algorithm does its own Feature Selection. The algorithm attempts to
determine what features are relevant to a given goal. That goal is defined by Training Data. In fact,
Training Data is the entire definition of the goal.
Here’s how it works. An internal part of the algorithm, called a loss function, describes a key part of how
the algorithm can learn a good representation of this goal. This is not the goal itself. The algorithm uses
this loss function to determine how close it is to the goal defined in the training data. The training data
is the “ground truth” for correctness of the model’s relationship to the human defined goal.5

Dependencies
In traditional software development there is a degree of dependency between the end user and the
engineering. The end user cannot truly say if the program is “correct”, and neither can the engineer.
Their definitions of correctness may be similar, but are most likely not exactly equal. It’s hard for an end
user to say what they want until a “prototype” of it has been built. Therefore both the end user and
engineer are dependent on each other. This is called a circular dependency. The ability to improve the
software comes from the interplay between both.
With Training Data, the AI Supervisors, control the meaning of the system when doing the literal
supervision. The Data Scientists control it when choosing abstractions such as label templates.
For example, if I as a supervisor were to label a tumor as cancerous when in fact it’s benign, I would be
controlling the output of the system in a detrimental way. In this context, it’s worth understanding there
is no validation possible to ever 100% eliminate this control. Engineering cannot, in a reasonable time
frame, look at all the data.

Historical Aside
There used to be this assumption that Data Science knew what ‘correct’ was. The theory was that they could
define some examples of correct, and then as long as the human supervisors generally stuck to that guide, they
knew what correct was. The problem is, how can an english speaking data scientist know if a translation to french
is correct? How can a data scientist know if a doctor’s medical opinion on an X-Ray image is correct? The short
answer is - they can’t. As the role of AI systems grows subject matter experts increasingly exercise control on the
system that supersedes Data Science.6

To understand why this goes beyond the “garbage in garbage out” phrase. Consider that in a traditional
program, while the end user may not be happy, the engineer can, through a concept called unit tests, at
least guarantee that the code is “correct”.
This is impossible in the context of training data, because the controls available to engineering, such as
a validation set, are still based on the control executed by the individual AI supervisors.

Note: Classic cases where there’s existing data that can’t be edited. Context of changing the underlying data
(unlike say sales statistics that are fixed).

Further, the AI supervisors are generally bound by the control exerted by engineering in defining the
abstractions they are allowed to use. It’s almost as though anything an end user writes, starts to
become part of the fabric of the system itself.
This blurring of the lines between “content” and “system” is important. This is distinctly different from
classic systems. For example, on a social media platform, your content may be the value, but it’s still
clear what is the literal system (the box you type in, the results you see, etc) and the content you post
(text, pictures, etc).
While this entire book is about the concepts around (control) of training data - it’s worth understanding
that:
Training Data abstractions define Data Sciences control Not just algorithm selection
Training Data literals define Supervisors control. Their control can supersede Data Science

Context Matter: Imagine a Perfect System

Let’s imagine for a moment we have a Deep Learning algorithm that is perfect. For any given examples
reasonably similar to our training set, say traffic lights, it will automatically detect said traffic lights
100% of the time, without failure. Is it perfect?
Unfortunately, our perfect system is not really perfect.
Because after celebrating our victory at detecting traffic lights - we realize we not only want to detect
the traffic light, but also if it’s red, red left, green, green left etc…
The system is perfect at detecting the abstractions we defined. Therefore the abstractions matter as
much as the accuracy of the detections. It’s worth pausing here and considering - even if the algorithm
is perfect - there’s still a need to understand the training data.
Continuing this example, we go back and update our Training Data with the new classes (red, green
etc). And again we hit a problem. We realize that sometimes the light is occluded. Now we must train
for occlusion. Oops and we forgot night time, and snow, and the list goes on.

Contexts in Training Data: Classic and Supervised

Classic and Supervised are the two major complementary camps within Training Data. Supervised has
recently been in the limelight, in part because there are more degrees of freedom. The most key
difference is exactly that - in the classic context there is only indirect human control, where as in the
supervised context there is direct human control. This is not to diminish the continued importance of
classic training data. It’s best to think of them as different tools covering different problems rather than
competing approaches.
 Figure 1-6. Classic and new approaches

Discovery
Training Data classically has been about discovery of new insights and useful heuristics. The starting
point is often text, tabular and time series data. This data is usually used as a form of discovery, such as
for recommender systems (Netflix suggested movies), anomaly detection, and car reconditioning costs.
Crucially there is no form of human “supervision”. In the modern deep learning context, there may not
even be feature engineering. To slightly oversimplify, the data is fixed, a very high level goal is defined,
and the algorithm goes to work.

Feature engineering
A practice of selecting specific columns (or slices) that are more relevant, for example the odometer column for a
vehicle reconditioning cost predictor
Monkey See, Monkey Do
With Supervised, we already know what the correct answer is, and the goal is to essentially copy that
understanding. This direct control makes supervised learning applicable to a new variety of use cases,
especially in the realm of “dense” data, such as video, images, and audio. Instead of the data being
fixed, we can even control generating new data, such as taking new images.
We will cover a refresher on the Classic context and in-depth comparison of how it relates to this new
Supervised context.

Training Data Sample Creation

Let’s explore, from the ground up, how to create a single sample of training data. This will help build
understanding of the core mechanics of what the literal supervision looks like.

Introduction
Imagine we are building an autonomous system, such as a traffic light detection system.
The system will have a deep learning model that has been trained on a set of training data.
This training data consists of:
Raw images (or video)
Labels

Here we will discuss a few different approaches and the appropriate training data.

Approach One: Binary Classification

As an example, two of the images in the set may look like this:

Figure 2 TK
To Supervise Example One, we need only two things:
1. To capture the relation to the file itself. E.g., that it’s a “sensor_front_2020_10_10_01_000.”
This is the “link” to the raw pixels. Eventually this file will be read, and the values converted into
 tensors eg position 0,0 having RGB values.
2. To declare what it is in a meaningful way to us. “Traffic_light” or `1`

And for Example Two: we could declare it as:

1. “sensor_front_2020_10_10_01_001”
2. “No_Traffic_light_not_present” or `0`
That’s it. While there is research on ‘zero shot’ and ‘one shot’ learning, in general we will use a set. So
for example we would have a list of three images, and three3 corresponding arrays detailing A and B

Let’s manually create our first set

You can do this with a pen and paper or white board. First draw a big box and call it “My First Set”.
Then Idraw a smaller box, and put a sketch of a traffic light and the number 1 inside it. Repeat that 2
more times drawing an image without a traffic light and a 0.

This is the core mapping idea. It can be done on pen and paper and can also be done in code.
Realistically we will need proper tools to create production training data, but from a conceptual
standpoint this is equally correct.
For example, consider this python code. Here we create a list and populate it with lists where the 0th
index is the file path and the 1st index is the ID. The completed result (assuming the literal bytes of the
.jpgs are available in the same folder) is a set of training data.7

Training_Data = [
[‘tmp/sensor_front_2020_10_10_01_000.jpg’,
1],
[‘tmp/sensor...001.jpg’, 0],
[‘tmp/sensor...002.jpg’, 0]]

This is missing a label map (what does 0 mean?). We can represent this as simple dictionary as:

Label_map = {
0 : “Traffic_light”,
1 : “No Traffic light”
}

Congrats! You have just created a Training Data Set, from scratch, with no tooling, and minimal effort!
With a little wrangling, this would be a completely valid starting point for a basic classification algorithm.
You can also see how, by simply adding more items to the list, you can increase the training data set.
While “real” sets are typically larger, and typically have more complex annotations, this is a great start!

