Palestine Polytechnic University

College of IT and Computer Engineering

Graduation Projects

Project name:
Admission Chatbot

Project team: Ali Jboor & Maher Salamin

Supervisor: Mr.Mohammad Jabari

May-2021
 Abstract

As a result of the rapid technological development and the development of the chatbot concept and
the time and effort it can save. Many specialized frameworks have emerged to undertake chatbot
creation and development. By relying on artificial intelligence, the chatbot has integrated machine
learning within it, and it has become more comprehensive and wider for various technological fields.
Therefore, we will create a chatbot for the University's Admission and Registration, the project aims
to build a chatbot to facilitate the process of accessing information related to students' inquiries
towards admissions, Registration and the university itself. The motivation for the work of this project
is that there is no university-level equivalent from previous graduation projects, as this project mainly
targets all palestinian tawjihi students and other palestinians, non palestinian students . As a
conclusion, it lies in answering frequent and common questions by people and providing the answer
to these questions at any time the person wants.

What is natural language processing (NLP)?

Natural language processing, which evolved from computational linguistics, uses methods from
various disciplines, such as computer science, artificial intelligence, linguistics, and data science, to
enable computers to understand human language in both written and verbal forms.

What is natural language understanding (NLU)?

Natural language understanding is a subset of natural language processing, which uses syntactic
and semantic analysis of text and speech to determine the meaning of a sentence.

2
Content list

1. Chapter 1: Introduction
1.1. Introduction ………………………………………………………………... 6
1.2. Problem Statement ………………………………………………………… 6
1.3. Objectives ………………………………………………………………….. 6
1.4. Scope ………………………………………………………………………. 6
1.5. Timeline …………………………………………………………………… 7

2. Chapter 2: Chatbots

2.1. Chatbot ………………………………,,,…………………………….……8

2.2. Types of chatbots …………………………………………………………..8

2.3. How do chatbots work …………………………………………………......8
2.4. Options to build chatbots …………………………………………………..9
2.5. Chatbot Platforms alternatives ……………………………………………10
2.6. Selected Platform ………………………………………………………….10
2.7. How to setup Rasa …………………………………………...………….…11
2.8. Supportive information……………………………………...………….......12

3. Chapter 3: Requirements
3.1. Requirements ……………………………………………………………. 15
3.2. Class diagram and use case diagram …………….……………...…..…... 16
3.3. Description and analysis system requirements…………………….……...17

4. Chapter 4: Models and interfaces

4.1. Basics Of Rasa Open Source Conversational AI ……………………….. 21
4.2. Architectural model …………………………………………………...… 21
4.3. Rasa framework files ……………………………………………………. 22
4.4. State and Sequence diagrams ………………………………………….... 24
4.5. Rasa NLU ……………………………………………………………...... 26
4.6. Rasa Core ………………………………………………………………... 27
4.7. User Interface ……………………………………………………………. 27

3
 5. Chapter 5: System demonstration
5.1. Admin panel ………………………………………………………...……29
5.2. Chatbot ……………………………………………………….….…….…32
6. Chapter 6: Testing
6.1. Testing criteria …………………..………………………………………..36
6.2. Test plan …………………………………………….…………………….37
7. Conclusion ………………………………………………………………………40
8. References ……………………………………………………………………….41
9. Appendix ……………………………………………………………………...…42
9.1. Conversation between user and bot used in test case …………..………....42
9.2. real-time visualization of user-bot conversation ………………………….42
9.3. Test case log file…………………………..…………………....……….....43
9.4. Rasa train log file …………………………………..………………..…….55
9.5. rasa run actions log file…………………………………………….………58

Figures and Tables list

Figures:

1. Figure 2.1: Python main screen in microsoft store.………………………………....11

2. Figure 2.2: Testing python version.…………………...…………………………….11
3. Figure 2.3: Rasa bot initial chatting.……………...…....……………………….......12
4. Figure 3.1: Use case diagram.………………………………………………...….…16
5. Figure 3.2: Class diagram …………………………………………………………..16
6. Figure 4.1: Basics Of Rasa Open Source Conversational with ML………….……...21
7. Figure 4.2: Architectural model …………………………………………………….21
8. Figure 4.3: Nlu.yml file content …………………………………………………….22
9. Figure 4.4: Domain and nlu files ………………………………………..…………..23
10. Figure 4.5: Domain and stories files ……………………………………..………….23
11. Figure 4.6: Sequence diagram …………………………………………..…………...25
12. Figure 4.7: State diagram …………………………..………………….…………….26
13. Figure 4.8: Component lifecycle ………………………………………..…………...26
14. Figure 4.9: Rasa core dialog handling ……………………………………..………...27
15. Figure 4.10: Chatting with bot Q1.....……………………………………..………….27
16. Figure 4.11: Chatting with bot Q2........………………………………………………28
17. Figure 5.1: Login screen. ……………..………...………………………….………...29
18. Figure 5.2: Dashboard screen. ……………………………………………....…...…..29
19. Figure 5.3: College menu screen. …………………………………………....……....30

4
20. Figure 5.4: college add/edit screen. ……………………….………………..….…......30
21. Figure 5.5: major menu screen. …………………………….……………...….……...30
22. Figure 5.6: major add/edit screen. ………………………….………………......……..31
23. Figure 5.7: branch menu screen. …………………….…………………..………...… 31
24. Figure 5.8: admin page screen..………………………….……………….….……......31
25. Figure 5.9: admin add screen..…………………….………………….…….…......…..32
26. Figure 5.10: rasa training log screen..………………………….…………….………..34
27. Figure 5.11: rasa action server screen ..……………………………………...………..34
28. Figure 5.12: rasa x screen …………………………...…………………….….……….34
29. Figure 6.1: test use case part I ………………………………………………..……….38
30. Figure 6.2: test use case part II ……………………………………………….……….39
31. Figure I: run-time conversation visualization ………………………………....…….43

Tables:

1. Table 1.1: Timeline.………………………….………………………………….…....7

2. Table 1.2: First Semester Timeline.………………………………………………......8
3. Table 3.1: Handling Admission policy table.………………………………………..17
4. Table 3.2: Answers form a table ……………………………………..……………. 17
5. Table 3.3: Unclear question handling table.…….………………..………………….17
6. Table 3.4: Adding new major ……………………………………………………….18
7. Table 3.5: Editing major data ……………………………………………………….18
8. Table 3.6: Delete major …………………………………………………….………18
9. Table 3.7: Add new college …………………………………………………………19
10. Table 3.8: Editing existing college ………………………………………………….19
11. Table 3.9: Delete college ……………………………………………………………19
12. Table 3.10: Add new tawjihi branch …………………………………………….…..20
13. Table 3.11: Add new tawjihi branch ……………………………………..………….20
14. Table 6.1: Tested features ……………………………………………………………37

5
 Chapter 1: Introduction

1.1 Introduction

Chatbot is a computer program that humans will interact with in natural spoken language and
including artificial intelligence techniques such as NLP (Natural language processing) that makes the
chatbot more interactive and more reliable.
Based on the recent epidemiological situation, the increasing demand and reliance on electronic
education has become very difficult to access to the university due to the curfew imposed, and this
has led to limited access to information for academics at the university.

This project aims to build a chatbot for Admission and Registration to answer every person who asks
about the university, colleges, majors and admission policy.

1.2 Problem Statement

At the start of each academic semester, registration opens for those wishing to join the university in
various disciplines, and telephone calls for admission and registration abound. This leads to an
increase in the loads and work for the employees of the Deanship of Admission and Registration as a
result of the constant pressure of those wishing to register and their families by flocking to the
Deanship, so the employees are not able to answer the phone calls and social media. This often leads
to many students who wish to register to be ignored.

