See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320342533

Using Search Engine Technology to Examine Relationship between Going

Concern Opinion and Its Financial Distress Metrics

Article  in  Journal of Emerging Technologies in Accounting · October 2017

DOI: 10.2308/jeta-51933

CITATION READS

1 419

4 authors, including:

Vikram Desai Joung Kim

Shivaji University, Kolhapur Nova Southeastern University
9 PUBLICATIONS   50 CITATIONS    22 PUBLICATIONS   374 CITATIONS   

SEE PROFILE SEE PROFILE

Rajendra P. Srivastava
University of Kansas
125 PUBLICATIONS   2,179 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Decision under uncertainities View project

All content following this page was uploaded by Rajendra P. Srivastava on 29 October 2017.

The user has requested enhancement of the downloaded file.

Indian Accounting Review
Vol. 20, No. 1, June 2016

Textual Analysis and Business Intelligence in

Big Data Environment : Search Engine versus XBRL

Rajendra P. Srivastava
School of Business
University of Kansas, USA
E-mail: [email protected]

ABSTRACT
The main objective of this paper is to discuss the role of search engines in
data analytics within XBRL Environment and beyond. It is frequently argued that
having XBRL formatted business reports makes it easy to access information and
hence analyze it. However, one major problem with the XBRL formatted data is
that the analytical tools will capture only those pieces of information that are
tagged. What if the user needs information that is not tagged or not required to
be tagged? Can the analytical tools still be effective to provide the analysis needed?
This is where search engines would become important. The present paper highlights
the importance of textual analysis and demonstrates the value of search engines
that go beyond the XBRL environment, especially in situation where the needed
pieces of information are not tagged. Several examples are presented in the paper
to show the value of a search engine such as Seek iNF (https://www.seekedgar.com )
developed at the University of Kansas, in performing textual analysis and developing
predictive models, especially where no programming skills are required.

Key words: Search engine for SEC Filings, textual analysis, predictive
models, Seek iNF.

I. INTRODUCTION
The main objective of this paper is to discuss the role of search engines
in textual analysis, data analytics, and business intelligence. Traditionally,
researchers have used financial and non-financial data to build models for
analyzing financial risk and financial performance. For example, Altman (1968),
Beaver (1996), Jones and Hensher (2004), Ohlson (1980), Shumway (2001),
Zmijewski (1984) and many others have used financial data to develop models

Acknowledgements: The author would like to thank the participants of the 2014 International
Symposium on Accounting Information Systems (organized by the Department of Accounting,
University of Melbourne, Australia, January 2-3, 2014) for their valuable suggestions on an
earlier version of the paper.

1
2 Indian Accounting Review

for assessing financial risk and financial performance using the regression
technique. Emery and Cogger (1982) used cash flow data to develop a
theoretical model to measure liquidity of a company, which in turn measures
the financial risk or potential for bankruptcy. Several researchers have used
the financial data to develop neural network models to forecast bankruptcy
and fraud [e.g., see Boritz and Kennedy (1995), Boritz, Kennedy, and
Albuquerque (1995), Fanning and Cogger (1998), O’Leary (1998)]. Lensberg,
Eilifsen, and McKee (2006) have used financial data to develop bankruptcy
prediction models using genetic algorithm, while McKee (2003) used similar
data to develop bankruptcy prediction models using rough sets. Shumway
(2001) used financial data to develop a simple hazard model for bankruptcy
prediction.Thus, a vast amount of work has been done in using the financial
data to develop predictive models and business intelligence for risk assess-
ments and bankruptcy prediction.
Callen, Khan and Lu (2013) use financial data to measure "accounting
quality" where "the precision with which financial reports convey information
to equity investors about the firm's expected cash flows" defines the accounting
quality. Recently, Lee, Churyk and Clinton (2013), (see also, Loughran and
Mcdonald, 2011) developed a fraud detection model based on the counts of
key words (positive versus negative words), and punctuation in Management
Discussion and Analysis (MD&A) disclosures and found it to be performing
better than many other models developed using financial data.
While financial and certain business data are being tagged using XBRL
(Extensible Business Reporting Language), not all key words, punctuations,
and present tense and past tense sentences, and counts of words and
sentences that are needed for determining sentiments and readability indices
to develop predictive models as developed by Lee, Churyk and Clinton (2013),
are being tagged nor they would ever be tagged by XBRL technology. While
there are many advantages of the XBRL technology from ease with which
information can be retrieved, shared, transported across platforms, used in
automated models, to cost saving and eliminating human errors in the above
process (e.g., see, Srivastava 2009 and Debreceny et al. 2005), it is practically
not possible to tag all words in a report through the XBRL technology. This
is where the need of a text search engine becomes important, which is the
topic of this paper. Text mining is a big area and an important area, especially
in this era of big data. There are several books written on this topic. The focus
of this paper is concentrated towards searching for financial and non-financial
information, along with exploring the use of text analytics for predictive models
in the context of accounting, auditing, and finance.
The Securities and Exchange Commission (SEC) in the USA and many
similar government agencies in other countries (e.g., Securities and Exchange
Board of India, SEBI) have mandated public companies to file their annual and
quarterly reports in the XBRL format along with the text/html filings. Thus,
all business relevant data, both financial and non-financial, are available in
the XBRL tagged format. And therefore, it would be easy to create models that
would provide timely information to analysts, regulators, and the general public
Srivastava 3