What are we doing?

Let’s unpack the human algorithm we are using.
Look at picture
Draw on knowledge of traffic lights
Map our knowledge of of traffic lights to sparse value eg “Traffic light is present”

Making it clear to the Machine

While it’s obvious to us - it’s not so obvious to a computer. For example, prior approaches in this space
such as Histogram of Oriented Gradients, and other edge detection mechanisms, have no real
understanding of “traffic light” and “not traffic light.”

Not so over simplified

It’s true that modern self-driving teams use more advanced approaches (that we will describe later).
For the sake of this example, we imagine the Traffic Lights are pre-cropped by some process. We may
make assumptions about the angle of the traffic lights. If we have known maps etc. this may be quite
reasonable and reliable.
Ultimately, other approaches typically build on classification, or add “spatial” properties to it. At the end
of the day, even if some prior process runs, there’s still a classification process being done.8
Approach Two: Upgraded Classification

Why are we using strings or integers?

We will introduce the concept of the Label Map
Generally speaking, most actual training will use integer values. However, those integer values typically
are only meaningful to us if attached to some kind of string label:
{ 0 : “None”,
1 : “Red”,
2: “Green”}
While mapping of this type is common to all systems, these label maps can take on additional
complexity. It’s also worth remembering that in general, the ‘label’ means nothing to the system. It’s
mapping the ID to the raw data. If these values are wrong it could cause a critical failure. Worse -
imagine if the testing relied on the same map!
That’s why when possible it’s good to “print output” eg that you can visually inspect the label matches
the desired ID. (specifically in regard to train/val/test) [Technical] As a test case, could also Assert on a
known ID matching String. [TODO move this to a different section maybe. TODO add a more general
comment about where the testing concepts are introduced as this isn’t really the main point of testing.]

Supervision vs Annotation
Annotation is a popular phrasing for anything to do with training data. Annotation generally implies adding or
drawing information - without regard to any concept of a system. In words to annotate is generally a “secondary”
action. This masks the importance and the context of the work being done. Supervision more accurately reflects
the overall scope and context of the work. It also better reflects the increasingly common context of correcting
(supervising!) an existing model or system.

Where is the Traffic Light?—Objectness score

The problem with the above approach is that we don’t know where the traffic light is. There is a
common concept called an objectness score, and other more complex ways to identify location. We will
cover location concepts in more detail.
Training Data Process Introduction
Now that we have covered the basics of how a single sample is created, and introduced some key
terms, let’s take a high-level look at the process.

Getting Started
This process will require several stages.

Raw Data and Tasks

Training data starts with identifying and capturing raw data. The next step is to design the Tasks, chiefly
the Labels and Attributes.

Train Model and Review Results

As soon as a minimal dataset is constructed it’s good to start training a model. This will give us clues to
help better design the Tasks.

Training Data Actions

These are actions that can be taken generally after some form of information from the model training
process.

Change the Labels and Attributes

This is one of the most common approaches. One example is to divide and conquer the label classes,
especially poorly performing classes. Essentially this is to both identify and improve the specific classes
that are the weakest. Here we have say Traffic Light. Performance is mixed. It’s unclear which examples
are needed to improve performance. When reviewing the results, we notice that Green seems to show
up more often in the failure cases. One option is to try to add more green ones to the general traffic
light set. A better option is to “split” the class Traffic Light into “Red” and “Green”. That way we can
very clearly see which is performing better. We can repeat this until the desired performance is reached.
For example, again splitting between large and small. There are a few intricacies and approaches to
implementation of this but they generally revolve around this idea.

Change the spatial type of instance

Imagine you started with choosing segmentation. Then realize the model is not training as desired. You
may be able to simply switch to an “easier” task like object detection, or even full image classification.
Alternatively perhaps object detection is yielding a bunch of overlapping boxes which aren’t helpful and
you need to switch to segmentation to accurately capture the meaning.
At the time of writing there are nearly a dozen popular spatial methods. While it may appear clear which
methods are less ideal for a certain case the optimal method is often less clear. This is also a bit of a
moving target as annotation tooling and model training methods change.
Create More Tasks
Annotating more data for better performance has almost become a cliche already. This is often
combined with the other approaches. For example, dividing the labels, or changing the spatial type, and
then supervising more. The primary consideration here is if more annotation will provide net lift.
Net Lift Introduction

For those with technical knowledge let’s first dispel a notion - this is not about balancing the dataset.
Try to forget the concept of balancing while considering this.
To illustrate the need for net lift consider a raw, unlabeled dataset, in which 10% is labeled. [Shown Fig
as Dataset]. As a baseline approach we will random sample data 3 points (10/30/80%). At each point
we will look at model performance. If the performance is unchanged we will stop.
By chance we draw all hearts each time. Each additional heart we supervise provides minimal value -
since we already have seen many hearts. Further, we don’t really understand the complete production
picture because we did not encounter circles or triangles.
Two different things here:
1. The idea of identifying previously unknown cases
2. The idea of wanting to maximize the value of each net annotation.

Change the Raw Data

Change the sensor angle
Change what part of the screen is being capture

Split the models / heads

A model may need less data for one class then another - sometimes by multiple orders of magnitude

Levels of System Maturity of Training Data Operations

The above described process is the overarching strategy. We will also zoom into the more tactical
concerns of operating and executing that process. This is Machine Learning Operations ( MLOps ) for
supervised data.
You will learn about the 5 major stages of operations, Data Prep, Tasks, Literal Human Control,
Datasets, and Export. Both the specific operational concerns of those stages and the 3 major levels of
system maturity for each. From early exploring, to proof of concept, through to production.

Training Data in the Ecosystem

Training Data sits in between Raw Data, such as sensors, and Modeling (Training and Prediction).
For example, you may create Training Data with one tool, and Train with a different tool. I will touch on
a high level map of adjacent areas, tools, and popular integration points.
Tooling
There are tools designed specifically for Training Data, such as Diffgram. I will talk about open source
and commercial options here. Expressly highlighting trade-offs of popular tools. While generally I will
aim to stay tooling agnostic I will ground some of the examples in specific tools.