1.3 Objectives
● Save effort and time for both the Admission and registration staff and students who wish to enroll.
● Provide detailed information about colleges and majors.
● Easy access to information.

1.4 Scope
● People who wish to enroll Palestine Polytechnic University
● Admission and registration staff

6
1.5 Timeline

Number Of Task Task name Required time in weeks

1 Proposal 2

2 Information planning and collecting 3

3 Background and Alternatives 2

4 Determine system requirements 2

5 System requirements analysis and 2

design

6 System programming and configure 9

7 System Test 3

8 System documentation All project period

Table 1.1: Timeline .

Table 1.2: First Semester Timeline.

7
 Chapter 2: Background and Alternatives

2.1 Chatbot :

A chatbot is a computer program that can simulate a conversation or chat with a user in natural
language through messaging applications, website, or mobile applications and interact with users
according to their input and should be available 24/7.
Chatbots are developed and became so popular due to the increased use of smart devices and IoT
technology.

2.2 Types of chatbots

a. Base-line chatbot:

It is a chatbot that is based on a database and uses if / then logic to create a

conversation flow and that takes a lot of time to ensure the understanding of the
question and the answer needed.[1]

b. AI chatbot:

This type of chatbot is more complex than base-line but it is more interactive and
personalized and needs big data training to be impressive if the problem is matched to
their capabilities.[2]

c. Hybrid Model:

A hybrid approach mixes the Base-line & AI chatbot to make it smart and his behavior
more expected by depending on database and Ai algorithm to work together.[3]

2.3 How do chatbots work?

Briefly and as mentioned in the definition, humans interact with chatbots.

There are two ways to interact with a chatbot:

a. Text

chatbot analyzes the inputted text and matches the text with predefined data called intents
which are categorized to manage the conversation. The user utterance is tagged with one of
these intents, even if what the user says stretches over two or more intents. Most chatbots will
take the intent with the highest score and take the conversation down that avenue.

8
 b. Voice

Some chatbots can interact and understand the voice of the user using a set of application
programming interfaces (api’s) that converts the recorded voice to the language and then
convert the voice to words of that language and then deal with the transformed text as
mentioned above.

2.4 Options to build a chatbot.

a. From Scratch

At first we have to identify the opportunities for our chatbot and decide its field and scope to
achieve efficiency and accuracy. and a precise understanding of the customer needs is required
to solve the operational challenges.
Then the design of the bot comes to be a significant stage to decide the user engagement with
your app or website.
and we can categorize chatbot interactions as structured and unstructured interactions.

❏ Structured interaction. You already know about this kind of interaction. You
know what your customers will ask and can design it easily — it's just like an
FAQ section of your app or website[4] . This information will link to your
contact information, services, products, etc.
❏ Unstructured interaction. The unstructured conversation flow includes freestyle
plain text. It's hard to predict what queries will emerge and it looks like an
extempore speech competition for your chatbot. the role of AI comes to lights
here, it decodes the context of the text based on NLP analysis. while the same
NLP will provide a voice to the chatbot.[5]

The later choice will need specialized chatbot developers with an understanding of
programming languages, machine learning, and AI. We can use some of the
code-based frameworks to build and handle the chatbot like wit.ai and api.ai.

b. Using platforms

It is similar to scratch chatbots but the only difference is that you do not have to hire a
specialized developer and use the chatbot builder platforms like Chatfuel, Botsify and
Rasa, it’s not hard or impossible to achieve it. but it’s not possible to create a
NLP-enabled chatbot that can deal with unstructured data.

9
2.5 Chatbot Platforms alternatives:
a. IBM Watson: is touted as a question-and-answer system that can be used to build
applications and chatbots. The IBM Watson platform allows us to create an application that
shares a dialog interaction between our chatbot and users on Quick n’ Easy Projector
Rentals. The IBM interface is simple to use, and no back-end coding is shown at first
glance. The chatbot can be easily integrated into other applications such as Slack, Facebook,
and Twilio.
b. Google Dialogflow: is an easy to understand conversational agent. Theoretically, we can
have a bot up and running by understanding 3 core concepts: intents, entities, and dialog
control. As stated earlier, these general concepts are followed across a majority of the
chatbot platforms we played with.
c. Rasa: is an open source chatbot that is equipped with a natural language processing tool.
The open source tool is called Rasa NLU. You can tweak and customize the machine
learning algorithm that Rasa uses so that you can create a model that provides the results
you desire. Rasa NLU can be run wherever you want it to, and none of your training data
has to be passed over to Google, Microsoft, Amazon, or Facebook to train your chatbot.

2.6 Selected Platform:

We will use the rasa framework to build the required chatbot, but why did we choose rasa ?
Will, rasa has a lot of advantages such as:
● Highly customisable with various pipelines can be employed to process user dialogues.
● The rasa framework can be run as a simple http server or can be used from python, using
APIs.
● It has the Rasa-nlu, when run on a server, can mimic other commercial NLP platforms such as
LUIS or wit.ai. This makes it easy to migrate an existing application to rasa-nlu.[3]
But as we know, nothing is perfect and rasa has its disadvantages like :
1. server requirements although spacy is a very fast NLP platform, it seems to be very memory
hungry.
2. Learning curve - Installation, configuration and training phases require machine learning
expertise (at least basic level)
3. Context based conversation not available out of the box - Rasa-nlu does not maintain the
context automatically. This has to be programmed separately into the chat service.

10
2.7 How To Setup Rasa and start a demo ChatBot[8]
● First, we have to know that Rasa needs python to work, so we’ll install python and the
version should be from 3.6 to 3.8 at maximum.

Figure 2.1:Python main screen in microsoft store.

Go to Microsoft Store and search for python 3.8. Then,Check if your Python environment is already
configured by opening the cmd : type python3 --version

Figure 2.2: Testing python version.

11
IF you did NOT find a version, then type this command :pip3 install -U pip

● Virtual Environment Setup

Create and activate the virtual environment using the below commands.
python3 -m venv ./venv
Activate the virtual environment:
.\venv\Scripts\activate
● Install Rasa open-source with the below command:
pip install rasa
● make a directory and move to it
mkdir test-chatbot && cd test-chatbot
Create the new project with rasa init command and start the conversation with the initial
demo chatbot.

Figure 2.3: Rasa bot initial chatting.

2.8 supportive information:

After a great effort in searching and exploring personalities related to the topic of chatbot and natural
language processing (NLU), we want to thank Mr. Jonathan Wheat, who is a certified developer for a
framework Rasa for all his efforts and time to communicate with us and to provide wonderful, useful
and inspiring answers about several Questions we asked him about Rasa and NLU. The following is
the information that he provided us.