for business decisions as soon as the financial reports are filed with the
regulatory agencies such as the SEC in the USA and SEBI in India. In fact,
already several vendors have developed such analytical tools that use XBRL
tagged data. Based on the availability of XBRL tagged business information
from the SEC filings, the SEC has built a "RoboCop" (Novack, Carney, Harker
2013) to find irregularities in companies' filings. However, since the XBRL
technology does not tag all the words, it is difficult to develop a model that
would use the tagged financial and non-financial data along with the certain
key words and number of or percentage of positive versus negative words
in a document such as 10-K, the annual report filed with the SEC. In fact,
SEC is looking into integrating text-based information into their "RoboCop"
(Novack, Carney, Harker 2013), which is based on the tagged data.
Rest of the paper is divided into four sections. Next section provides a brief
discussion on the background research, especially the textual analysis
research. Section III describes the features of the search engine, Seek iNF,that
has been developed at the University of Kansas. Section IV provides
illustrations on how Seek iNF can be used to perform textual analysis without
any programming skills in PERL or Python. Finally, Section V provides a
summary and conclusion.

II. BACKGROUND ON TEXTUAL ANALYSIS RESEARCH

In this section, we briefly describe some of the important researches
in the textual (i.e., content) analysis domain and show the importance of such
work not only in developing predictive models for assessing financial risk,
business risk, fraud risks, etc., but also in developing business intelligence
to understand the quality of financial reporting and financial disclosures.
Companies disclose certain piece of information, some voluntarily and some
by mandate, in their management report to communicate to the public about
what they have accomplished and what they plan to achieve in the future. The
language how they communicate to public and regulatory agencies may vary
from when the company is profitable versus when the company is having
financial distress. Li (2010) provides a good review of the textual analysis
related to corporate disclosures. Disclosures are becoming complex, as
mentioned earlier, some due to regulatory requirements and some due to
management intention as stated by Li (2010, p. 144):
. . . managers' communication patterns could reveal certain manage-
rial characteristics and thus have significant implications for under-
standing management decisions. Recent developments in behavioral
economics emphasize the cognitive biases of human beings and the
roles of these biases in decision making (Kahneman, 2003). . . As a
communication vehicle for management, textual disclosures can
provide a means for researchers to assess managers' behavioral biases
and understand firm behavior.
4 Indian Accounting Review

In a recent article, Monga and Chasan (2015) find that :

Companies are spending an increasing amount of time and energy
beefing up their regulatory filings to meet disclosure requirements.
The average 10-K is getting longer-about 42,000 words in 2013,
up from roughly 30,000 words in 2000.
Tetlock, Saar-Tsechansky, and Macskassy (2008) examine whether the use
of a simple quantitative measure of language can predict individual firms'
accounting earnings and stock returns. They mention the following two
advantages of using such a textual analysis (page 1438).
First, by quantifying language, researchers can examine and judge the
directional impact of a limitless variety of events, whereas most studies
focus on one particular event type, such as earnings announcements,
mergers, or analysts' recommendations. Analyzing a more complete set
of events that affect firms' fundamental values allows researchers to
identify common patterns in firm responses and market reactions to
events. Equally important, examining all newsworthy events simulta-
neously limits the scope for "dredging for anomalies"- the phrase used
by Fama (1998) to describe running event studies on different types
of events until one obtains "significant" results.
Second, linguistic communication is a potentially important source of
information about firms' fundamental values. Because very few stock
market investors directly observe firms' production activities, they get
most of their information secondhand. Their three main sources are
analysts' forecasts, quantifiable publicly disclosed accounting vari-
ables, and linguistic descriptions of firms' current and future profit-
generating activities. If analyst and accounting variables are incom-
plete or biased measures of firms' fundamentals, linguistic variables
may have incremental explanatory power for firms' future earnings and
returns.
Loughran and Mcdonald (2011) analyze a large sample of 10-Ks during
1994-2008, and show that almost three-fourths of the word count identified
as negative by the Harvard Dictionary do not have negative meaning in a
financial context. In addition, they developed an alternative negative word list
that better reflects the tone of financial text along with five other word
classifications (positive, uncertain, litigious, strong modal and weak modal).
They linked the word lists to 10-K filing returns, trading volume, subsequent
return volatility, fraud, material weakness, and unexpected earnings.
Lee, Churyk and Clinton (2013, p. 35) state that "Conventional fraud
detection measures using ratio analysis and other financial data were either
unable to detect the fraud or unable to detect it soon enough to avoid
catastrophic outcomes". They develop a fraud detection model based on the
counts of key words (positive versus negative words), and punctuation in
Management Discussion and Analysis (MD&A) disclosures and found it to be
performing better than many other models developed using financial data.
Srivastava 5