Applied vs Research Sets

The modern needs and form of training data continue to rapidly evolve. When people think of datasets
popular research datasets, like MS Common Objects in Context (COCO) 20149, Imagenet 200910, (both
vision) and the General Language Understanding and Evaluation (GLUE) 20181112 sets come to mind.
These sets are designed for research purposes, and by design, they evolve relatively slowly. In general,
it’s the research that’s designed to change around these sets, not the sets themselves. The sets are
used as benchmarks for performance. For example here, the same set is used in 5 different years. This
is useful for research comparisons. The core assumption is the sets are static.
In the context of a practical commercial product, things change. In general the rule of thumb is that
data more than 12 months old is unlikely to accurately represent the current state. Generally the
assumption for practical sets is that they are only static for the moment of literal training, but are
otherwise always changing. Further commercial needs are very different from research. Time to create
the sets, time to update, costs to create the sets, implications of mistakes, etc. are very different from a
research context.
Discovering Diverse Content Through
Random Scribd Documents
looked cross, as if he were in a dreadful hurry and hated to be
bothered. Altogether things were changed for the worse, and
Josephine’s heart would perhaps have broken if it hadn’t been for
the dear companionship of Rudanthy, who smiled and slept in a
placid waxen manner that was restfully familiar.
Besides, all journeys have an end; and the six days’ trip of the little
San Diegan came to its own before the door of a stately mansion,
gay with the red brick and white marble which mark most Baltimore
homes, and the ring of an electric bell that the expressman touched:
“A ‘parcel’ for Joseph Smith. Billed from San Diego, Cal. Live here,
eh?”
It was a colored man in livery who replied:
“Yes, suh. Mister Joseph Smith, he done live here, suh.”
“Sign, please. That is, if you can write.”
“Course I can write. I allays signs parcels for Mister Smith, suh.
Where’s the parcel at, suh?” returned the liveried negro.
“Sign. I’ll fetch it,” came the prompt answer.
Old Peter signed, being the trusted and trustworthy servant of his
master, and returned the book to the agent’s hands, who himself
returned to the carriage, lifted out Josephine and Rudanthy,
conveyed them up the glistening steps, and left them to their fate.
 CHAPTER III.
ARRIVAL.

Peter stared, but said nothing. Not even when the agent ran back
from the carriage with a little satchel and a strap full of shawls and
picture-books. The hack rolled away, the keen March wind chilled the
young Californian, who stood, doll in hand, respectfully waiting
admission to the warm hall beyond the door. Finally, since the
servant seemed to have been stricken speechless, she found her
own voice, and said:
“Please, boy, I’d like to see my Uncle Joe.”
“Your—Uncle—Joe, little miss?”
“That’s what I said. I must come in. I’m very cold. If this is
Baltimore, that the folks on the cars said was pretty, I guess they
didn’t know what they were talking about. I want to come in,
please.”
The old man found his wits returning. This was the queerest “parcel”
for which he had ever signed a receipt in an express-book, and he
knew there was some mistake. Yet he couldn’t withstand the
pleading brown eyes under the scarlet hat, even if he hadn’t been
“raised” to a habit of hospitality.
“Suah, little lady. Come right in. ’Tis dreadful cold out to-day. I ’most
froze goin’ to market, an’ I’se right down ashamed of myself leavin’
comp’ny waitin’ this way. Step right in the drawin’-room, little missy,
and tell me who ’tis you’d like to see.”
Picking up the luggage that had been deposited on the topmost of
the gleaming marble steps, which, even in winter, unlike his
neighbors, the master of the house disdained to hide beneath a
wooden casing, the negro led the way into the luxurious parlor. To
Josephine, fresh from the chill of the cloudy, windy day without, the
whole place seemed aglow. A rosy light came through the red-
curtained windows, shone from the open grate, repeated itself in the
deep crimson carpet that was so delightfully soft and warm.
“Sit down by the fire, little lady. There. That’s nice. Put your dolly
right here. Maybe she’s cold, too. Now, then, suah you’se fixed so
fine you can tell me who ’tis you’ve come to see,” said the man.
“What is your name, boy?” inquired Josephine.
“Peter, missy. My name’s Peter.”
“Well, then, Peter, don’t be stupid. Or are you deaf, maybe?” she
asked.
“Land, no, missy. I’se got my hearin’ fust class,” he replied,
somewhat indignantly.
“I have come to see my Uncle Joe. I wish to see him now. Please tell
him,” she commanded.
The negro scratched his gray wool and reflected. He had been born
and raised in the service of the family where he still “officiated,” and
knew its history thoroughly. His present master was the only son of
an only son, and there had never been a daughter. No, nor wife, at
least to this household. There were cousins in plenty, with whom Mr.
Joseph Smith was not on good terms. There were property interests
dividing them, and Mr. Joseph kept his vast wealth for his own use
alone. Some thought he should have shared it with others, but he
did not so think and lived his quiet life, with a trio of colored men-
servants. His house was one of the best appointed on the wide
avenue, but, also, one of the quietest. It was the first time that old
Peter had ever heard a child’s voice in that great room, and its clear
tones seemed to confuse him.
“I want to see my Uncle Joe. I want to see him right away. Go, boy,
and call him,” Josephine explained.
This was command, and Peter was used to obey, so he replied:
“All right, little missy, I’ll go see. Has you got your card? Who shall I
say ’tis?”
Josephine reflected. Once mamma had had some dear little visiting
cards engraved with her small daughter’s name, and the child
remembered with regret that if they had been packed with her
“things” at all, it must have been in the trunk, which the expressman
said would arrive by and by from the railway station. She could
merely say:
“Uncles don’t need cards when their folks come to see them. I’ve
come from mamma. She’s gone to the pickley land to see papa. Just
tell him Josephine. What’s that stuff out there?”
She ran to the window, pulled the lace curtains apart, and peered
out. The air was now full of great white flakes that whirled and
skurried about as if in the wildest sort of play.
“What is it, Peter? Quick, what is it?” she demanded.
“Huh! Don’t you know snow when you see it, little missy? Where you
lived at all your born days?” he cried, surprised.
“Oh, just snow. Course I’ve seen it, coming here on the cars. It was
on the ground, though, not in the air and the sky. I’ve lived with
mamma. Now I’ve come to live with Uncle Joe. Why don’t you tell
him? If a lady called to see my mamma do you s’pose big Bridget
wouldn’t say so?”
“I’se goin’,” he said, and went.
But he was gone so long, and the expected uncle was so slow to
welcome her, that even that beautiful room began to look dismal to
the little stranger. The violent storm which had sprung up with such
suddenness, darkened the air, and a terrible homesickness
threatened to bring on a burst of tears. Then, all at once, Josephine
remembered what Doctor Mack had said:
“Don’t be a weeper, little lady, whatever else you are. Be a smiler,
like my Cousin Helen, your mamma. You’re pretty small to tackle the
world alone, but just do it with a laugh and it will laugh back upon
you.”
Not all of which she understood, though she recalled every one of
the impressive words, but the “laughing part” was plain enough.
“Course, Rudanthy. No Uncle Joe would be glad to get a crying little
girl to his house. I’ll take off my coat and yours, darling. You are
pretty tired, I guess. I wonder where they’ll let us sleep, that black
boy and my uncle. I hope the room will have a pretty fire in it, like
this one. Don’t you?”
Rudanthy did not answer, but as Josephine laid her flat upon the
carpet, to remove her travelling cloak, she immediately closed her
waxen lids, and her little mother took this for assent.
“Oh, you sweetest thing! How I do love you!”
There followed a close hug of the faithful doll, which was witnessed
by a trio of colored men from a rear door, where they stood, open-
eyed and mouthed, wondering what in the world the master would
say when he returned and found this little trespasser upon his
hearth-stone.
When Rudanthy had been embraced, to the detriment of her jute
ringlets and her mistress’ comfort, Josephine curled down on the rug
before the grate to put the doll asleep, observing:
“You’re so cold, Rudanthy. Colder than I am, even. Your precious
hands are like ice. You must lie right here close to the fire, ’tween
me and it. By-and-by Uncle Joe will come and then—My! Won’t he
be surprised? That Peter boy is so dreadful stupid, like’s not he’ll
forget to say a single word about us. Never mind. He’s my papa’s
twin brother. Do you know what twins are, Rudanthy? I do. Big
Bridget’s sister’s got a pair of them. They’re two of a kind, though
sometimes one of them is the other kind. I mean, you know,
sometimes one twin isn’t a brother, it’s a sister. That’s what big
Bridget’s sister’s was. Oh, dear. I’m tired. I’m hungry. I liked it better
on that nice first railway car where everybody took care of me and
gave me sweeties. It’s terrible still here. I—I’m afraid I’m going to
sleep.”
In another moment the fear of the weary little traveller had become
a fact. Rudanthy was already slumbering; and, alas! that was to
prove the last of her many naps. But Josephine was unconscious of
the grief awaiting her own awakening; and, fortunately, too young to
know what a different welcome should have been accorded herself
by the relative she had come so far to visit.
Peter peeped in, from time to time, found all peaceful, and retired in
thankfulness for the temporary lull. He was trembling in his shoes
against the hour when the master should return and find him so
unfaithful to his trust as to have admitted that curly-haired intruder
upon their dignified privacy. Yet he encouraged himself with the
reflection:
“Well, no need crossin’ no bridges till you meet up with ’em, and this
bridge ain’t a crossin’ till Massa Joe’s key turns in that lock. Reckon I
was guided to pick out that fine duck for dinner this night, I do.
S’posin’, now, the market had been poor? Huh! Every trouble sets
better on a full stummick ’an a empty. Massa Joe’s powerful fond of
duck, lessen it’s spoiled in the cookin’. I’ll go warn that ’Pollo to be
mighty careful it done to a turn.”
Peter departed kitchen ward, where he tarried gossipping over the
small guest above stairs and the probable outcome of her advent.
“Nobody what’s a Christian goin’ to turn a little gell outen their doors
such an evenin’ as this,” said Apollo, deftly basting the fowl in the
pan.
 “I’M JOSEPHINE!”
“Mebbe not, mebbe not. But I reckon we can’t, none of us, callate
on whatever Massa Joe’s goin’ to do about anything till he does it.
He’s off to a board meeting, this evening, and I hope he sets on it
comfortable. When them boards are too hard, like, he comes home
mighty ’rascible. Keep a right smart watch on that bird, ’Pollo, won’t
you? whiles I go lay the table.”
But here another question arose to puzzle the old man. Should he,
or should he not, prepare that table for the unexpected guest? There
was nobody more particular than Mr. Smith that all his orders should
be obeyed to the letter. Each evening he wished his dinner to be
served after one prescribed fashion, and any infraction of his rules
brought a reprimand to Peter.
However, in this case he determined to risk a little for hospitality’s
sake, reflecting that if the master were displeased he could whisk off
the extra plate before it was discovered.
“Massa Joe’s just as like to scold if I don’t put it on as if I do. Never
allays account for what’ll please him best. Depends on how he takes
it.”
Busy in his dining-room he did not hear the cab roll over the snowy
street and stop at the door, nor the turn of the key in the lock. Nor,
lost in his own thoughts, did the master of the house summon a
servant to help him off with his coat and overshoes. He repaired
immediately to his library, arranged a few papers, went to his
dressing-room and attired himself for dinner, with the carefulness to
which he had been trained from childhood, and afterward strolled
leisurely toward the great parlor, turned on the electric light, and
paused upon its threshold amazed, exclaiming:
“What is this? What in the world is—this?”
The sudden radiance which touched her eyelids, rather than his
startled exclamation, roused small Josephine from her restful nap.
She sat up, rubbed her eyes, which brightened with a radiance
beyond that of electricity, and sprang to her feet. With outstretched
arms she flung herself upon the astonished gentleman, crying:
“Oh, you beautiful, beautiful man! You darling, precious Uncle Joe!
I’m Josephine! I’ve come!”
 CHAPTER IV.
A MULTITUDE OF JOSEPHS.

