Applying Soft Computing Techniques in Information Retrieval

International Journal of Advanced Engineering, Management and Science (IJAEMS) [Vol-4, Issue-5, May- 2018]
https://dx.doi.org/10.22161/ijaems.4.5.8 ISSN: 2454-1311
www.ijaems.com Page | 386
Applying Soft Computing Techniques in
Information Retrieval
Namrata Nagpal
Amity Institute of Information technology, Amity University, India
Email: nnagpal@lko.amity.edu
Abstract— There is plethora of information available over
the internet on daily basis and to retrieve meaningful
effective information using usual IR methods is becoming a
cumbersome task. Hence this paper summarizes the
different soft computing techniques available that can be
applied to information retrieval systems to improve its
efficiency in acquiring knowledge related to a user’s
query.
Keywords— fuzzy logic, information retrieval, IR models,
neural networks, soft computing.
I. INTRODUCTION
There is an urgent requirement of effective Information
Retrieval Systems(IRS) has aroused with the quick
growing development of the Internet, and plethora of
availability of online text based information. The goal of
an IR System is to retrieve relevant information regarding
a user's query.
The efficiency of an Information Retrieval System can be
calculated using parameters which are prime in order to
meet the requirement of the system and accomplish its
goal instantly. The keywords put by the user while
forming a query as per his requirement is mostly vague
and uncertain. Also the document representation as well
as the process by which the query is matched to the
document is also uncertain. Hence the effectiveness of an
Information Retrieval System crucially depends upon the
system's capability to deal with the vague and uncertain
retrieval process.
There are various techniques utilized for above
efficiency; we majorly focus on the significance of soft
computing related to artificial neural networks, fuzzy
logic, genetic algorithms, or rough sets, to name a few.
The paper states about the significance of information
retrieval and various IR models in section 2. Section 3
deals with soft computing techniques available whereas
Section 4 specifies the application of soft techniques to
improve the information retrieval process.
II. INFORMATION RETRIEVAL
Information retrieval refers to a task of finding the relevant
documents from given a set of documents and applying a
user query. Information retrieval applications also require
characteristics like accuracy, speed, consistency and ease
of use in recovering relevant documents that satisfy user
queries. Some required characteristics for efficient
information retrieval are as under [2]:
 Accuracy: It refers to the relevant correct
response of information retrieved in recall i.e., the
percentage of relevant information retrieved with
high precision.
 Consistency: Consistency of data retrieval should
be maintained via indexing of the text by the
groups of indexers or by the authors.
 Ease of use: Information retrieval has become the
blood in the vein of the users in current scenario
using Internet on daily basis. Hence, it is more of
a responsibility now of information retrieval
systems to cater user requests with utmost
efficiency and effectiveness.
 Speed: It refers to the instant time taken in return
to searching the given information using fast
search techniques like NLP or run as "batch"
indexing.
2.1 IR MODELS
Information Retrieval (IR) is by far the most sought after
process used nowadays with the netizens. It is the process
which shows that a data collection is represented, stored,
then searched for the purpose of discovering knowledge in
response to user’s query. Information Retrieval (IR) came
into existence in 1950s out of sheer necessity of acquiring
information from users’ perspective. Since then, there has
been a constant development in this field in search of the
betterment of IR systems. Several IR systems like
google.com have become an everyday commodity and are
used by wide variety of users. Thus, Information retrieval
has become an integral part of our lives and an ever
growing research area too. The quality of information
retrieval can be measured using two main features:
 Precision: This is the percentage of documents
retrieved related to the user’s query.

 Recall: This is the percentage of documents that
are query related and are actually retrieved.
The fundamental IR models can be classified into Boolean,
vector & probabilistic models. Each model has been
introduced in its basic format as under:
1. The Boolean model: The Boolean model is the
most primitive information retrieval method and
one of the most criticized one too. This model
takes a query term as an unambiguous definition
of a set of documents. The Boolean model makes
use of Boolean algebra operators like AND, OR,
NOT in formulating a query. The major
disadvantage found in this Boolean system is that
the model is not able to give ranking to the
returned list of documents [4]. This model
associates a document and its set of keywords as
a part of query separated by AND, OR, NOT. The
retrieval function in the end determines if the
document is relevant or not.
2. Vector space model: In the Vector Space Model,
documents and query are represented as a Vector
and the angle between the two vectors are
computed using the similarity cosine function as:
Sim(dj,q) = dj.q / | dj|. |q|
The Vector Space Model has introduced a weight
scheme called if-idf weighting. The new scheme
has weights defined as:
 Term frequency (tf): This factor
determines the the number of times a
term has occurred in the document/
query.
 Inverse document frequency (idf): This
factor measures the inverse of the
number of documents that holds a given
query / document term [4].
3. Probabilistic Model: The probabilistic model
attempts to rank the documents by their
probability of relevance of a given query. Binary
vectors ~d and ~q are used to denote the
document queries. Each binary vector indicates if
a given term or document attribute occurs in the
document/ query.
III. SOFT COMPUTING TECHNIQUES
Soft Computing is a field in CS/IT dealing with the fusion
of methodologies that are designed to model and provide
solutions to real world problems mathematically. It aims to
exploit the uncertainty, imprecision and approximate
reasoning so as to achieve low-cost solutions that are
robust and tractable [1]. Its solutions are majorly
unpredictable, uncertain and between 0 and 1.
Soft computing is but different from hard computing; it
can tolerate imprecision, approximation uncertainty,
partial truth. The human mind holds the pivotal position in
soft computing. In fact soft computing deals in
implementation of optimization techniques to find
probable solutions of hard core problems.
Various techniques used under soft computing are genetic
algorithms (GA), fuzzy logic (FL), neural networks
(ANN), Bayesian networks, machine learning, etc. some of
which are described as under:
 Neural networks: Artificial neural network (ANN) is
a very demanding field of computer science research
in today’s era. ANN is an information system whose
working and thinking is based on the working of the
brain. It is similar to brain in two possible ways:
a) ANN acquires knowledge via a set learning
process.
b) Knowledge is stored in interconnected links called
synaptic weights.
ANN is mostly used in areas related to weather
prediction, pattern recognition, data recognition,
stock market prediction ,image processing, image
compression, to name a few. ANN works best in
areas which do not have set algorithms. ANN
architecture comprises of three layers: input,
hidden and output layer; each layer having many
nodes. Back propagation algorithm is the most
common method used in ANN networks [5].
 Fuzzy Logic: Fuzzy logic is a rule-based system that
relies on the practical experience of an operator.
Fuzzy logic was first devised by Dr. Zadeh of
University of California at Berkeley during 1960s.
Fuzzy logic deals with the concepts which can’t be
exactly explained in terms of binary 0 or 1. It defines
the way of reasoning the Boolean values by giving
results between 0 and 1. It is quite similar to the way
our brain functions. Fuzzy systems are a part of
developing human like capabilities. It reflects to
those areas where our solution is not absolute but
something fuzzy. Fuzzy logic is a form of artificial
intelligence software; therefore, it would be
considered to be a subset of A.
 Genetic algorithms(GA) : Genetic algorithm is a
subset of AI and fuzzy logic. It is majorly used to
deduce different optimization problems related to
real-life applications. The basic idea of any GA is to
copy the normal selection of any user the way he
would to find a suitable solution for given
application. Genetic algorithm is basically machine
learning model motivated by the biological evolution
model. It solves both the types of problems –

constrained and unconstrained and repeatedly
modifies series of individual solutions [6].
The utility of genetic algorithms can be seen in varied
fields like climatology, control engineering, automated
manufacturing & design, biomedical engineering, games
theory, code-breaking, and electronic design.
IV. APPLYING SOFT COMPUTING TO
INFORMATION RETRIEVAL
A lot of effort has been made to improvise the
performance of Information retrieval systems in recent
years. Till now the researchers are trying hard to explore
the information retrieval systems further by applying new
methodologies. Applying soft computing techniques is
giving positive results in increasing the efficiency of IR
systems. Using the artificial neural networks and fuzzy
logic concepts with information retrieval generate
altogether a new field called soft information retrieval [7].
Fuzzy set theory is hence applied to increase the flexibility
if IR systems. The main vagueness of IR system can be
tapped well with the use of fuzzy logic. The main concern
of applying fuzzy set theory to IR is:
 How to define the Boolean model representing
the documents and the query language.
 How to define the associative mechanism like
fuzzy clustering or fuzzy thesaurus.
To solve above mentioned problems, Boolean models have
been extended to represent a document as a fuzzy set of
terms. Each term has a numeric weight specified which
describes the association of keywords to the document‘s
content.
Various other fuzzy methods devised to improve the IR
systems can be applying numeric query weights, linguistic
query weights, and aggregation operators, fuzzy thesauri
of terms or fuzzy clustering of documents.
Another successful method of implementing soft
information retrieval is to apply ANN to the IR systems.
The efficiency of information retrieval can be improved by
applying the Supervised and Unsupervised learning
method of neural networks.
Supervised learning method of ANN incorporates an
“external teacher" [7]. This teacher then specifies the
required output of the Neural Networks. During the initial
learning phase, the ANN takes the values of the weights to
obtain the required output.
An unsupervised learning procedure is not bothered about
any learning or teaching feedback. It aims at applying self-
learning. This method is also called “self-organization"
because the system relies only upon local information and
internal control by using the input patterns. Hence
unsupervised learning has become more popular in IR
specifically for documents or terms classification &
clustering.
V. CONCLUSION
Soft computing techniques have become the talk of the
town in recent years and a promising area for the
researchers to explore. The implementation of soft
computing on IR systems to improvise its efficiency has
become more important as the power of computer
processing devices has increased while the cost has
reduced.
The soft computing techniques have become an integral
part of our daily life in the form of applying fuzzy logic,
expert systems and artificial neural networks to the
appliances like cookers, washing machines and
refrigerators. Many commercial and industrial applications
of soft computing are widely in use and is expected to
grow exponentially over the years to come..
The application of soft computing techniques on
information retrieval should increase drastically when
combined with IoT devices in years to come.
REFERENCES
[1] http://shodhganga.inflibnet.ac.in/bitstream/10603/101
61/11/11_chapter%203.pdf
[2] http://www.doc.ic.ac.uk/~nd/surprise_97/journal/vol2/
hks/infor_ret.html
[3] A. Roshdi, A. Roohparvar, “Review: Information
Retrieval Techniques and Applications”, International
Journal of Computer Networks and Communications
Security, VOL. 3, NO. 9, SEPTEMBER 2015
[4] D.Hiemstra,P. de Vries, “Relating the new language
models of information retrieval to the traditional
retrieval models”, published as CTIT technical report
TR-CTIT-00-09, May 2000.
[5] https://www.doc.ic.ac.uk/~nd/surprise_96/journal/vol
4/cs11/report.html#Introduction%20to%20neural%20
networks
[6] Dogan Ibrahim, “An overview of soft computing”,
12th International Conference on Application of
Fuzzy Systems and Soft Computing, ICAFS 2016, 29-
30 August 2016, Vienna, Austria.
[7] Fabio Crestani, Gabriella Pasi, “Soft Information
Retrieval: Applications of Fuzzy Set Theory and
Neural Networks”.

Applying Soft Computing Techniques in Information Retrieval

More Related Content

What's hot

Similar to Applying Soft Computing Techniques in Information Retrieval

Recently uploaded

Applying Soft Computing Techniques in Information Retrieval