In a preliminary investigation, they found that companies that were involved

in fraudulent financial reporting used on average fewer negative words
compared to similar companies. Their early fraud detection model based on
the textual (i.e., content) analysis of MD&A in 10-K is:
FRAUDi = 2.89757 – 0.83408 (POSITIVE EMOTIONi) – 0.48315
(PRESENT TENSEi) + .0001 (TOTAL WORDSi) – 2.80753(COLONSi)
While Lee, Churyk and Clinton (2013) validate their model empirically, the
model makes logical sense and conveys about managements' behavior when
they commit fraud. It seems the Management is trying to misguide the users
of the financial report and hide the fraud by using more words, fewer present
tense, less 'positive emotion' and fewer colons in the management discussion
and analysis part of 10-K.
Li, Lundholm, and Minnis (2013) develop a model to compute management's
perception of the intensity of competition using textual analysis of firms' 10-
K filings. They find that the measure of competition varies both across-industry
and within-industry, and also show that each measure is related to the firm's
future rates of diminishing marginal returns. Their measure is based on the
count of the number of occurrences of words "competition, competitor,
competitive, compete, competing," including those words with an "s" appended,
less any case where "not," "less," "few," or "limited" precedes the word by three
or fewer words. Their measure of competition is:
PCTCOMP = 1000*NCOMP/NWORDS,
where NCOMP = number of words in 10-K as described above and
NWORDS = Total number of words without numbers. They wrote their own
program to count the occurrences of these words and phrases because there
was no database that would provide such counts. However, recently, the search
engine, Seek iNF, developed at the University of Kansas, has several built-in
features that provides such counts without any required programming skills
(for details see the web site https://www.SeekEdgar.com:8443/ ). This will be
further elaborated in the next section.
One can develop predictive models and business intelligence using the
textual analysis of a report along with the financial and other non-financial
data for assessing risks such as business risk, litigation risk, and the risk of
financial distress. But to do that one needs effective text mining tools and search
engines. There are several such tools available in the market place from providers
such as LexisNexis (http://www.lexisnexis.com/hottopics/lnacademic/) and
IBM (http://www-03.ibm.com/software/products/en/category/content-
analytics). However, rather than describing all different search engines and text
analyzing systems, I describe, next, the search engines, Seek iNF, that provides
tremendous opportunity not only to perform textual analyses and develop
predictive models for assessing various risks, such as, financial risk, litigation
risk, and fraud risk, but also to create new knowledge.
6 Indian Accounting Review

III. SEEK iNF – A TOOL FOR INFORMATION

RETRIEVAL AND TEXTUAL ANALYSIS
As discussed in the previous section, textual analysis of annual reports
and other companies' filings is playing an important role in developing
predictive models and business intelligence. However, the lack of Perl and
Python programming skills in business faculty has restricted the research
in textual analysis to very few, as evidenced by the published research.
For example, Li, Lundholm, and Minnis (2013) wrote their own program to
extract the information they needed to conduct their research.
In this section, I plan to describe a search engine, Seek iNF (Search Engine
for Extracting Knowledge from Industrial Filings) that does not require any
programming skills to extract information from any report or filing whether it
is in the traditional text format, or any other format such as HTML, ASCII,
PDF or XBRL. Seek iNF is an outgrowth of a project FRAANK (Financial
Reporting and Auditing Agent) that started in 1996 at the Ernst & Young Center
for Auditing Research and Advanced Technology in the School of Business at
the University of Kansas (see Bovee et al. 2005 for details about FRAANK). The
basic idea was to develop an intelligent system that will fetch a report from
a website and parse the text to extract the desired information. FRAANK
succeeded, in no time, in fetching 10-K (annual reports) and 10-Q (quarterly
reports) of the US public companies from the SEC Edgar database and parsing
the financial information from the balance sheet, income statement, and cash
flow statement and dumping it into MS Excel Spreadsheet. The University of
Kansas received the US Patent for FRAANK in 2010.
FRAANK did not have any intelligence in terms of getting specific piece
of information from the reports. Once FRAANK became functional, my
colleagues started requesting for specific piece of information such as 'a list
of all the companies that have mentioned this issue … in the footnote in
10-K'. We will tweak the program in FRAANK and run it on the weekend and
have the data ready by Monday. Next, an idea was suggested that why not
we develop a system where users query the system and get the data they want
on their own without requesting us. This is when Seek iNF was born in 2007.
Seek iNF is a search engine based on the Cloud technology. There is no
need to have any programming skills to extract information from a report that
is in the database. Currently, Seek iNF uses its own database of SEC Filings
and Public Company Accounting Oversight Board (PCAOB) inspection reports
that are pre-processed for efficient retrieval of information. At the present, this
database consists of 14 million filings and 20 million documents for the years
1994 – 2016 and daily updated automatically (see Table 1 for details ). It has
built-in features for identifying exact phrases containing words like articles and
prepositions. In addition, it has built-in Boolean logic, counters of words,
phrases, and sentences, and provides distribution of all the words in a
document and six readability indices, all without any need or programming.
Srivastava 7