“So I perceive!” responded the master of the house, when he could

rally from this onslaught of affection. “I’m sure I’m very pleased to
welcome you. I—when—how did you arrive?”
“I’m a ’xpress ‘parcel,’” she answered, laughing, for she had learned
before this that she had made her long journey in rather an unusual
fashion. “Mamma had to go away on the peacock-blue ocean; and
Doctor Mack couldn’t bother with me, ’cause he’s going to the folks
that eat almonds together and give presents; and there wasn’t
anybody else ’xcept big Bridget, and she’d spent all her money, and
mamma said you wouldn’t want a ‘wild Irish girl’ to plague you.
Would you?”
“I’m not fond of being plagued by anybody,” said the gentleman,
rather dryly. He was puzzled as much by her odd talk as her
unexpected appearance, and wondered if children so young were
ever lunatics. The better to consider the matter he sat down in the
nearest chair, and instantly Josephine was upon his knee. The
sensation this gave him was most peculiar. He didn’t remember that
he had ever taken any child on his lap, yet permitted this one to
remain there, because he didn’t know what better to do. He had
heard that one should treat a lunatic as if all vagaries were real.
Opposition only made an insane person worse. What worse could
this little crazy creature, with the lovely face and dreadful manners,
do to a finical old bachelor in evening clothes than crush the creases
out of his trouser knees?
The lap was not as comfortable as Doctor Mack’s, and far, far from
as cosey as mamma’s. Uncle Joe’s long legs had a downward slant to
them that made Josephine’s perch upon them rather uncertain. After
sliding toward the floor once or twice, and hitching up again, she
slipped to her feet and leaned affectionately against his shoulder,
saying:
“That’s better. I guess you’re not used to holding little girls, are you,
Uncle Joe?”
“No, Josephine. What is your other name?” said he.
“Smith. Just like yours. You’re my papa’s dear twin, you know.”
“Oh, am I?” he asked.
“Course. Didn’t you know that? How funny. That’s because you
haven’t mamma to remind you, I s’pose. Mamma remembers
everything. Mamma never is naughty. Mamma knows everything.
Mamma is dear, dear, dear. And, oh, I want her, I want her!”
Josephine’s arms went round the gentleman’s neck, and her tears
fell freely upon his spotless shirt-front. She had been very brave, she
had done what she promised Doctor Mack, and kept a “laughing
front” as long as she could; but now here, in the home of her papa’s
twin, with her “own folks,” her self-control gave way, and she cried
as she had never cried before in all her short and happy life.
Mr. Smith was hopelessly distressed. He didn’t know what to say or
do, and this proved most fortunate for both of them. For whatever
he might have said would have puzzled his visitor as greatly as she
was puzzling him. Happily for both, the deluge of tears was soon
over, and Josephine lifted a face on which the smiles seemed all the
brighter because of the moisture that still bedewed it.
“Please ’xcuse me, Uncle Joe. I didn’t mean to cry once, but it—it’s
so lovely to have you at last. It was a long, long way on the railway,
uncle. Rudanthy got terribly tired,” explained the visitor.
“Did she? Who is Rudanthy?”
“You, my uncle, yet don’t know Rudanthy, that has been mine ever
since I was? Mamma says she has to change heads now and then,
and once in awhile she buys her a new pair of feet or hands; but it’s
the same darling dolly, whether her head’s new or old. I’ll fetch her.
It’s time she waked up, anyway.”
Josephine sped to the rug before the grate, stooped to lift her
playmate, paused, and uttered a terrified cry.
“Uncle! Uncle Joe, come here quick—quick!”
Smiling at his own acquiescence, the gentleman obeyed her
demand, and stooped over her as she also bent above the object on
the rug. All that was left of poor Rudanthy—who had travelled three
thousand miles to be melted into a shapeless mass before the first
hearth-fire which received her.
Josephine did not cry now. This was a trouble too deep for tears.
“What ails her, Uncle Joe? I never, never saw her look like that. Her
nose and her lips and her checks are all flattened out, and her eyes
—her eyes are just round glass balls. Her lovely curls”— The little
hands flew to the top of the speaker’s own head, but found no
change there. Yet she looked up rather anxiously into the face above
her. “Do you s’pose I’d have got to look that dreadful way if I hadn’t
waked up when I did, Uncle Joe?”
“No, Josephine. No, indeed. Your unhappy Rudanthy was a waxen
young person who was indiscreet enough to lie down before an open
fire. You seem to be real flesh and blood, and might easily scorch,
yet would hardly melt. Next time you take a nap, however, I’d advise
you to lie on a lounge or a bed.”
“I will. I wouldn’t like to look like her. But what shall I do? I don’t
know a store here,” she wailed.
“I do. I might be able to find you a new doll, if you won’t cry,” came
the answer which surprised himself.
“Oh, I shan’t cry any more. Never any more—if I can help it. That’s a
promise. But I shouldn’t want a new doll. I only want a head. Poor
Rudanthy! Do you s’pose she suffered much?” was the next anxious
question.
“It’s not likely. But let Rudanthy lie yonder on the cool window sill. I
want to talk with you. I want you to answer a few questions. Sit
down by me, please. Is this comfortable?”
Josephine sank into the midst of the cushions he piled for her on the
wide sofa and sighed luxuriously, answering:
“It’s lovely. This is the nicest place I ever, ever saw.”
“Thank you. Now, child, tell me something about other places you
remember, and, also, please tell me your name.”
Josephine was surprised. What a very short memory this uncle had,
to be sure. It wouldn’t be polite to say so, though, and it was an
easy question to answer.
“My name is Josephine Smith. I’m named after you, you know,
’cause you’re my papa’s twin. I’m sent to you because”—and she
went on to explain the reasons, so far as she understood them, of
her long journey and her presence in his house. She brought her
coat and showed him, neatly sewed inside its flap, a square of
glazed holland on which was written her name, to whom consigned,
and the express company by which she had been “specially shipped
and delivered.”
It was all plain and straightforward. This was the very house
designated on the tag, and he was Joseph Smith; but it was, also, a
riddle too deep for him to guess.
“I see, I see. Well, since you are here we must make the best of it. I
think there’s a mistake, but I dare say the morning will set it all
right. Meanwhile, it’s snowing too fast to make any inquiries to-
night. It is about dinner time, for me. Have you had your dinner?”
asked the host.
“I had one on the train. That seems a great while ago,” said the
guest.
“I beg pardon, but I think there is a little smut upon your pretty
nose. After a railway journey travellers usually like to wash up, and
so on. I don’t know much about little girls, yet”—he rather timidly
suggested.
“I should be so glad. Just see my hands, Uncle Joe!” and she
extended a pair of plump palms which sadly needed soap and water.
“I’m not your”—he began, meaning to set her right concerning their
relationship; then thought better of it. What would a child do who
had come to visit an unknown uncle and found herself in the home
of a stranger? Weep, most likely. He didn’t want that. He’d had
enough of tears, as witness one spoiled shirt-front. He began also to
change his mind regarding the little one’s manners. She had
evidently lived with gentlefolks and when some one came to claim
her in the morning he would wish them to understand that she had
been treated courteously.