12
Our Question: What are the actual parts that are required and needed to be programmed for chatbot
using Rasa Framework ?.
“ There are 3 major parts to the Rasa Framework:
- The Rasa Server: This is the actual software that runs and interprets the user's input and handles the
responses based on a trained model (more on that below).
- The Action Sever: This handles some of the complex form logic if you need it, and/or API calls, etc.
This is written in Python, and there are many examples in the Rasa repo to look at. Depending on
how complex you need your bot to be, you may or may not need to write anything for this.
- The chat interface - This is not part of the normal Rasa framework, you'll probably want to grab one
of the many chat interfaces, and/or connect Rasa to FaceBook Messenger, Slack, Telegram or other
service. Rasa has connectors for these major services to allow you to interact with your bot using
commercial chat interfaces.”[4]

Training data.
“The training data for Rasa 2.0 is all formatted as .yml files. There are 3 types of training data.
- There is the main domain.yml file that contains a list of intents, entities, slots, responses and some
actions/form configuration - these equate to what the bot thinks you mean (intent), the text the bot
responds with (responses), the bots "memory" (slots and entities), and some setup if you're using the
Action server above.
- The next file is an nlu.yml file - this is your training data, it lists all of your intents and the text,
words, phrases that the bot can learn from and interpret as that specific intent. The easiest example is
a "greet" intent. That would have things like - hi, hello, how are you, what's up?, how are you doing?
as the training data and Rasa will process all that when you train.
- The last part of the training data is the stories.yml file - these are actual story flows that start with an
intent and you script how you want the conversation to flow. These are fluid, they're not super strict
"wired up" flows, but they help teach the bot how a conversation should look. If done properly, it will
handle tangents (someone talking about one topic and switching to another and the bot will keep
track). You can, and should, have as many stories as possible, different permutations and flows to
give Rasa the best idea of a conversation. It'll surprise you with how it works, and will (again if done
properly with enough data) converse very well.
To train the bot, it is as simple as typing - rasa train. and getting a cup of coffee or a beer
depending on how big your bot is. In the beginning it'll train pretty quickly, but as you add more and
more data, it'll take longer and longer. I have one that takes over an hour to train because of the data
involved.”[4]

13
Our Question: Is there any majority of advanced python code that I need to write?, or are there
ready-made components and all I have to do is use them ?
“As mentioned above, the only python you'll need is for complex logic and any API calls (ie -
getting the weather, or traffic or server status, etc). When I say complex logic, it is possible to set up
a strict flow to questions (think form fields) for the bot to ask. You can branch depending on answers
and ask follow-up questions based on previous answers. This is all written in Python. You don't need
any specialized python skills, I don't think I've done anything outside of my basic python knowledge.
It is mostly python Classes, functions, etc, and there is a specific pattern you need to follow to get a
FormAction to work properly, so there is a lot of "boilerplate" code if that makes sense.”[4]

Our Question: If I have to build a chatbot from scratch, which algorithms should I use and what skills
and information should I have about NLP ?, and what components do I need to achieve my work ?.
“Wow, it all depends on how robust you want to make it. You'll definitely need to level up on
your Python and/or C, C++. It also depends on what you want the bot to do, there are some simple
classification algorithms that could map an intent with your phrase(s) and then when a new phrase
comes in, it could, with some confidence, predict what intent it maps to, and then have the bot
respond. You could implement a SVM (Support Vector machine) to do something like this, but for
me that's WAY too much effort, and I don't have a doctorate in Machine Learning, or Data Science.
Rasa has it all there for you. What's great is that if there is some cool new algorithm, policy or
pipeline you want to use, you can add it into Rasa and use it. I don't want to discourage you from
rolling your own, but there's a lot to it. There are tutorials out there if that's what you want to do
though.”[4]

Our Question: the tutorials and outer resources that can help me get the work done and overcome this
challenge?.
“Rasa has a bunch of examples of bots in their gitHub repo as well as a few video series (Rasa
Master Class, Algorithm Whiteboard and Breakout Diaries) on YouTube - all free. They also host
webinars and (pre-covid) had a chatbot conference, this year it was online - I think most of those talks
are on youtube as well.”[4]

14
 Chapter 3: System Requirements

3.1 Requirements:
● Functional Requirements:
1. The system must provide clear information about Admission policy.
2. The system must provide clear and fully detailed information about university
colleges.
3. The system must provide clear and fully detailed information about colleges’
programs.
4. The system must provide clear and fully detailed information about colleges’ majors.
5. The system should clarify information about the permitted secondary school branch
for each major.
6. The system should clarify the minimum GPA in high school for each major.
7. The system should clarify the duration of study for each major.
8. The system should clarify the parallel study policy for each major.
9. The system should provide the graduation plans for each major.
10. The system should provide information about placement tests.
11. The system should provide information about first installment costs for each major
and the credit hour price.
● Non-Functional :
1. The system shall handle multiple users inputs, If two or more students are chatting
with the bot, none of the students has to wait too long to be answered by the system.
2. The bot should have a delay in response, to let the student feel like he/she is talking to
a human instead of a bot. A little late response from the chatbot makes the student feel
as though he is talking to a human.
3. The system should have the appropriate data set. The correct data set is the basis for
the chatbot, when the data set is correct and tuned, the chatbot will be trained on it to
give the best possible result.
4. The system should have Data Training. Data training mainly depends on the content
targeted at Admission and Registration deanship.
5. The system should prevent abusive language.
6. The system should be used on http://reg.ppu.edu/ website.
7. Ability to extend the project to include all colleges.

15
 Figure 3.1: Use case diagram.

3.2 Class Diagram and use case diagram

In this part will explain the class diagram that is used to describe the system classes and relation
between them.

Figure 3.2: Class diagram

16
3.3 Description and analysis system requirements:

Table 3.1: Handling Admission policy table .

Table 3.2: Answers form table.

Table 3.3: Unclear question handling table.

17
Table 3.4: Adding new major

Table 3.5: Editing major data

Table 3.6: Delete major

18
Table 3.7: Add new college

Table 3.8: Editing existing college

Table 3.9: Delete college

19
Table 3.10: Add new tawjihi branch

Table 3.11: Add new tawjihi branch

20
 Chapter 4: Models and interfaces

4.1 Basics Of Rasa Open Source Conversational AI:

The diagram below shows the general rasa open source conversational associated with machine
learning. [5]

Figure 4.1. Basics Of Rasa Open Source Conversational with ML .

4.2 Architectural model :

“The diagram below provides an overview of the Rasa Open Source architecture. The two primary
components are Natural Language Understanding (NLU) and dialogue management.
NLU is the part that handles intent classification, entity extraction, and response retrieval. It's shown
below as the NLU Pipeline because it processes user utterances using an NLU model that is generated
by the trained pipeline. The dialogue management component decides the next action in a
conversation based on the context. This is displayed as the Dialogue Policies in the diagram.”[6]

Figure 4.2: Architectural model.

21
4.3 Rasa framework files:

At the moment of writing this report, we have been using the latest version of the Rasa framework
v2.x and in this release the existing file structure and file extensions have changed from previous
versions. Dealing with the framework changed and became better and smoother by using files with
.yml extensions.
● nlu.yml:
○ Intents: The file contains the Natural language understanding ( NLU ) model training
examples. This includes intents, which are user goals, and example utterances that
represent those intents. The NLU training data also labels the entities, or important
keywords, the assistant should extract from the example utterance.
○ Entities: The important keywords that an assistant should take note of. For example,
the message ‘My name is Maher’ has the name ‘Maher’ in it. An assistant should
extract the name and remember it throughout the conversation to keep the interaction
natural.
○ Utterances: Anything a user says. A single utterance is an entire sentence passed as
input to the Chatbot to intent. By definition, an utterance holds an intent and could
potentially include one or more entities. During each transaction with the Chatbot, an
utterance is sent to the Chatbot’s understanding unit to parse and interpret.

Figure 4.3. Nlu.yml file content.

22
● Domain.yml: Defines the environment in which the assistant operates, including:
○ What the user means: specifically, what intents and entities the model can understand.
○ What responses the model can provide: such as utterances or custom actions.
○ What to say next: what the model should be ready to respond with.
○ What info to remember: what information an assistant should remember and use
throughout the conversation.

Figure 4.4. Domain and nlu files.