TABLE 1

List of Filings and Exhibits in the Seek iNF Database

File Type Or Exhibit Type Mean
AAERS Accounting and Auditing Enforcement Releases
6-K, 6-K/A Current report of foreign issuer pursuant to Rules
13a-16 and 15d-16 Amendments
8-K, 8-K/A Current report filing (Disclosure of significant eco-
nomic events), and amendment
10-K, 10-K/A Annual report pursuant to Section 13 and 15(d), and
amendment
10-KT, 10-KT/A Annual-Transition report pursuant to Rule 13a-10
or 15d-10, and amendment
10-K405, 10-K405/A Annual Reports – Reg. S-K Item 405, and amend-
ment
10-KT405, 10-KT405/A Annual-Transition Reports - Reg. S-K Item 405, and
amendment
10-KSB, 10-KSB/A Annual Reports by Small Businesses, and amend-
ment
10-KSB40, 10-KSB40/A Annual Reports by Small Businesses – Reg. S-B Item
405, and amendment
10-Q, 10-Q/A Quarterly report pursuant to Section 13 or 15(d) and
amendment
10-QT, 10-QT/A Quarterly-Transition report pursuant to Rule 13a-10
or 15d-10 and amendment
10-QSB, 10-QSB/A Quarterly Reports by Small Businesses and amend-
ment
13D,13D/A (Schedule) Schedule filed to report acquisition of beneficial
ownership of 5% or more of a class of equity
securities
13G, 13G/A (Schedule) Schedule filed to report acquisition of beneficial
ownership of 5% or more of a class of equity
securities by passive investors and certain institu-
tions
13F-HR, 13F-HR/A Initial Quarterly Form 13F Holdings Report filed by
institutional managers. Initial Quarterly Form 13F
Combination Report filed by institutional managers
13F-NT, 13F-NT/A Initial Quarterly Form 13F Notice Report filed by
institutional managers
20-F, 20-F/A Annual and transition report of foreign private
issuers pursuant to Section 13 or 15(d)
40-F, 40-F/A Annual reports filed by certain Canadian issuers
pursuant to Section 15(d) and Rule 15d-4
424B1 - 424B8 Prospectus filed pursuant to Rule 424(b)(1) – Rule
424(b)(8)
8 Indian Accounting Review

TABLE 1 (Contd.)

File Type Or Exhibit Type Mean

N-CSR, N-CSR/A Certified annual shareholder report of registered
management investment companies filed on Form
NCSR
N-CSRS, N-CSRS/A Certified semi-annual shareholder report of regis-
tered management investment companies filed on
Form NCSR
CORRESPRESPONSE A correspondence can be sent as a document with
another submission type or can be sent as a
separate submission. Response by SEC to the above
correspondence
DEF 14A, DEFA 14A Definitive proxy statements and amended proxy
statement
Exhibit-21 (10-K) List of Subsidiaries
Exhibit-95 (10-K) Mine Safety Disclosure
Form 3, Form 3/A Registration statement for specified transactions by
(F-3, F-3/A) certain foreign private issuers
Form 4, Form 4/A Registration statement for securities issued by
(F-4, F-4/A) foreign private issuers in certain business combina-
tion transactions
Form 5, Form 5/A Annual statement of changes in beneficial ownership
(F-5, F-5/A) of securities
S-1, S-1/A General form of registration statement for all com-
panies including face-amount certificate companies
PCAOB PCAOB reports inspecting registered public account
Inspection Reports ing firms to assess compliance with the Sarbanes-
Oxley Act, the rules of the Board, the rules of the
Securities and Exchange Commission, and profe-
ssional standards, in connection with the firm's
performance of audits, issuance of audit reports, and
related matters involving U.S. companies, other
issuers, brokers and dealers.
PCAOB Board Orders in Settlements with Registered Firms
Settled Disciplinary Orders or their Associated Persons
PCAOB Adjudicated Opinions, Orders, and Other Final Board Action
Disciplinary Actions Imposing Sanctions in Contested Disciplinary
Proceedings and SEC Actions on Review of those
Sanctions

Basically, Seek iNF deals with the following four Dimensions: (1) Search
all or few documents in the database for specific issues or concerns, (2) Obtain
a piece of information whether financial or non-financial, (3) Perform text
analytics, and (4) Download all the searched data in HTML document and Excel
Spreadsheet file for further analysis. In the next section, I will describe these
Srivastava 9

dimensions in detail with examples. These four dimensions along with the
following Power Features, Seek iNF provides unique opportunity to researchers
to perform textual analysis and develop predictive models and business
intelligence in an efficient and effective manner which has not been possible
because of lack of programming skills in researchers except a few.
Seek iNF has several Power Features that make the search and retrieval
of information much more efficient and effective which has never been possible
without the knowledge of programing by researchers. These features are
discussed next.
Search Documents with Multiple Exact Phrases: This feature allows
users to search all or few documents in the database for presence or absence
of multiple phrases containing words like 'a', 'in', 'if', 'no', 'of', 'the', etc.
Such a feature makes it easier to find which of the companies have mentioned
in their 10-K a given phrase, say 'wrongful termination'. One can input this
phrase in the search engine and get 4,511 companies in few seconds that have
mentioned the phrase in their 10-Ks. I will elaborate on this feature further
through examples in the next section. Using the built-in Boolean logic makes
the search process much more effective, especially using the multiple exact
phrases.
You can search any document with a combination of exact phrases with
"AND", "OR", or any combination of these conditions. Seek iNF uses "+" for AND,
"|" for OR, and "-" for negation with or without space before and after the
symbol. Here are examples of how one would type the three phrases represented
by A, B, & C in the exact phrase slot in Step 1 (see Figure 1 ).