So he rang for Peter, who appeared as suddenly as if he had come
from the hall without.
“Been listening at the doorway, boy? Take care. Go up to the guest
room, turn on the heat and light, and see that there are plenty of
fresh towels. Take this young lady’s things with you. She will
probably spend the night here. I hope you have a decent dinner
provided.”
“Fine, Massa Joe. Just supreme. Yes, suh. Certainly, suh,” answered
the servant.
“Uncle Joe, is there a bathroom in this house?” asked she.
“Three of them, Josephine.”
“May I use one? I haven’t had a bath since I was in San Diego, and
I’m—mamma would not allow me at table, I guess; I’m dreadful
dirty.”
If Josephine had tried to find the shortest way to Mr. Smith’s heart
she could not have chosen more wisely.
“To be sure, to be sure. Peter, make a bath ready next the guest
room. Will an hour give you time enough, little lady?”
“I don’t want so long. I’m so glad I learned to dress myself, aren’t
you? ’Cause all the women to this house seem to be men, don’t
they?”
“Yes, child. Poor, unfortunate house!”
“It’s a beautiful house, Uncle Joe; and you needn’t care any more.
I’ve come, now. I, Josephine. I’ll take care of you. Good-by. When
you see me again I’ll be looking lovely, ’cause I’ll put on the new
white wool dress that mamma embroidered with forget-me-nots.”
“Vanity!” thought Mr. Smith, regretfully, which shows that he didn’t
as yet understand his little visitor, whose “lovely” referred to her
clothes alone, and not at all to herself.
The dinner hour at 1000 Bismarck Avenue was precisely half-past
six. Even for the most notable of the few guests entertained by the
master of the house he rarely delayed more than five minutes, and
on no occasion had it been served a moment earlier. The old-
fashioned hall clock had ticked the hour for generations of Smiths
“from Virginia,” and was regulated nowadays by the tower timepiece
at Mt. Royal station. It was fortunate for Josephine that just as the
minute hand dropped to its place, midway between the six and
seven on the dial, she came tripping down the wide stair, radiant
from her bath and the comfort of fresh clothing, and eager to be
again with the handsome Uncle Joe, who was waiting for her at the
stair’s foot with some impatience.
Her promptness pleased him, and the uncommon vision of her
childish loveliness pleased him even more. He had believed that he
disliked children, but was now inclined to change his opinion.
“I’m glad you are punctual, Miss Josephine, else I’d have had to
begin my dinner without you. I never put back meals for anybody,”
he remarked.
“Would you? Don’t you? Then I’m glad, too. Isn’t the frock pretty?
My mamma worked all these flowers with her own little white hands.
I love it. I had to kiss them before I could put it on,” she said, again
lifting her skirt and touching it with her lips.
“I suppose you love your mamma very dearly. What is she like?”
He was leading her along the hall toward the dining-room, and Peter,
standing within its entrance, congratulated himself that he had laid
the table for two. He glanced at his master’s face, found it good-
natured and interested, and took his own cue therefrom.
“She is like—she is like the most beautiful thing in the world, dear
Uncle Joe. Don’t you remember?” asked the astonished child.
“Well, no, not exactly.”
“That’s a pity, and you my papa’s twin. Papa hasn’t nice gray hair
like yours, though, and there isn’t any shiny bare place on top of his
head. I mean there wasn’t when he went away last year. His hair
was dark, like mamma’s, and his mustache was brown and curly. I
think he isn’t as big as you, Uncle Joe, and his clothes are gray, with
buttony fixings on them. He has a beautiful sash around his waist,
sometimes, and lovely shoulder trimmings. He’s an officer, my papa
is, in Company F. That’s for ’musement, mamma says. For the
business, he’s a ’lectrickeller. Is this my place? Thank you, Peter.”
Mr. Smith handed his little visitor to her chair, which the old butler
had pulled back for her, with the same courtly manner he would
have shown the pastor’s wife. Indeed, if he had been asked he
would have admitted that he found the present guest the more
interesting of the two.
Peter made ready to serve the soup, but a look from the strange
child restrained him. She added a word to the look:
“Why, boy, you forgot. Uncle Joe hasn’t said the grace yet.”
Now, Mr. Smith was a faithful and devout church member, but was in
the habit of omitting this little ceremony at his solitary meals. He
was disconcerted for the moment, but presently bowed his head and
repeated the formula to which he had been accustomed in his youth.
It proved to be the same that the little girl was used to hearing from
her own parents’ lips, and she believed it to be the ordinary habit of
every household. She did not dream that she had instituted a new
order of things, and unfolded her napkin with a smile, saying:
“Now, I’m dreadful hungry, Uncle Joe. Are you?”
“I believe I am, little one.”
Peter served with much dignity and flourish; but Josephine had
dined at hotel tables often enough to accept his attentions as a
matter of course. Her quiet behavior, her daintiness, and her chatter,
amused and delighted her host. He found himself in a much better
humor than when he returned through the storm from an
unsatisfactory board meeting, and was grateful for the mischance
which had brought him such pleasant company.
As for old Peter, his dark face glowed with enthusiasm. He was
deeply religious, and now believed that this unknown child had been
sent by heaven itself to gladden their big, empty house. He didn’t
understand how his master could be “uncle” to anybody, yet, since
that master accepted the fact so genially, he was only too glad to do
likewise.
It was a fine and stately dinner, and as course after course was
served, Josephine’s wonder grew, till she had to inquire:
“Is it like this always, to your home, Uncle Joe?”
“What do you mean?” he asked.
“Such a birthday table, and no folks, ’xcept you and me.”
“It is the same, usually, unless Peter fails to find a good market.
Have you finished? No more cream or cake?” he explained and
questioned.
“No, thank you. I’m never asked to take two helpings. Only on the
car I had three, sometimes, though I didn’t eat them. Mamma
wouldn’t have liked it.”
“And do you always remember what ‘mamma’ wishes?”
“No. I’m a terrible forgetter. But I try. Somehow it’s easier now I
can’t see her,” she answered.
“Quite natural. Suppose we go into the library for a little while. I
want to consult the directory.”
She clasped his hand, looked up confidingly, but felt as if she should
fall asleep on the way thither. She wondered if it ever came bedtime
in that house, and how many hours had passed since she entered it.
“There, Miss Josephine, I think you’ll find that chair a comfortable
one,” said the host, when they had reached the library, rich with all
that is desirable in such a room. “Do you like pictures?”
“Oh, I love them!”
“That’s good. So do I. I’ll get you some.”
But Mr. Smith was not used to the “loves” of little girls, and his
selection was made rather because he wanted to see how she would
handle a book than because he thought about the subject chosen. A
volume of Dore’s grotesque drawings happened to be in most
shabby condition, and he reflected that she “couldn’t hurt that
much, anyway, for it’s to be rebound.”
Afterward he opened the directory for himself, and Josephine
thought it a dull-looking book. For some time both were interested
and silent; then Uncle Joe cried out with startling suddenness:
“Three thousand Smiths in this little city; and seventy-five of them
are Josephs! Well, my child, you’re mighty rich in ‘uncles’!”
 CHAPTER V.
A WILD MARCH MORNING.