● data/Stories.yml: example end-to-end conversations. A transcription of a conversation
between a user and a Chatbot, where user inputs are expressed as corresponding intents (and
entities where necessary), and the Chatbot’s responses are expressed as corresponding action
names.

Figure 4.5: Domain and stories files.

23
 ● actions/actions.py: all programming is here. The action class describes how the Chatbot
fulfills various intents. Each action has access to the tracker, which contains data about the
conversation, and implements a run function which holds the logic for carrying out said
action.
● Tracker: A tracker object maintains the current state of the conversations. It keeps track of the
events that have happened so far, such as utterances and actions, as well as other data such as
the slots and entities.
● Slots: Slots are variables a Chatbot requires to perform a specific task. Slots are essential to
interpret a user’s input and adequately execute the action. Slots are commonly filled using
Entities. Slots serve as the building block for a Chatbot’s context manager
● processing pipeline/Config.yml: The configuration file defines the components and policies
that the model will use to make predictions based on user input.
● Endpoints.yml: A web server that reacts to the call, runs the code, and optionally, returns
information to modify the dialogue state. The full configuration of the custom action server is
provided in this file.
● Credentials.yml: This file holds the configuration details for connecting to external
applications, like messaging channels.
4.4 State and Sequence diagrams:

In the following figure, we will show the state diagram and what is the state of the data entry process
by the user and where it passes.
In the beginning, when the user enters a question through the Facebook Messenger platform, this
question moves via Facebook Messenger API's to the section of classifying intens and starting the
process of extracting the necessary entities. In the event that not enough words expressing intents are
revealed; The bot sends a message to the user asking for more information. When the bot gets enough
information to reveal intents and entities, it sends the available data to a special section called
Actions.
In the Actions section, the bot specifies the appropriate response to be sent to the user after
performing matching operations for extracted entities based on the highest corresponding percentage
of these entities. After that, the bot sends the resulting information and shows it as a message via the
Facebook Messenger platform to the user.

24
In the figure below, we'll show the sequence diagram. In this diagram, we will carry out a complete
tracking process for all stages of the system, how it handles and processes user input.
In the beginning, the user enters the data through the Facebook Messenger platform and is sent via
Facebook Messenger API's to the Rasa framework, which will perform several operations to process
the available data. Where the framework sends the data to the Rasa NLU file, where both the
classification of intentions and the extraction of entities from the sent data are performed in this file,
where the data is divided into words and from these divided words the intent is extracted as these
intentions are already predefined words Inside the Rasa NLU file. For example, if we ask what's the
weather like tomorrow? When sending this question, the framework identifies the intentions through
some of the key words in the question, and here the key word is weather. Then they are recognized
and likewise the entities are extracted which are tomorrow in this case.
After that, the framework sends the classified and extracted intentions and entities to Rasa Core, to
determine the appropriate actions and send the appropriate response to the user via the Facebook
Messenger platform.

Figure 4.6: Sequence diagram.

25
 Figure 4.7: State diagram.
4.5 Rasa NLU:
The NLU module is answerable for concluding all language parsing tasks. The foremost important of
those tasks are intent detection and named entity recognition. The implementation of the NLU
module is modeled as a pipeline that processes the input text in consecutive steps called components.
So, it’s a crucial thing to grasp the components’ lifecycle and the way they interact with one another.
The figure below shows the Rasa NLU components’ lifecycle. Before starting the pipeline, a context
object is passed to among components in order that they can dissipate information. This object allows

Figure 4.8: Component lifecycle.

the output of 1 component to be used because of the input of the text.
In order to do intent classification, the next has to be pre-processed first into a format that is more
machine-learning friendly. “The most widely used method in literature is the word vector embedding.
This method represents words as vectors in a high dimensional space with the proposition that
semantically similar words would fall closer to each other in terms of distance in this high
dimensional space. Thus, the first step of the NLU pipeline is to tokenize the input.

26
Following that, the input is passed to a regex featurizer which extracts features that match predefined
regular expressions, such as dates and numbers, as part of a simple entity detection algorithm. The
next component in the pipeline runs a Conditional Random Field model for entity extraction based on
Scikit Learn. The next component maps the extracted entities to their synonyms provided by a
training file. The following component converts the data to a bag of words, suitable for intent
classification. The final component of the pipeline does intent classification using a model based on
StarSpace.”[7]

4.6 Rasa Core :Dialogue Handling/Policies:

Rasa core is used to implement the dialogue handling component to predict which action to be
executed and update the state for this action and give the output as well as used for next input.

Figure 4.9: Rasa core dialog handling.

4.7 User Interface:

Asking about enrollment and payments.

Figure 4.10: chatting with bot Q1.

27
● Asking about majors and view specific major information

Figure 4.11: chatting with bot Q2.

28
 Chapter five: System demonstration

5.1. Admin panel:

5.1.1. Implementation details: This interface has been developed so that those involved in the
Deanship of Admission and Registration at Palestine Polytechnic University add, amend
or delete what is necessary regarding each of the colleges in the university in all cases.
The same applies to the majors offered by each college as soon as the decision is made to
establish a new specialization, change the characteristics of one of the pre-existing
specializations, or cancel a specialization in itself. And also concerning the Tawjihi
branches approved by the Ministry of Education, it is possible to add or delete what is
necessary from these branches as necessary.
5.1.2. Implementation issues: We did not face many problems while developing this panel due
to our prior knowledge and ability to develop websites using this combination of PHP
with HTML, CSS, Bootstrap, and a little Javascript. However, some of the problems that
we encountered made us go through a difficult time sometimes, which is also difficult to
track, because we did not use a framework to implement this panel and no error log is
referred to when encountering some problems.
5.1.3. Snapshots and and UI description:

● Login: This page is the first page to be shown when accessing the site

Figure 5.1:Login screen.

● Dashboard: This page will be the first thing that appears to the user when he logs
into the site correctly, whether he is an administrator or a viewer.

Figure 5.2: Dashboard screen.

29
● Colleges menu: This page displays all the colleges in the university, and it appears
for both the viewer and the administrator. The administrator can add a new college,
remove a pre-existing college, or modify pre-existing college data.

Figure 5.3: college menu screen.

● College data: This page displays the fields assigned to each college, name, and
description. Only an administrator can access this page.

Figure 5.4: college add/edit screen.

● Majors menu: On this page, all the majors that the student can study on the
campus of Palestine Polytechnic University are displayed, and both the viewer and
the administrator can access this page.

Figure 5.5: major menu screen.

30
● Major data: Only administrators can view this page, this page contains all
necessary fields for majors at Palestine Polytechnic University. Starting with the
title of the major to the program attached to this major through each of the college
supervising this specialization and each of the total credit hours, the hourly rate
specified in JOD, the value of the first installment, minimum GPA, duration of
the assigned study and the approved guideline branches for each major.

Figure 5.6: major add/edit screen.

● Branch menu: Here all the branches of the Tawjihi that its students can enroll in
the colleges' majors are displayed. Can be viewed by the viewer and administrator.

Figure 5.7: branch menu screen.

● Admin page: All users of the site are shown on this page. Both the viewer and the
administrator can access this page.

Figure 5.8: admin page screen.

31
 ● Admin data: On this page, special data is added to create a new user for the site.
Only an administrator can access this page.

Figure 5.9: admin add screen.

5.1.4. Important information: As a summary, this panel gives the administrator full access to
all the available data regarding each of the colleges, majors, and Tawjihi branches, with
the possibility of adding, deleting or modifying them. With the possibility of adding new
users or deleting previous users.