FIGURE 1
Screenshot of the main page of Seek iNF
10 Indian Accounting Review

l For AND logic, type the phrases separated by + sign as: A + B + C

l For OR logic, type the phrases separated by | symbol as: A | B | C
l For AND and OR, such as A & (B or C), type A + (B | C)
l For AND andnegation such as "A & B & not C", type as A + B – C
l For OR and Negation such as "(A or B) & not C", type as (A | B) – C

Information Retrieval with Proximity Search: This feature searches

documents for information with two words within any number of words. The
number of words between two words could be anywhere between two to several
thousands. This is very useful in searching for information; especially when
the two words of interest are separated by few words in the text. For example,
suppose you are interested in finding executive bios. This information is usually
published in the proxy statement (DEF 14A) usually in a table. In order to get
this table, one can use the Proximity search feature with two words 'name'
and 'age' within two words. You will get all the tables in a few seconds. This
example is further elaborated in the next section.
Information Retrieval with few words before and few words after a
phrase: This feature provides users a tool to obtain a unique piece of
information along with the traditional data that are available through a
database. For example, one can find addresses of companies or fiscal year end
using this feature, which is not currently available through any database
provider. Several examples are used to illustrate this feature in the next section.
Counts of Multiple words, phrases, and sentences: This feature provides
a tool to gather counts of multiple words, multiple phrases, and sentences,
along with distribution of all the words. This kind of information is useful in
textual analysis and developing predictive models and business intelligence.
This will be further elaborated in the next section through examples.
Counts Occurrence of two words within few words: This feature is
useful again in developing models for business intelligence. For example, one
can use this feature to compute the metric measuring competition as developed
by Li et al. (2013). I will further elaborate this feature through examples in
the next section.
Readability Indices: Seek iNF has been programmed to provide the
following six readability indices: Gunning-Fog Index, Smog Index, Flesch
Reading Ease, Flesch-Kincaid Grade Level, Automatic Readability Index,
Coleman-Liau Index (see Table 2 for definition), for all the documents present
in the Seek iNF database. This means these measures are available for all the
SEC Filings and PCAOB Inspection Reports (about 20 million documents).
With this feature one can perform textual analysis on all the filings and
documents without knowing any programming.
Srivastava 11

TABLE 2

Readability Indices
Readability Index Formula
Gunning-Fog Index 0.4x[(words/sentences) + 100x(complex words)/
words]a
Smog Index 1.0430xSQRT[(number of polysyllables)x30/
(number of sentences)] + 3.729b
Flesch Reading Ease 206.835 – 1.015[(total words)/(total sentences)]
– 84.6[(total syllables)/(Total words)]c
Flesch-Kincaid Grade Level 0.39[(total words)/(total sentences)] + 11.8[(total
c
syllables)/(total words)] – 15.59
Automatic Readability Index 4.71(characters/words) + 0.5(words/sentences) –
21.43 d
e
Coleman - Liau Index 0.0588L – 0.2965S – 15.8

a. https://en.wikipedia.org/wiki/Gunning_fog_index. A "complex word" is

defined as the word with three or more syllables. Do not include proper
nouns, familiar jargon, or compound words. Do not include common
suffixes (such as -es, -ed, or -ing) as a syllable.
b. https://en.wikipedia.org/wikiSMOG.
c. https://en.wikipedia.org/wiki/Flesch%E2%80%93 Kincaid_read-ability
_tests.
d. https://en.wikipedia.org/wiki/Automated_readability_index
e. https://en.wikipedia.org/wiki/Coleman%E2%80%93Liau_index. L is the
average number of letters per 100 words and S is the average number
of sentences per 100 words.

Download all the searched data in Excel Spreadsheet: This feature

allows users to download all the output information in a CSV file and also in
an HTML file, irrespective of whether the searched output is in the form of
HTML snippets, or tables, or whether it is the counts of multiple phrases/
words. A CSV file can be opened in Excel and thus makes it easier to perform
statistical analysis of the output data.
In addition to the above features, Seek iNF provides the frequency
distribution of all the words in a document which can be downloaded in an
Excel file for textual analysis. The users can search for specific piece of
information by companies' names or CIKs (Central Index Keys by SEC), or by
SIC (Standard Industrial Classification). Also, one can search for information
by specific resolution such as "paragraph", "Table", "Footnote" "Audit Report",
"SOX404 Management Report", "SOX 404 Audit Report" and "MD&A".
12 Indian Accounting Review

IV. ILLUSTRATIONS
In this section, I demonstrate the power of the search engine, Seek iNF,
in research, especially in accounting, finance, and other business
disciplines.Given the ease with which one can gather information using Seek
iNF, the knowledge, imagination, and creativity of the researcher will play an
important role in future research.

Example 1: Searching for documents containing the exact phrases

Did you know that there are public companies that make statements in
their annual reports (10-Ks) that they were not profitable and may never be
profitable and still investors invest money? Why is that? How would you find
out who they are? Manually, it is impossible! But, Seek iNF can answer this
question in seconds. Insert the following phrases "We may never become
profitable | We May Never Achieve or Sustain Profitability | we may never
achieve or maintain profitability | We may never be profitable" in the exact
phrase slot, and select "all companies", the time period, and 10-Ks, and get
the result in seconds (see Figure 1). Figure 2 shows that such companies
started around 1998 and has been growing in number since then. One can
explore several questions as to not only who these companies are but to how
can they survive so long even after making such a statement. Some have
survived 10-12 years.