Josephine was half-asleep. A woman would have thought about her

fatigue and sent her early to bed. “Uncle Joe” thought of nothing
now save the array of common and uncommon names in the city
directory. He counted and recounted the “Smiths,” “Smyths,” and
“Smythes,” and jotted down his figures in a notebook. He copied,
also, any address of any Smith whose residence was in a locality
which he considered suitable for relatives of his small guest. He
became so absorbed in this study that an hour had passed before he
remembered her, and the extraordinary quiet of her lively tongue.
Josephine had dozed and waked, dozed and waked, and dreamed
many dreams during that hour of silence. Her tired little brain was all
confused with the weird pictures of tortured men gazing at her from
the trunks of gnarled trees, and thoughts of a myriad of uncles, each
wearing eyeglasses, and sitting with glistening bald head beneath a
brilliant light. The light dazzled her, the dreams terrified her, and the
little face that dropped at length upon the open page of the great
folio was drawn and distressed.
“For goodness sake! I suppose she’s sleepy. I believe that children
do go to bed early. At least they should. If I’m to be a correct sort of
‘uncle,’ even for one night, I must get her there. I wonder how!”
considered the gentleman.
The first thing was to wake her, and he attempted it, saying:
“Josephine! Josephine!”
The child stirred uneasily, but slumbered on.
“Uncle Joe” laid his hand upon her shoulder rather gingerly. He was
much more afraid of her than she could ever be of him.
“Miss Josephine! If you please, wake up.”
She responded with a suddenness that startled him.
“Why—where am I? Oh! I know. Did I go to sleep, Uncle Joe?”
“I should judge that you did. Would you like to go to bed?”
“If you please, uncle.”
He smiled faintly at the odd situation in which he found himself,
playing nurse to a little girl. A boy would have been less
disconcerting, for he had been a boy himself, once, and remembered
his childhood. But he had never been a little girl, had never lived in a
house with a little girl, and didn’t know how little girls expected to be
treated. He volunteered one question:
“If somebody takes you to your room, could you—could you do the
rest for yourself, Josephine?”
“Why, course. I began when I was eight years old. That was my last
birthday that ever was. Big Bridget was not to wait on me any more
after that, mamma said. But she did. She loved it. Mamma, even,
loved it, too. And nobody need go upstairs with me. I know the way.
I remember it all. If— May I say my prayers by you, Uncle Joe?
Mamma”—
One glance about the strange room, one thought of the absent
mother, and the little girl’s lip quivered. Then came a second
thought, and she remembered her promise. She was never to cry
again, if she could help it. By winking very fast and thinking about
other things than mamma and home she would be able to help it.
Before he touched her shoulder to wake her, Mr. Smith had rung for
Peter, who now stood waiting orders in the parting of the portière,
and beheld a sight such as he had never dreamed to see in that
great, lonely house: Josephine kneeling reverently beside his
master’s knee, saying aloud the Lord’s Prayer and the familiar “Now I
lay me.”
Then she rose, flung her arms about the gentleman’s neck, saw the
moisture in his eyes, and asked in surprise:

“NOW I LAY ME.”