5.2. Chatbot
5.2.1. Implementation details:
The project consists of several parts, the most important of which are:
1- Actions section, which is the folder that contains 3 files in the Python programming
language.The most important file is the Actions file, which contains custom Actions that
are built according to the need and purpose of the chatbot.
2- The data section and contains 3 important files that cannot be dispensed with: the
natural language understanding file, which contains the training data necessary for the bot,
which it is expected to receive during its operation from the user, and the rules file, which
contains a certain structure that makes the bot act obligatory according to what exists,
regardless of the circumstances in terms of the received data, and the story file, which
contains scenarios of conversations with users, and all conversations are recorded within
this file in .yml extension.
3- There is also a very important file, which is the domain that defines the universe in
which your assistant operates. It specifies the intents, entities, slots, responses, forms, and
actions your bot should know about. It also defines a configuration for conversation
sessions.
4- There are also configuration files, endpoints, and credentials that are responsible for
the overall properties and some permissions allowed for the bot and linking with chat
channels such as Slack and Facebook Messenger.
5- There is also the Models section, in which all models are stored after each bot
training. Every model we can say is like the nucleus or brain of the bot. The bot cannot
work and listen to the user’s messages and respond to them without the model.

32
 6- After each bot training process, and to get the latest results, you must choose the
newest model. Older models can be selected so that they can be compared with the new
model in terms of additions.

5.2.2. Implementation issues: We faced many problems. In each stage we passed, we

encountered many problems, some of which took days to solve. We will mention the most
important problems that we encountered and stood in our way:
1- The problem of installing the framework, during the installation on Windows we
did not encounter any serious problems except for the problem of installing Rasa X, which
is the GUI. The last version was not very compatible with Windows at the time, so we had
to use one of the old versions so that we could work on Rasa X.
2- Installing the framework on Linux was one of the biggest problems we faced, and
the problem was divided into smaller problems: the issue of the Python programming
language version, the problem of pip packages, and the problem of environment variables
and dependencies.
3- We also encountered problems with the actions, especially that the file was sensitive
to the issue of functions' names and how to deal with it and not respond except by relying
on the correct and acceptable naming rules.
4- The biggest problem we faced was the problem of integrity error while working
with Rasa X and its connection with the Action Server on Windows. The problem
revolved around the inability to record or respond to commands and conversations
between the user and the bot in the dedicated database. We repeatedly searched, asked,
and investigated this problem that has not been resolved at the time of writing this report,
as it was a problem related to the Windows operating system. We did not encounter the
problem while running the bot on Linux.
5- One of the problems that we faced was about slots and not filling them out
according to the rules in the Domain file. While talking to the bot, some answers are
extracted and processed, and responded to according to what was extracted and we
encountered the problem on Windows.
6- The bot training process is one of the most frequent problems in terms of time and
space, so every training process for the bot requires at least 4 minutes at least. After
completing the training process, a model with a size of more than 20 MB will be
produced. And you must activate the latest model after each training process.
7- Not all buttons appear when the bot is connected to chat channels such as Slack and
Facebook Messenger.
8- Uploading the bot to Docker was one of the most difficult problems for us due to
our lack of knowledge of Docker and the mechanism of uploading projects to it and the
necessity of using Linux.

33
5.2.3. Snapshots and UI description:
● Rasa training log screen:

Figure 5.10: rasa training log screen.

● Rasa actions server terminal screen:
Rasa action server runs custom actions for a Rasa Open Source conversational
assistant. When your assistant predicts a custom action, the Rasa server sends a
POST request to the action server with a json payload including the name of the
predicted action, the conversation ID, the contents of the tracker and the contents
of the domain. When the action server finishes running a custom action, it returns a
json payload of responses and events. See the API spec for details about the
request and response payloads. The Rasa server then returns the responses to the
user and adds the events to the conversation tracker.

Figure 5.11: rasa action server screen.

● Rasa X terminal screen:

Figure 5.12: rasa x screen.

34
5.2.4. Important information:
● When training the bot, it is important to make sure that the structure in the domain
file is in the right format or it will fail.
● Training time is machine dependent.
● It’s ok to delete old models, but it’s not recommended.
● Intents in the NLU file have to contain at least two training examples. One training
example may lead to training failures.
● Intents have to be part of the domain. If not, it may lead to issues in the
conversations.
● Any change in the actions file requires to stop the action server and restart it again
by typing rasa run actions from the terminal ONLY to acquire the changes.
● Any change in nlu, rules, stories, config, credentials, domain, and endpoints files
requires to stop the bot and restart it again in case of using terminal. In case of
using Rasa X GUI, the bot won’t stop or need to be restarted.
● In case Rasa X was stopped, the ONLY way to start it again is using the terminal
by typing rasa x.
● User needs a password to have access to the internal Rasa X GUI, it is
automatically generated in every attempt to start Rasa X from the terminal.
● Rasa X password is not changed when attempting to train the bot using the GUI.
● It’s ok to have a User Warning about the model confidence being set to softmax
instead of linear norm. Rasa framework developers recommend the softmax
instead of the linear norm.

35
 Chapter 6: Testing

6.1. Testing criteria:

6.1.1. Chatbots testing criterias:
Criteria derived from the Rasa framework, which is a set of commands that are
executed to obtain degrading results based on the contents and how the bot is built. It
is done by executing specific commands, data validation verifies that no major errors
or inconsistencies appear in your domain, NLU data, or story data. If data validation
leads to errors, training the model may also fail or result in poor performance, so it is
always a good idea to do this check before training the model. Including the
--fail-on-warnings flag, this step will fail warnings that indicate more minor issues.
Another thing you can do to get the bot to be tested is to write test stories.
Testing your trained model on test stories is the best way to be confident in how your
assistant will behave in certain situations. Test stories, written in a modified story
format, allow you to present complete conversations and test that, given certain user
input, your model will behave expectedly. This is especially important when you start
to present more complex stories than user conversations. Test stories are similar to the
stories in your training data but include the user's message as well.

Conversation flow: in this criteria, the following things should be checked while
testing:
● Does a chatbot clearly understand the questions?
● Does it always give instant responses to these questions?
● Are the answers relevant to the given questions?
● Should a user ask a series of questions until he gets the answer?
● Does a chatbot engage the user to continue the conversation?[9]

Business-specific questions: “Each industry has its specific terminology, notions,

nuances. So, a tester should have a list of domain-specific questions to check if the
chatbot is able to answer those.”[9]

Confusion handling: “Confusion may arise if a user enters some expression with
double meaning or an unknown word for a chatbot. The latter should be taught to
answer in such a situation. The tester’s goal is to check if the chatbot can handle
misunderstandings, exceptional conversational scenarios, and unusual patterns. This
capability of a chatbot is showing how “emotionally intelligent” it is.”[9]

6.1.2. The classic criteria in the software test. By testing the software at and around critical
points, and what it accepts and what it does not accept. The software is also presented to a
group of people to be tested and examined based on interactions and data with humans.

36
6.2. Test plan

6.2.1. Features to be tested: Tested features will be those features that the bot has been trained
on and given all the necessary information and data about these features.

User input/intent Bot action

admission_policy utter_admission_policy

choose_college utter_choose_college

program utter_program

expected_major expected_major_form

parallet_study utter_parallel_study

placement_tests utter_placement_tests

first_installment utter_first_installment
Table 14: Tested features

6.2.2. Features not to be tested:

● Feature of return and change the value of the entered mark.
● The feature of showing a welcome message to the user when the chat
window is opened.
● The feature of changing the intended intent of a previous message.