FIGURE 2

Graph of Percentage of Annual Reports with ‘‘Never Profitable’’and

Going Concern Opinion by Year

Another interesting question comes to mind is what type of audit opinions

were issued by the corresponding auditors. You would expect all such
companies should have received a Going Concern audit opinion. This is easy
Srivastava 13

to find out using Seek iNF. Just insert the phrases pertaining to 'we may never
be profitable' and pertaining to 'Going Concern opinion' together in the slot
of "With exact phrase" in Step 1. More specifically, type the following phrases
"(we may never become profitable | we may never achieve or sustain
profitability | we may never achieve or maintain profitability | we may never
be profitable) + (In my opinion | In our opinion) + (substantial doubt about|
substantial doubts about | substantial doubts regarding | substantial doubt
regarding) + going concern1" in the "With exact phrase" slot and the two words:
"opinion concern", in the "Proximity Search" slot within 500 words and select
"all companies", the time period, and 10-Ks. Figure 2 shows an interesting
result - not all such companies received GC opinions. Actually, less than 50%
have received a going concern opinion. Why is that? What about audit quality?
Many interesting research questions emanate from this result.

Example 2: Proximity Search

This feature also provides a powerful tool to capture useful information
that is not available from any other source. Here I will demonstrate two
examples that I have been asked frequently. One deals with executives'
compensation and the other deals with executives' bios. My purpose here is
to demonstrate the process. As a researcher one needs to identify what
combinations of words and phrases will give exactly what one wants for one’s
research using this feature or in combination of other features.
2(a). Find Executive Compensation Tables. The executive compensation
data are presented in a table in the proxy statement (DEF 14A). After looking
at few compensation tables, I found out that the two words, 'salary' and 'bonus',
were appearing within two words in almost all the tables in DEF 14As. Suppose,
I want to obtain the executive compensation data for all the companies for the
year 2015. Here is what I needed to do. Type in "salary bonus" in 'Proximity
search', within 2 words, select "All" companies, the year 2015, 'Table' as the
resolution, and DEF 14A as the filings, and submit the query. In few seconds,
the system will return the result, with a displayof first 20 companies on the
left of the display window. In the middle, the system will show the compensation
table of the first company. Figure 3 is a screenshot of the compensation data
displayed by the systemin the middle for one of the companies. One can
download the compensation table displayed in the middle in the Excel
Spreadsheet by selecting "Download Table in CSV" from the menubar on the
top right side of the display window. One can download the entire searched
result, i.e., all the tables in the CSV format by submitting a Request Form.
The system will automatically send you the link for the data when it is
ready (see the details at: https://www.seekedgar.com:8443/SeekiNF_search_
Engine.pdf).

1 Seek iNF uses '|' symbol for OR logic, '+' symbol for AND logic, and '–' for negation.
14 Indian Accounting Review

FIGURE 3

Screenshot of Compensation Table

2(b). Find Executive Bios from DEF 14A. Similar to the previous example
of executive compensation, executive bios are in a table in the proxy statement
(DEF 14A). Again one needs to find the pattern, which will identify such a table.
After looking at few tables that contained names and ages of executives, I found
out that 'Name' and 'Age' were appearing within two words almost in all such
tables. Next, I used the two words 'Name' and 'Age' in "Proximity search" within
2 words to get all the tables that contained executives' bios for all the
companies. Figure 4 is a screenshot of the display window for one of the
companies. One can download this table in Excel by selecting "Download Table
in CSV" from the menu bar displayed on the top right side of the display
window. One can download the entire search result in Excel by submitting a
Request Form and filling out the search criteria again. The system will
automatically process the request and inform the user with the link when the
data are ready to be downloaded.
Srivastava 15

FIGURE 4

Screenshot of Seek iNF Output of Executive Bios

Example 3: Search with few words before and few words after a phrase
This feature of searching for information with few words before and few
words after a phrase is again a pretty powerful tool for getting information that
is not available by any other source unless you know how to program in PERL
or Python to fetch that information. Suppose you want to find the "fiscal year
end" of all the public companies filing with the SEC. This information is
provided by the companies in their 10-Ks, the annual reports filed with the
SEC. Looking at few 10-Ks, it seems that this information is listed right after
the phrase "fiscal year end". Thus, if one uses the search criteria "fiscal year
end" in"With exact phrase" and selects the option to display the searched data
with zero word before and 1 word after, the system will display the desired
result. Figure 5 represents the "fiscal year end" data in an Excel file for a set
of companies obtained through the process of submitting a "Request Form"
(https://www.seekedgar.com:8443/SeekiNF_search_Engine.pdf ).

FIGURE 5

‘‘Fiscal year end’’ data in Excel for a given set of Companies

using the feature ‘‘Proximity search’’ and ‘‘Zero Word Before and
One Word After’’

Example 4: A Measure of Competition Based on Textual Analysis of 10 K

As mentioned earlier, Li, Lundholm, and Minnis (2013) compute
management's perception of the intensity of the competition using textual
analysis of the firm's 10-K filing. They count the number of occurrences of the
following words: competition, competitor, competitive, compete, competing,
including those words with an "s" appended, and then remove any case where
"not," "less," "few," or "limited" precedes the word by three or fewer words.
16 Indian Accounting Review