“Do you feel bad, Uncle Joe? Aren’t you happy, Uncle Joe? Can’t I
help you, you dear, dear man?”
The “dear” man’s arms went round the little figure, and he drew it
close to his lonely heart with a jealous wish that he might always
keep it there. All at once he felt that he hated that other unknown,
rightful uncle to whom this charming “parcel” belonged, and almost
he wished that no such person might ever be found. Then he
unclasped her clinging arms and—actually kissed her!
“You are helping me very greatly, Josephine. You are a dear child.
Peter will see that your room is all right for the night. Tell him
anything you need and he’ll get it for you. Good-night, little girl.”
“Good-night, Uncle Joe. Dear Uncle Joe. I think—I think you are just
too sweet for words! I hope you’ll rest well. Good-night.”
She vanished through the curtains, looking back and kissing her
finger-tips to him, and smiling trustingly upon him to the last. But
the old man sat long looking after her before he turned again to his
books, reflecting:
“Strange! Only a few hours of a child’s presence in this silent place,
yet it seems transfigured. ‘An angel’s visit,’ maybe. To show me that,
after all, I am something softer and more human than the crusty old
bachelor I thought myself. What would her mother say, that absent,
perfect ‘mamma,’ if she knew into what strange hands her darling
had fallen? Of course, my first duty to-morrow is to hunt up this
mislaid uncle of little Josephine’s and restore her to him. But—Well,
it’s my duty, and of course I shall do it.”
The great bed in the guest room was big enough, Josephine
thought, to have held mamma herself, and even big Bridget without
crowding. It was far softer than her own little white cot in the San
Diegan cottage, and plunged in its great depths the small traveller
instantly fell asleep. She did not hear Peter come in and lower the
light, and knew nothing more, indeed, till morning. Then she roused
with a confused feeling, not quite realizing where she was or what
had happened to her. For a few moments she lay still, expecting
mamma’s or big Bridget’s face to appear beneath the silken curtains
which draped the bed’s head; then she remembered everything, and
that in a house without women she was bound to do all things for
herself.
“But it’s dreadful dark everywhere. I guess I don’t like such thick
curtains as Uncle Joe has. Mamma’s are thin white ones and it’s
always sunshiny at home—’xcept when it isn’t. That’s only when the
rains come, and that’s most always the nicest of all. Then we have a
dear little fire in the grate, and mamma reads to me, and big Bridget
bakes and cooks the best things. We write letters to papa, and
mamma sings and plays, and—it’s just lovely! Never mind,
Josephine. You’ll be back there soon’s papa gets well again, and
Uncle Joe was sort of cryey round his eyes last night. Mamma said I
was to be like his own little daughter to him and take care of him
and never make him any trouble. So I will.”
There was no prouder child in that city that morning than the little
stranger within its gates. She prepared her bath without aid,
brushed her hair and dressed herself entirely. It was true that her
curls did not look much as they did after mamma’s loving fingers had
handled them, and the less said about those on the back of her head
the better. Nor were the buttons in the right places to match the
buttonholes, and the result was that the little frock which had always
been so tidy hung at a curious angle from its wearer’s shoulders.
But who’d mind a trifle like that, in a beginner?
Not Uncle Joe, who saw only the fair front of his visitor, as she ran
down the hall to meet him, emerging from his own chamber. Indeed,
he was not now in a mood to observe anything save himself, though
he answered Josephine’s gay “Good morning” with another rather
grimly spoken.
The child paused, astonished. There were no longer tears in his
eyes, but he looked as if a “good cry” would be relief. His face was
distorted with pain, and every time he put one of his feet to the floor
he winced as if it hurt him. He seemed as dim and glum as the day
outside, and that was dreary beyond anything the little Californian
had ever seen. The snow had fallen steadily all the night, and the
avenue was almost impassable. A few milk-carts forced their way
along, and a man in a gray uniform, with a leather bag over his
shoulder, was wading up each flight of steps to the doorways above
them and handing in the morning mail.
“Aren’t you well, Uncle Joe? Didn’t you rest well?” she inquired
solicitously.
“No, I’ve got that wretched old gout again,” he snapped.
“What’s that?”
“It’s a horrible, useless, nerve-racking ‘misery’ in my foot. It’s being
out in that storm yesterday, and this senseless heap of snow on the
ground. March is supposed to be spring, but this beastly climate
doesn’t know what spring means. Ugh!” he groaned.
“Doesn’t it?” she asked, amazed by this statement.
“Hum, child. There’s no need of your repeating everything I say in
another question. I’m always cross when I’m gouty. Don’t heed me.
Just enjoy yourself the best you can, for I don’t see how I’m to hunt
up your uncle for you in such weather.”
Josephine thought he was talking queerly, but said nothing; only
followed him slowly to the breakfast room, which Peter had done his
best to make cheerful.
Mr. Smith sat down at table and began to open the pile of letters
which lay beside his plate. Then he unfolded his newspaper, looked
at a few items, and sipped his coffee. He had forgotten Josephine,
though she had not forgotten him, and sat waiting until such time as
it should please him to ask the blessing.
For the sake of her patient yet eager face, Peter took an unheard-of
liberty: he nudged his master’s shoulder.
“Hey? What? Peter!” angrily demanded Mr. Smith.
“Yes, suh. Certainly, suh. But I reckon little missy won’t eat
withouten it.”
It was almost as disagreeable to the gentleman to be reminded of
his duty, and that, too, by a servant, as to suffer his present physical
pangs. But he swallowed the lesson with the remainder of his coffee,
and bowed his head, resolving that never again while that brown-
eyed child sat opposite him should such a reminder be necessary.
As before, with the conclusion of the simple grace, Josephine’s
tongue and appetite were released from guard, and she
commented:
“This is an awful funny Baltimore, isn’t it?”
“I don’t know. Do you always state a thing and then ask it?”
returned Uncle Joe, crisply.
“I ’xpect I do ask a heap of questions. Mamma has to correct me
sometimes. But I can’t help it, can I? How shall I know things I don’t
know if I don’t ask folks that do know, you know?”
“You’ll be a very knowing young person if you keep on,” said he.
“Oh! I want to be. I want to know every single thing there is in the
whole world. Papa used to say there was a ‘why’ always, and I like
to find out the ‘whys.’”
“I believe you. Peter, another chop, please.”
“With your foot, Massa Joe?” remonstrated the butler.
“No. With my roll and fresh cup of coffee,” was the retort.
“Excuse me, Massa Joe, but you told me last time that next time I
was to remember you ’bout the doctor saying ‘no meat with the
gout.’”
“Doctors know little. I’m hungry. If I’ve got to suffer I might as well
be hung for a sheep as a lamb. I’ve already eaten two chops.
Another, Peter, and a juicy one.”
The order was obeyed, though the old negro knew that soon he
would be reprimanded as much for yielding to his master’s whim as
he had already been for opposing it.
“Doctor Mack knows everything,” said Josephine.
“Huh! Everybody belonging to you is perfect, I conclude,” said the
host, with some sarcasm.
“I don’t like him, though. Not very well. He gives me medicine
sometimes, though mamma says I don’t need it. I’m glad he’s gone
to eat those philopenas. Aren’t you?”
“I don’t care a rap where he goes,” answered Uncle Joe testily.
Josephine opened her eyes to their widest. This old man in the
soiled green dressing-gown, unshaven, frowning and wincing in a
horrible manner, was like another person to the handsome
gentleman with whom she had dined overnight. He was not half so
agreeable, and— Well, mamma often said that nobody in this world
had a right to be “cross” and make themselves unpleasant to other
people. She was sorry for poor Uncle Joe, and remembered that he
had not had the advantage of mamma’s society and wisdom.
“Uncle Joe, you look just like one of them picture-men that was shut
up in a tree trunk. You know. You showed them to me last night. I
wish you wouldn’t make up such a face,” she observed.
Mr. Smith’s mouth flew open in sheer amazement, while Peter tossed
his hands aloft and rolled his eyes till the whites alone were visible.
In all his service he had never heard anybody dare to speak so
plainly to his master, whose temper was none of the mildest. He
dreaded what would follow, and was more astonished than ever
when it proved to be a quiet:
“Humph! Children and fools speak truth, ’tis said. You’re a sharp-
eyed, unflattering little lady, Miss Josephine; but I’ll try to control my
ugly visage for your benefit.”
The tone in which this was said, rather than the words themselves,
was a reproof to the child, who immediately left her place, ran to her
uncle’s side, and laid her hand pleadingly upon his arm.
“Please forgive me, poor Uncle Joe. I guess that was saucy. I—I
didn’t think. That’s a way I have. I say things first, and think them
afterward. I guess it isn’t a nice way. I’ll try to get over that. My!
won’t that be fun? You trying not to make up faces, and I trying not
to say wrong things. I’ll tell you. Have you got a little box
anywhere?”
“Yes, I presume so. Go eat your breakfast, child. Why?”
“’Cause. Did you know there was heathens?” she asked gravely.
“I’ve heard so. I’ve met a few.”
“You have? How delightful!” came the swift exclamation.
“I didn’t find it so. Why, I say?” he inquired.
“Each of us that forgot and broke over must put a penny, a cent, I
mean, in the box. It must be shut tight, and the cover gum-
mucilaged down. You must make a hole in the cover with your
penknife, and when you screw up your face, just for nothing, you
put a penny in. I’ll watch and tell you. Then I’ll put one in when I
say wrong things. I’ve a lot of money in my satchel. Mamma and
Doctor Mack each gave me some to buy things on the way. But
there wasn’t anything to buy, and I can use it all, only for Rudanthy’s
new head. Can we go buy that to-day, Uncle Joe?”
“No. Nobody knows when I’ll get out again, if this weather holds.
The idea of a snowstorm like this in March. In March!” angrily.
“Yes, suh,” responded Peter respectfully, since some reply seemed
expected.
“Here, boy. Carry my mail to the library. Get a good heat on. Fetch
that old soft shawl I put over my foot when it’s bad, and, for
goodness sake, keep that child out of the way and contented,
somehow.”
Josephine had gone to the window, pulled the draperies apart, and
was looking out on a very different world from any she had ever
seen. White was every object on which her eye rested, save the red
fronts of the houses, and even these were festooned with snowy
wreaths wherever such could find a resting place. The scene
impressed and almost frightened her; but when, presently, it
stopped snowing, and a boy ran out from a neighboring house,
dragging a red sled through the drifts, her spirits rose. It had been
one long, long week since she had exchanged a single word with
any child, and this was an opportunity to be improved. She darted
from the room, sped to the hall door, which stood ajar for Lafayette’s
convenience in clearing off the steps, and dashed outward.
Her feet sank deep into the cold, soft stuff, but she didn’t even
notice that, as she cried, eagerly:
“Little boy! Oh, little boy! Come here quick! I want somebody to play
with me.”
A moment’s pause of surprise, that a child should issue from “old Mr.
Smith’s,” and the answer came cheerily back:
“Wish I could; but I’m going sledding.”
“I’ll go with you! I never went a-sledding in all my”—
The sentence was never finished, for somebody jerked her forcibly
back within doors just as a great express wagon crawled to a pause
before the entrance.
 CHAPTER VI.
MEMORIES AND MELODIES.