6.2.3. Testing environment and tools description:

Facebook Messenger and Slack were used as test environments for the
bot but were abandoned for reasons mentioned in the challenges section.
Therefore, the testing of the bot was completed on the local host environment
by hosting the bot using NG Rock software, which creates a link that enables
users outside the local environment to access and interact with the bot.
Not many testing tools were used, so to speak, a set of commands dedicated to
bot testing were used by the developers of the framework, namely:
● rasa test.
● rasa data validate.
● rasa test nlu.
● rasa data split nlu.

37
6.2.4. Test case diagram:
This test case is a simulation of a segment of a conversation between a user and the bot
that satisfies the use case model in Chapter 3. This short conversation consists of messages
sent by the user, texts, and questions that the bot replied.
The following flow chart shows in detail what is being processed during this short
conversation, given that the diagram is very large, we have divided it into two parts to
facilitate the process of follow-up and scrutiny.

Part 1

Figure 6.1: test use case part I

38
Part 2

Figure 6.2: Test use case II

39
 Conclusion

This bot was built to respond to the inquiries of the Tawjihi students regarding each of the university's
faculties and their specializations, with extracted information for each specialization, familiarizing
students with the level exams that students submit about their enrollment in the university,
introducing the educational qualification diploma program and the mechanism for joining it. Giving
students notes on the electronic enrollment application package, the locations of approved banks,
and how to fill out the application. Introduce Bagrut students to the conditions and notes that must be
taken into account in the event of joining Palestine Polytechnic University and the mechanism for
calculating grades. Introduce students to the procedures followed to reserve a seat and what
documents are required after the student is accepted. Introducing students to the system of
transferring to Palestine Polytechnic University from another university on the undergraduate
system. Informing students of the university’s teaching system and language. Introducing students to
the student exchange system with other universities. Introducing students to the system of grants,
exemptions, and financial aid provided to students. Informing students of cases in which the student
loses his university seat. Introducing students to the installment refund system for new students and
its conditions.

Challenges: We faced many challenges, and the biggest challenge during the work was the arrest
of my colleague Ali Jboor, a month before the deadline for the project. Dealing with the framework is
one of the biggest challenges, as it is new and not widespread. And also the challenge of learning
the necessary Python language to complete the construction of this bot. We can say that the current
health situation contributed to creating communication problems between team members and
problems related to slow internet and power outages during work.
We also mention the challenge we faced, so we divided the work into two parts, part for the admin
panel and a part to follow up on the work on the bot. Unfortunately, Ali was arrested and I had to do
the project alone in his absence, especially after we agreed on a day to share the results of the work,
but that day Ali was arrested.

interesting decisions:
1- The idea of ​the project itself was very interesting. We considered this project and set it as a
challenge to our abilities and ourselves. To be based on learning to use and implement programs
using a non-renowned framework for a project that is the most important in a university student's
career.
2- The idea of ​changing the operating system used to run the bot was one of the most crucial
decisions in the workflow of the project. Windows was the best and most worthy on paper, according
to the sources. But there was one problem that we had encountered for such a long time that made
an operating system change necessary.

Recommendations: It is possible to modify and increase the efficiency of the bot to the fullest extent
if the time factor and the human factor are available. Unfortunately, we were not able to deliver the
bot to the maximum extent that we drew and expected due to the circumstances that befell us.
Additional matters necessary for students related to registration, student status, and forms to official
documents can be added.

Future improvements: The project scope may be expanded to include all corners of the university,
including faculties and deanships of registration and follow-up of all matters that the student is
interested in during their academic life. The ability to communicate using voice messages.

40
 References
1. "Chatbots - Artificial Solutions." Chatbots | Conversational AI Platform for Enterprise -
Teneo. Accessed 8 Nov. 2020.
2. "How Do Bots and Chatbots Work."
https://www.cxtoday.com/contact-centre/how-do-bots-and-chatbots-work/
3. "Building Chatbots - A comparison of Rasa-NLU and Dialogflow." 17 Apr. 2018, Building
Chatbots - A comparison of Rasa-NLU and Dialogflow . Accessed 8 Nov. 2020.
4. "Jonathan Wheat" Jonathan Wheat - DEV Community. Accessed 10 Nov. 2020.
5. Understanding the Basics of Rasa - Open source conversational AI
https://smazee.com/blog/basics-of-rasa
6. Rasa Architecture Overview
https://rasa.com/docs/rasa/arch-overview/
7. An Architecture for Dynamic Conversational Agents for Citizen Participation and Ideation by:
Saifeldin Ahmed
(PDF) An Architecture for Dynamic Conversational Agents for Citizen Participation and
Ideation (researchgate.net)
8. Installation: https://rasa.com/docs/rasa/installation
9. How to test a chatbot. Accessed May 31-2021:
https://apphawks.com/blog/a-compelling-guide-on-chatbot-testing

41
Appendix

1- Conversation between user and bot used in section 6.2.4:

User: Hello
Bot: Welcome
User: What specialties are available?
Bot: Enter your branch in high school
User: scientific
Bot: Enter your high school average
User: 59.6
Bot: What program would you like to enroll in at Palestine Polytechnic University?
User: (presses the Bachelor button) / writes the word Bachelor
Bot: We regret that you did not meet the admission requirements for any of the
disciplines affiliated with the Bachelor's Program

2- In the following diagram, a real-time visualization of the conversation progression appears during
the bot's interaction with the user's input.

Figure I : run-time conversation visualization

42
3- Test case log file content:
(venv) maher@maher-PC:~/Documents/admission-bot$ rasa interactive
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:97: UserWarning:
Action 'utter_university_colleges' is listed as a response action in the domain file, but there is no
matching response defined. Please check your domain.
More info at https://rasa.com/docs/rasa/responses
Nothing changed. You can use the old model stored at
'/home/maher/Documents/admission-bot/models/20210531-231738.tar.gz'.
2021-06-01 17:03:51 INFO rasa.model - Loading model models/20210531-231738.tar.gz...
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/utils/train_utils.py:593:
UserWarning: constrain_similarities is set to `False`. It is recommended to set it to `True` when using
cross-entropy loss. It will be set to `True` by default, Rasa Open Source 3.0.0 onwards.
category=UserWarning,
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/utils/train_utils.py:565:
UserWarning: model_confidence is set to `softmax`. It is recommended to try using
`model_confidence=linear_norm` to make it easier to tune fallback thresholds.
category=UserWarning,
2021-06-01 17:04:07 INFO root - Rasa server is up and running.
Processed story blocks: 100%|████████| 26/26 [00:00<00:00, 1961.68it/s, # trackers=1]
Processed rules: 100%|███████████████| 10/10 [00:00<00:00, 3479.88it/s, # trackers=1]
Bot loaded. Visualisation at http://localhost:5006/visualization.html .
Type a message and press enter (press 'Ctrl-c' to exit).
? Your input -> ‫مرحبا‬
? Your NLU model classified '‫ 'مرحبا‬with intent 'arabic_greet' and there are no entit
ies, is this correct? Yes
------
Chat History

# Bot You
───────────────────────────────────────────────────
1 action_listen
───────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00

43
Current slots:
colleges: None, major: None, branch_of_tawjihi: None, mark_of_branch: None, program: None,
requested_slot: None, session_started_metadata: None

------
? The bot wants to run 'utter_arabic_greet', correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬

Current slots:
colleges: None, major: None, branch_of_tawjihi: None, mark_of_branch: None, program: None,
requested_slot: None, session_started_metadata: None
------
? The bot wants to run 'action_listen', correct? Yes
? Your input -> ‫ما هي التخصصات المتاحة؟‬
? Your NLU model classified '‫ 'ما هي التخصصات المتاحة؟‬with intent 'expected_major' an
d there are no entities, is this correct? Yes
------
Chat History
# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬

44
 intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00

Current slots:
colleges: None, major: None, branch_of_tawjihi: None, mark_of_branch: None, program: None,
requested_slot: None, session_started_metadata: None
------
? The bot wants to run 'expected_major_form', correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":

45
 "expected_major_form"}
slot{"requested_slot":
"branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬

Current slots:
colleges: None, major: None, branch_of_tawjihi: None, mark_of_branch: None, program: None,
requested_slot: branch_of_tawjihi, session_started_metadata: None
------
? The bot wants to run 'action_listen', correct? Yes
? Your input -> ‫علمي‬
? Your NLU model classified '‫ 'علمي‬with intent 'tawjihi_branch' and there are no enti
ties, is this correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":
"expected_major_form"}
slot{"requested_slot":

46
 "branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬
action_listen 1.00
─────────────────────────────────────────────────────────
6 ‫علمي‬
intent: tawjihi_branch 1.00

Current slots:
colleges: None, major: None, branch_of_tawjihi: None, mark_of_branch: None, program: None,
requested_slot: branch_of_tawjihi, session_started_metadata: None

------
? The bot wants to run 'expected_major_form', correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":
"expected_major_form"}
slot{"requested_slot":

47
 "branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬
slot{"branch_of_tawjihi": "‫}"علمي‬
slot{"requested_slot": "mark_of_branch"}
‫أدخل معدلك في الثانوية العامة‬

Current slots:
colleges: None, major: None, branch_of_tawjihi: ‫علمي‬, mark_of_branch: None, program: None,
requested_slot: mark_of_branch, session_started_metadata: None

------
? The bot wants to run 'action_listen', correct? Yes
? Your input -> 59.6
? Your NLU model classified '59.6' with intent 'branch_mark' and there are no enti
ties, is this correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":

48
 "expected_major_form"}
slot{"requested_slot":
"branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬
slot{"branch_of_tawjihi": "‫}"علمي‬
slot{"requested_slot": "mark_of_branch"}
‫أدخل معدلك في الثانوية العامة‬
action_listen 1.00
─────────────────────────────────────────────────────────
6 59.6
intent: branch_mark 0.55

Current slots:
colleges: None, major: None, branch_of_tawjihi: ‫علمي‬, mark_of_branch: None, program: None,
requested_slot: mark_of_branch, session_started_metadata: None

------
? The bot wants to run 'expected_major_form', correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00

49
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":
"expected_major_form"}
slot{"requested_slot":
"branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬
slot{"branch_of_tawjihi": "‫}"علمي‬
slot{"requested_slot": "mark_of_branch"}
‫أدخل معدلك في الثانوية العامة‬
slot{"mark_of_branch": 59.6}
slot{"requested_slot": "program"}
‫ما هو البرنامج الذي ترغب بااللتحاق به في‬
‫جامعة بوليتكنك فلسطين؟‬
Buttons:
1:
‫)بكالوريوس (بكالوريوس‬
2: ‫)دبلوم (دبلوم‬
Type out your own message...

Current slots:
colleges: None, major: None, branch_of_tawjihi: ‫علمي‬, mark_of_branch: 59.6, program: None,
requested_slot: program, session_started_metadata: None

------
? The bot wants to run 'action_listen', correct? Yes
? ‫ بكالوريوس (بكال‬:1 ‫ما هو البرنامج الذي ترغب بااللتحاق به في جامعة بوليتكنك فلسطين؟‬
‫)وريوس‬
------
Chat History

# Bot You
─────────────────────────────────────────────────────────

50
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":
"expected_major_form"}
slot{"requested_slot":
"branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬
slot{"branch_of_tawjihi": "‫}"علمي‬
slot{"requested_slot": "mark_of_branch"}
‫أدخل معدلك في الثانوية العامة‬
slot{"mark_of_branch": 59.6}
slot{"requested_slot": "program"}
‫ما هو البرنامج الذي ترغب بااللتحاق به في‬
‫جامعة بوليتكنك فلسطين؟‬
Buttons:
1:
‫)بكالوريوس (بكالوريوس‬
2: ‫)دبلوم (دبلوم‬
Type out your own message...
action_listen 1.00
─────────────────────────────────────────────────────────
6 ‫بكالوريوس‬
intent: university_programs 1.00

51
Current slots:
colleges: None, major: None, branch_of_tawjihi: ‫علمي‬, mark_of_branch: 59.6, program: None,
requested_slot: program, session_started_metadata: None

------
? The bot wants to run 'expected_major_form', correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬
intent: expected_major 1.00
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":
"expected_major_form"}
slot{"requested_slot":
"branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬
slot{"branch_of_tawjihi": "‫}"علمي‬
slot{"requested_slot": "mark_of_branch"}
‫أدخل معدلك في الثانوية العامة‬

52
 slot{"mark_of_branch": 59.6}
slot{"requested_slot": "program"}
‫ما هو البرنامج الذي ترغب بااللتحاق به في‬
‫جامعة بوليتكنك فلسطين؟‬
Buttons:
1:
‫)بكالوريوس (بكالوريوس‬
2: ‫)دبلوم (دبلوم‬
Type out your own message...
slot{"program": "‫}"بكالوريوس‬
slot{"requested_slot": null}
active_loop{"name": null}

Current slots:
colleges: None, major: None, branch_of_tawjihi: ‫علمي‬, mark_of_branch: 59.6, program:
‫بكالوريوس‬, requested_slot: None, session_started_metadata: None

------
? The bot wants to run 'action_tawjihi_branch_and_mark', correct? Yes
------
Chat History

# Bot You
─────────────────────────────────────────────────────────
1 action_listen
─────────────────────────────────────────────────────────
2 ‫مرحبا‬
intent: arabic_greet 1.00
─────────────────────────────────────────────────────────
3 utter_arabic_greet 1.00
‫أهال وسهال بك‬
action_listen 1.00
─────────────────────────────────────────────────────────
4 ‫ما هي التخصصات المتاحة؟‬

53
 intent: expected_major 1.00
─────────────────────────────────────────────────────────
5 expected_major_form 1.00
active_loop{"name":
"expected_major_form"}
slot{"requested_slot":
"branch_of_tawjihi"}
‫أدخل فرعك في الثانوية العامة‬
slot{"branch_of_tawjihi": "‫}"علمي‬
slot{"requested_slot": "mark_of_branch"}
‫أدخل معدلك في الثانوية العامة‬
slot{"mark_of_branch": 59.6}
slot{"requested_slot": "program"}
‫ما هو البرنامج الذي ترغب بااللتحاق به في‬
‫جامعة بوليتكنك فلسطين؟‬
Buttons:
1:
‫)بكالوريوس (بكالوريوس‬
2: ‫)دبلوم (دبلوم‬
Type out your own message...
slot{"program": "‫}"بكالوريوس‬
slot{"requested_slot": null}
active_loop{"name": null}
action_tawjihi_branch_and_mark 1.00
‫نأسف لعدم استيفائك شروط القبول الي من‬
‫التخصصات التابعة لبرنامج بكالوريوس‬

Current slots:
colleges: None, major: None, branch_of_tawjihi: ‫علمي‬, mark_of_branch: 59.6, program:
‫بكالوريوس‬, requested_slot: None, session_started_metadata: None

------
? The bot wants to run 'action_listen', correct? Yes
? Your input ->

54
4- rasa train log file content:
(venv) maher@maher-PC:~/Documents/admission-bot$ rasa train
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:97: UserWarning:
Action 'utter_university_colleges' is listed as a response action in the domain file, but there is no
matching response defined. Please check your domain.
More info at https://rasa.com/docs/rasa/responses
2021-06-01 23:41:58 INFO rasa.model - Data (domain) for Core model section changed.
2021-06-01 23:41:58 INFO rasa.model - Data (messages) for NLU model section changed.
2021-06-01 23:41:58 INFO rasa.model - Data (domain) for Core model section changed.
2021-06-01 23:41:58 INFO rasa.model - Data (messages) for NLU model section changed.
Training NLU model...
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:97: UserWarning:
Please configure the number of 'epochs' in your configuration file. We will change the default value
of 'epochs' in the future to 1.
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/utils/train_utils.py:593:
UserWarning: constrain_similarities is set to `False`. It is recommended to set it to `True` when using
cross-entropy loss. It will be set to `True` by default, Rasa Open Source 3.0.0 onwards.
category=UserWarning,
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/utils/train_utils.py:565:
UserWarning: model_confidence is set to `softmax`. It is recommended to try using
`model_confidence=linear_norm` to make it easier to tune fallback thresholds.
category=UserWarning,
2021-06-01 23:42:00 INFO rasa.shared.nlu.training_data.training_data - Training data stats:
2021-06-01 23:42:00 INFO rasa.shared.nlu.training_data.training_data - Number of intent
examples: 267 (40 distinct intents)

2021-06-01 23:42:00 INFO rasa.shared.nlu.training_data.training_data - Found intents:

'dploma_tahel', 'arabic_night_greet', 'registration', 'affirm', 'bot_condition',
'refund_system_for_new_student', 'arabic_goodbye', 'placement_tests', 'system_study',
'university_programs', 'Losing_a_seat', 'arabic_morning_greet', 'tawjihi_branch',
'continue_conversation', 'transfer_to_ppu', 'graduation_plans', 'choose_major', 'bot_challenge', 'deny',
'nlu_fallback', 'goodbye', 'parallel_study', 'continue_pajrot', 'abuse_words',
'pajrot_calculate_mark_and_major', 'filling_the_application', 'colleges',
'installments_and_credit_hours', 'choose_college', 'branch_mark', 'arabic_greet',

55
'pajrot_enrollment_Conditions', 'ppu_grants', 'restart', 'greet', 'admission_policy', 'thanks',
'language_study', 'required_documents', 'expected_major'
2021-06-01 23:42:00 INFO rasa.shared.nlu.training_data.training_data - Number of response
examples: 0 (0 distinct responses)
2021-06-01 23:42:00 INFO rasa.shared.nlu.training_data.training_data - Number of entity
examples: 0 (0 distinct entities)
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:97: UserWarning:
Intent 'restart' has only 1 training examples! Minimum is 2, training may fail.
/home/maher/Documents/venv/lib/python3.7/site-packages/rasa/shared/utils/io.py:97: UserWarning:
Intent 'continue_pajrot' has only 1 training examples! Minimum is 2, training may fail.
2021-06-01 23:42:00 INFO rasa.nlu.model - Starting to train component WhitespaceTokenizer
2021-06-01 23:42:00 INFO rasa.nlu.model - Finished training component.
2021-06-01 23:42:00 INFO rasa.nlu.model - Starting to train component CountVectorsFeaturizer
2021-06-01 23:42:00 INFO rasa.nlu.featurizers.sparse_featurizer.count_vectors_featurizer - 366
vocabulary slots consumed out of 1366 slots configured for text attribute.
2021-06-01 23:42:00 INFO rasa.nlu.model - Finished training component.
2021-06-01 23:42:00 INFO rasa.nlu.model - Starting to train component DIETClassifier
Epochs: 100%|████████████| 300/300 [09:24<00:00, 1.88s/it, t_loss=1.46, i_acc=0.996]
2021-06-01 23:51:25 INFO rasa.nlu.model - Finished training component.
2021-06-01 23:51:25 INFO rasa.nlu.model - Starting to train component FallbackClassifier
2021-06-01 23:51:25 INFO rasa.nlu.model - Finished training component.
2021-06-01 23:51:25 INFO rasa.nlu.model - Successfully saved model into
'/tmp/tmpj_0w3csd/nlu'
NLU model training completed.
Training Core model...
Processed story blocks:
100%|██████████████████████████████████████████████████████
█████████████████████| 26/26 [00:00<00:00, 217.94it/s, # trackers=1]
Processed story blocks:
100%|██████████████████████████████████████████████████████
█████████████████████| 26/26 [00:02<00:00, 9.74it/s, # trackers=25]
Processed story blocks:
100%|██████████████████████████████████████████████████████
█████████████████████| 26/26 [00:06<00:00, 3.94it/s, # trackers=50]

56
Processed story blocks:
100%|██████████████████████████████████████████████████████
█████████████████████| 26/26 [00:07<00:00, 3.38it/s, # trackers=50]
Processed rules:
100%|██████████████████████████████████████████████████████
████████████████████████████| 10/10 [00:00<00:00, 375.80it/s, # trackers=1]
Processed trackers:
100%|██████████████████████████████████████████████████████
████████████████████████| 25/25 [00:00<00:00, 127.51it/s, # actions=135]
Processed actions: 135it [00:00, 1502.46it/s, # examples=135]
Processed trackers:
100%|██████████████████████████████████████████████████████
██████████████████████| 525/525 [00:11<00:00, 44.32it/s, # actions=1960]
Epochs:
100%|██████████████████████████████████████████████████████
████████████████| 100/100 [08:18<00:00, 4.98s/it, t_loss=7.88, loss=7.5, acc=0.904]
Processed trackers:
100%|██████████████████████████████████████████████████████
█████████████████████████| 10/10 [00:00<00:00, 338.96it/s, # actions=23]
Processed actions: 23it [00:00, 2582.98it/s, # examples=21]
Processed trackers:
100%|██████████████████████████████████████████████████████
████████████████████████| 25/25 [00:00<00:00, 117.29it/s, # actions=138]
Processed trackers:
100%|██████████████████████████████████████████████████████
███████████████████████████████████████| 10/10 [00:00<00:00,
156.09it/s]
Processed trackers:
100%|██████████████████████████████████████████████████████
████████████████████████████████████████| 35/35 [00:00<00:00,
59.45it/s]
2021-06-02 00:01:55 INFO rasa.core.agent - Persisted model to '/tmp/tmpj_0w3csd/core'
Core model training completed.

57
Your Rasa model is trained and saved at
'/home/maher/Documents/admission-bot/models/20210602-000157.tar.gz'.
(venv) maher@maher-PC:~/Documents/admission-bot$

5- rasa run actions log file content:

(venv) maher@maher-PC:~/Documents/admission-bot$ rasa run actions

2021-06-01 23:32:55 INFO rasa_sdk.endpoint - Starting action endpoint server...
2021-06-01 23:32:56 INFO rasa_sdk.executor - Registered function for 'action_session_start'.
2021-06-01 23:32:56 INFO rasa_sdk.executor - Registered function for 'action_colleges'.
2021-06-01 23:32:56 INFO rasa_sdk.executor - Registered function for 'action_college_major'.
2021-06-01 23:32:56 INFO rasa_sdk.executor - Registered function for
'action_tawjihi_branch_and_mark'.
2021-06-01 23:32:56 INFO rasa_sdk.executor - Registered function for 'action_major_details'.
2021-06-01 23:32:56 INFO rasa_sdk.executor - Registered function for
'validate_expected_major_form'.
2021-06-01 23:32:56 INFO rasa_sdk.endpoint - Action endpoint is up and running on
http://0.0.0.0:5055

58