They used their own programming skills to get these counts. However, I want
to show here how one can get the required counts easily using the Seek iNF
search capabilities without having any knowledge of PER or Python. As given
earlier, their measure of competition is
PCTCOMP = 1000*NCOMP/NWORDS,
where NCOMP = number of words in 10K as described above and NWORDS =
Total number of words without numbers. Seek iNF yields these counts in no
time. We obtain NCOMP in two steps. First, we count the occurrence of the
words: competition, competitor, competitive, compete, competing, competitions,
competitors, competes, and subtract from it the "Proximity" count which counts
the occurrence of the following two words within three or less words:
not competition, less competition, few competition, limited competition,
not competitor, less competitor, few competitor, limited competitor,
not competitive, less competitive, few competitive, limited competitive,
not compete, less compete, few compete, limited compete, not competing,
less competing, few competing, limited competing, not competitions,
less competitions, few competitions, limited competitions, not competitors,
less competitors, few competitors, limited competitors, not competes,
less competes, few competes, limited competes.
Let us compute PCTCOMP for the following five companies: Qwest Corp,
Verizon Communications Inc, AT&T Inc., Level 3 Communications Inc., General
Communication Inc., with CIKs: 68622, 732712, 732717, 808461, 794323, for
10 years (2006-2015). Since "Request Form" will allow you to download only
five years data at a time, you need to submit two separate requests. After typing
your name, email and University/Company, select the menu item "Phrase(s)/
Word(s) Count" for the words counts, and "Proximity Count" for the second part
of the count to be subtracted from the first count to determine NCOMP. Seek
iNF provides the total word count as a default. Thus, we can easily calculate
the competition metric PCTCOMP.
Srivastava 17

FIGURE 6

Changes in the Competition Metric for Five Companies (ATT&T Inc.,

General Communication Inc., Level 3 Communications Inc.,
Qwest Corp, and Version Communication Inc.)
for the Years 2005–2015

Figure 6 depicts a graph of the competition metric, PCTCOMP, for five

companies in the telecom industry. As you can see Level 3 has had very little
variation in its measure of competition where as for Verizon and AT&T it varies
a lot during this period. There are a lot of research questions that one could
explore about the economic reasons for these variations in the competition
metric for individual companies. I do not explore these questions and try to
find answers. I recommend this to be future research questions. One can
perform all kinds of textual analyses and readability analyses on all the
14 million SE Filings and 20 million documents in a matter of minutes.

Example 5: Unique Financial Information from 10K

There are many unique financial line items that are not available in any
of the current databases. For example, suppose you want to find "Air Traffic
Liability" for airline companies for the year 2015. Here is what one can do to
get this information through Seek iNF:Step1: Type the phrase Air Traffic
Liability in the slot "With the exact phrase". Step 2: Since we want all the
companies, leave the default choice "All" and select the desired time frame
2015-2015.Step 3: Select "Table". Step 4: Select 10-K. Step 5: Click on SUBMIT
button.
You can download all the data in Excel Spreadsheet two ways, one directly
from the display window by selecting the menu item "Display Table in CSV"
just above the display window on the right for each company, and the other
18 Indian Accounting Review

by submitting a "Request Form" as done in Example 4. Figure 7 is a screenshot

of the display window for one company. One can download this table in Excel
too as mentioned above.

FIGURE 7

Unique Line Item, ‘‘Air traffic Liability’’ in Balance Sheet

V. SUMMARY AND CONCLUSION

This paper discusses the role of search engines in textual analysis and
in developing predictive models and business intelligence in a big data
environment and within XBRL environment. The paper first provides a
summary of the background research on textual analysis and predictive models
already developed and their importance in the current environment. It seems
if managements want to misguide the users of the financial reports and hide
the fraud, they would use more words, fewer present tense, less 'positive
emotion' and fewer colons in the management discussion and analysis part
of the 10-K as observed by Lee, Churyk and Clinton (2013). Although XBRL
technology is being required by regulators globally for financial reporting
purposes because of certain advantages, this technology lacks in providing the
kind of data needed for textual analyses. Thus, search engines specifically
designed to search textual information become important. Seek iNF is such
a search engine that has been developed at The University of Kansas and is
being further developed by SeekEdgar LLC (see https://www.seekedgar.com
for details). Several examples have been discussed to illustrate the power of
Seek iNF.
The text parsing search engines with built-in intelligence for searching for
specific piece of information, financial and non-financial, is going to play an
important role in future research for developing predictive models and business
intelligence. I would emphasize that the knowledge, imagination and creativity
of the researcher would drive the future research; not the pre-canned data.
Srivastava 19

References

Altman, E. (1968). Financial ratios, Discriminant Analysis and the Prediction of Corporate
Bankruptcy.Journal of Finance 23: 589-609.
Atiya, A. F. (2001). Bankruptcy Prediction for Credit Risk Using Neural Networks: A Survey
and New Results. IEEE Transactions on Neural Networks, Vol. 12, No. 4, July:
929-935.
Beaver, R. (1996). Financial ratios as predictors of failure.In Empirical Research in
Accounting: Selected Studies 1966, Journal of Accounting Research, vol. 4,
pp. 71-111.
Boritz, J. and D. Kennedy. (1995). Effectiveness of neural network types for prediction of
business failure.Expert SystemsApplications, vol. 9: 504-512.
Boritz, J., D. Kennedy, and A. Albuquerque. (1995). Predicting corporate failure using a neural
network approach.Intelligent System in Accounting, Finance, and Management,
Vol. 4: 95-111.
Bovee, M., A. Kogan, R. P. Srivastava, M. A. Vasarhelyi, K. M. Nelson, (2005). Financial
Reporting and Auditing Agent with Net Knowledge (FRAANK) and eXtensible Business
Reporting Language (XBRL). Journal of Information Systems, Vol. 19, No. 1 (Spring):
pp. 19-41.
Callen, J. L. M. Khan and H. Lu. (2013). Accounting Quality, Stock Price Delay, and Future
Stock Returns. Contemporary Accounting Research, Vol. 30, Issue 1, Spring:
269-295.
Debreceny, R. S., A. Chandra, J. J. Cheh, D. Guithues-Amrhein, N. J. Hannon, P. D.
Hutchison, D. Janvrin, R. A. Jones, B. Lamberton, A. Lymer, M. Mascha, R. Nehmer,
S. Roohani, R. P. Srivastava, S. Trabelsi, T. Tribunella, G. Trites, and M. A.
Vasarhelyi. (2005). Financial Reporting in XBRL on the SEC's EDGAR System:
A Critique and Evaluation. Journal of Information Systems 19 (2):191-210.
Emery, G. and K. Cogger. (1982). The Measurement of Liquidity. Journal of Accounting
Research, Vol. 20, No. 2 Pt. I Autumn: 290-303.
Fanning, K. M. and K. O. Cogger. (1998). Neural network detection of management fraud using
published financial data. International Journal of Intelligent Systems in Accounting,
Finance & Management, Vol. 7, No. 1, March: 21-41.
Jones and Hensher. (2004). Predicting Firm Financial Distress: A Mixed Logit Model.
The Accounting Review, Vol. 79, No. 4, 2004, pp. 1011-1038.
Lee, C., N. T. Churyk, and B. D. Clinton. (2013). Detect Fraud Before Catastrophe. Strategic
Finance, March: 33-37.
Lensberg, T., A. Eilifsen, and T. E. McKee. (2006). Bankruptcy theory development and
classification via genetic programming. European Journal of Operational Research,
Vol. 169, Issue 2, March: 677-697.
Li, F. (2010). Textual Analysis of Corporate Disclosures: A Survey of the Literature. Journal
of Accounting Literature, Vol. 29: 143-165.
Li, F., R. Lundholm, and M. Minnis. (2013). A Measure of Competition Based on 10-K Filings.
Journal of Accounting Research, Vol. 51, No. 2 (May): 399-436.
Loughran, T. and B. Mcdonald. (2011). When is a Liability not a Liability? Textual Analysis,
Dictionaries, and 10-Ks. The Journal of Finance, Vol. 6, Issue 1, February: 35-65.
McKee, T. E. (2003). Rough sets bankruptcy prediction models versus auditor signaling rates.
Journal of Forecasting, Vol. 22, Issue 8, December: 569-586.
Monga, V. and E. Chasan. (2015). The 109,894-Word Annual Report: As regulators require
more disclosures, 10-Ks reach epic lengths; how much is too much? The Wall Street
Journal, Business CFO Journal. Updated June 1.
Novack, J., J. Carney, and F. Harker. (2013). How SEC's New RoboCop Profiles Companies
For Accounting Fraud.http://www.forbes.com/sites/janetnovack/2013/08/09/
how-secs-new-robocop-profiles-companies-for-accounting-fraud/
 20 Indian Accounting Review

Ohlson, J. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal
of Accounting Research 18: 109-131.
O'Leary, D. (1998). Using Neural Networks to Predict Corporate Failure. International Journal
of Intelligent Systems in Accounting, Finance & Management, Vol.7, pp. 187-197.
Shumway, T. (2001). Forecasting Bankruptcy More Accurately: A Simple Hazard Model.
The Journal of Business, Vol. 74, No. 1, January: 101-124.
Srivastava, R. P. (2005). Financial Reporting in XBRL on the SEC's EDGAR System: A Critique
and Evaluation. Working Party of the AAA Information Systems and Artificial
Intelligence/Emerging Technologies Section (with 17 other members, with equal
participation). Journal of Information Systems, Vol. 19, No. 2, Fall: 191-210.
Srivastava, R. P., (2009). XBRL (Extensible Business Reporting Language): A Research
Perspective. Indian Accounting Review, Vol. 13, No. 1, pp. 14-32.
Tetlock, Paul C., M. Saar-Tsechansky, and S. Macskassy, (2008). More than words:
Quantifyinglanguage to measure firms' fundamentals, Journal of Finance 63,
1437-1467.
Zmijewski, M. (1984). Methodological issues related to the estimation of financial distress
prediction models. Journal of Accounting Research 22 (Supplement): 59-82.

28TH ASIAN-PACIFIC CONFERENCE

ON INTERNATIONAL ACCOUNTING ISSUES

Maui, Hawaii, USA

November 6– 9, 2016

Home Page: www.apconference.org

For more information, please contact the Conference Headquarters:

Dr. Ali Payvandi, Conference Chairman, [email protected]

Crystal Cui, Conference Program Coordinator, [email protected]

Asian-Pacific Conference on International Accounting Issues

Craig School of Business
California State University-Fresno
5245 North Backer Avenue, M/S PB7
Fresno, California 93740-0007, USA

Tel: (559) 278-4723

Tel: (559) 278-2602
Fax: (559) 278-7838

View publication stats