“My trunk! my trunk! My darling little blue trunk!”

“Massa Joe says for you to go right straight back to the library,
missy. He says you done get the pneumony, cuttin’ up that way in
the snow, and you not raised in it. He says not to let that boy in
here. I—I’s sorry to disoblige any little lady what’s a-visitin’ of us,
but”—
“It’s my trunk, Peter. Don’t you hear?”
“Yes, missy. But Lafayette, that’s his business, hauling luggage. I’se
the butler, I is.”
Josephine retreated a few paces from the door. She had lived in the
open air, but had never felt it pinch her nose as this did. Her feet,
also, were cold, and growing wet from the snow which was melting
on them. But Peter was attending to that. He was wiping them
carefully with his red handkerchief, and Josephine lifted first one,
then the other, in silent obedience to his touch. But her interest was
wholly in the trunk, which had now been deposited in the vestibule,
and from which Lafayette was carefully removing all particles of
snow before he carried it up over the carpeted stair.
Mr. Smith limped to the library door and looked out. He had meant
to send word that the trunk should be retained at the railway station
for the present, or until he should find out to whom Josephine had
really been “consigned,” and asked, in vexation:
“Come already, has it? Humph! If it had been something I wanted in
a hurry, they’d have taken their own time about delivering it. Said
they couldn’t handle goods in a storm, and such nonsense. I don’t
see, Peter, as it need be taken upstairs. Have it put in the storeroom,
where it will be handier to get at when she leaves.”
Both Peter and Josephine heard him with amazement.
“What is that, Uncle Joe? That ‘when I leave.’ Have I—have I been
so—so saucy and forgetful that—that you can’t let me stay?”
“No, no, child. I merely meant— There, don’t look so distressed. You
are here for the day, anyway, because none of us can go trudging
about in such weather. I’ll telephone for— There. No matter. It’s
right. It’s all right. Don’t, for goodness sake, cry. Anything, anything
but that. Ugh! my foot. I must get out of this draught,” he almost
yelled.
Josephine was very grave. She walked quietly to Uncle Joe’s side,
and clasped the hand which did not hold a cane with both her own.
“It’s dreadful funny, seems to me. Aren’t we going to stay in this
house all the time? I wish—I’m sorry I spoke about the box and the
heatheny money. But if you don’t mind, I must, I must, get into my
trunk. The key is in my satchel in my room. Mamma put it there with
the clean clothes I wore last night. She said they would last till the
trunk came; but that as soon as ever it did I must open it and take
out a little box was in it for you. The very, very moment. I must
mind my mamma, mustn’t I?”
“Yes, child, I suppose so,” he slowly returned.
Mr. Smith was now in his reclining chair, with his inflamed foot
stretched out in momentary comfort. He spoke gently, rather sadly,
in fact, as he added:
“My child, you may open your trunk. I will never counsel you to do
anything against your mother’s wishes. She seems to be a sensible
woman. But there has been a mistake which I cannot understand. I
am Joseph Smith. I have lived in this house for many years, and it is
the street and number which is written on the tag you showed me.
Do you understand me, so far?”
“Course. Why not?”
“Very well. I’m sorry to tell you that I have no twin brother, no ‘sister
Helen,’ and no niece anywhere in this world. I have many cousins
whom I distrust, and who don’t like me because I happen to be
richer than they. That’s why I live here alone, with my colored ‘boys.’
In short, though I am Joseph Smith, of number 1000 Bismarck
Avenue, I am not this same Joseph Smith to whom your mamma
sent you. To-morrow we will try to find this other Joseph Smith, your
mislaid uncle. Even to-day I will send for somebody who will search
for him in my stead. Until he is found you will be safe with me, and I
shall be very happy to have you for my guest. Do you still
understand? Can you follow what I say?”
“Course,” she instantly responded.
But after this brief reply Josephine dropped down upon the rug and
gazed so long and so silently into the fire that her host was impelled
to put an end to her reflections by asking:
“Well, little girl, of what are you thinking?”
“How nice it would be to have two Uncle Joes.”
“Thank you. That’s quite complimentary to me. But I’m afraid that
the other one might prove much dearer than I. Then I should be
jealous,” he returned, smiling a little.
Josephine looked up brightly.
“I know what that means. I had a kitten, Spot, and a dog, Keno; and
whenever I petted Spot Keno would put his tail between his legs and
go off under the sofa and look just—mis’able. Mamma said it was
jealousy made him do it. Would you go off under a table if the other
Uncle Joe got petted? Oh! I mean—you know. Would you?”
Though this was not so very lucid, Mr. Smith appeared to
comprehend her meaning. Just then, too, a severe twinge made him
contort his features and utter a groan.
Josephine was on her feet and at his side instantly, crying out: