Improving Query Expansion by Automatic Query Disambiguation in

Intelligent Information Retrieval

Oussama Ben Khiroun

, Bilel Elayeb

1,2

, Ibrahim Bounhas

Fabrice Evrard

and Narjès Bellamine Ben Saoud

1,5

RIADI Research Laboratory, ENSI Manouba University, 2010, Manouba, Tunisia

Emirates College of Technology, P.O. Box: 41009, Abu Dhabi, U.A.E

LISI Lab. of computer science for industrial systems, ISD Manouba University, 2010, Manouba, Tunisia

Informatics Research Institute of Toulouse (IRIT), 02 Rue Camichel, 31071 Toulouse, France

Higher Institute of Informatics (ISI), Tunis El Manar University, 2080 Ariana, Tunisia

Keywords: Semantic Query Expansion, Word Sense Disambiguation, Information Retrieval, Possibility Theory,

Semantic Graph.

Abstract: We study in this paper the impact of Word Sense Disambiguation (WSD) on Query Expansion (QE) for

monolingual intelligent information retrieval. The proposed approaches for WSD and QE are based on

corpus analysis using co-occurrence graphs modelled by possibilistic networks. Indeed, our model for

relevance judgment uses possibility theory to take advantages of a double measure (possibility and

necessity). Our experiments are performed using the standard ROMANSEVAL test collection for the WSD

task and the CLEF-2003 benchmark for the QE process in French monolingual Information Retrieval (IR)

evaluation. The results show the positive impact of WSD on QE based on the recall/precision standard

metrics.

1 INTRODUCTION

Users of information retrieval (IR) systems choose

generally short queries to express their needs. The

submitted request is matched with the indexes of

documents, according to a specified matching model

and search results are returned sorted by descendant

order of the computed relevance scores. These

results may contain noise (irrelevant documents) due

to the shortness of query.

One possible solution which can enhance results

consists in expanding the context of the query, thus

satisfying the user. Query expansion (QE) consists

in enriching the user's query by adding new terms to

better express his need (Elayeb et al., 2011;

Carpineto and Romano, 2012). Another solution

arises when one or more terms in query have more

than one sense (ambiguous). If we expand the query

using wrong sense information, search results would

be probably irrelevant to the user (Krovetz, 1997;

Paskalis and Khodra, 2011).

Then it is necessary to identify, in a second step,

the exact sense of ambiguous words, what is called

word sense disambiguation (WSD). It is defined as

the ability to identify the meaning of words in

context using one or more sources of knowledge to

associate the most appropriate senses with

ambiguous terms (Navigli, 2009).

WSD is an important field of natural language

processing (NLP). However, WSD is also used in

information retrieval and proved its impact to

improve the search process (Liu et al., 2005; Zhong

and Ng, 2012).

Many studies about query expansion and WSD

were conducted (Chifu and Ionescu, 2012). We

present in this work a comparative study of the

contribution of WSD to IR and its impact on query

expansion based on possibilistic networks. The

presented results focus on queries issued from the

CLEF-2003 corpus and containing ambiguous words

from the ROMANSEVAL benchmark for WSD in

French language.

In this paper, we propose, assess and compare a

new possibilistic query expansion approach using

word sense disambiguation on a graph of co-

occurrence. As a background of our work, we

153

Ben Khiroun O., Elayeb B., Bounhas I., Evrard F. and Bellamine Ben Saoud N..

Improving Query Expansion by Automatic Query Disambiguation in Intelligent Information Retrieval.

DOI: 10.5220/0004822401530160

In Proceedings of the 6th International Conference on Agents and Artiﬁcial Intelligence (ICAART-2014), pages 153-160

ISBN: 978-989-758-015-4

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

present in Section 2 some related works. The

proposed possibilistic approach is detailed in Section

3 and a set of experimentations, results and

interpretations is made in Section 4.

2 RELATED WORK

We present in this section some useful IR concepts

and related works from the literature.

Query Expansion (QE) is one of the strategies

implemented in IR systems to improve their

performance and better satisfy users. (Carpineto and

Romano, 2012) classify QE into two main

techniques: interactive query expansion (IQE),

which relies on user guidance, and automatic query

expansion (AQE). In both cases, QE can be achieved

by various techniques such as exploitation of

external linguistic resources (thesauri, dictionaries,

etc.), corpus analysis and relevance feedback

techniques (Manning et al., 2008).

QE approaches based on relevance feedback can

be classified into three main categories (Manning et

al., 2008). The first approach is “user relevance

feedback” which includes user judgment of the

returned results. The second one is “indirect

relevance feedback” (called often implicit relevance

feedback) using indirect sources of evidence such as

number of hits on web page’s links. The last

approach is “pseudo relevance feedback” (also

known as blind relevance feedback). In this method,

the IRS uses the top k retrieved documents which are

the most relevant to expand the initial query. Thus, a

set of candidate terms from these documents is

added using often variants of Rocchio algorithm

(Rocchio, 1971).

Although relevance feedback may reduce noise

in IR results, all these techniques do not provide

direct way to exactly identify the meaning of the

query terms, thus needing other approaches for

query disambiguation.

Word sense disambiguation (WSD) is a

commonly known task in natural language

processing (NLP) problems and IR (Banerjee and

Pedersen, 2002). According to (Navigli, 2009),

WSD heavily relies on knowledge sources which are

classified into two groups: structured resources

(such as thesauri, electronic dictionaries, etc.) and

unstructured resources (such as corpora documents).

Pinto and Pérez-sanjulián (2008) studied the

impact of applying WSD on automatic QE using

WordNet as external linguistic resource for both

WSD and QE. Experiments were conducted using

short and long queries from the TREC-8 text

collection. Results proved that QE applied on both

short and long queries is not able to improve

retrieval performance without identifying the correct

meaning of ambiguous word from the set of

extracted synonyms from WordNet (Miller et al.,

1990). The search performance was better for short

queries.

Paskalis and Khodra (2011) analyzed many

scenarios on IR process by using QE, WSD,

stemming and a relevance feedback technique. WSD

was applied using an extended implementation of

Lesk algorithm (Banerjee and Pedersen, 2002;

Banerjee and Pedersen, 2003). For the QE task, they

used two components: a co-occurrence based

thesaurus built automatically from the documents

collection and pseudo relevance feedback by

assuming a set of top documents as relevant and

injecting representative terms in the original query

(Manning et al., 2008).

Elayeb et al. (2011) and Ben Khiroun et al.

(2012) proposed respectively QE and WSD

approaches based on possibilistic networks.

However, they did not apply their WSD algorithm

on query disambiguation. They also used

dictionaries as lexical resources. To the best of our

knowledge, no research about using co-occurrence

graphs for both WSD and QE tasks were conducted

in the French language. Besides, the possibilistic

approach has not been tested for query

disambiguation. Thus, we need to experiment

possibilistic networks for enhancing IR results, by

studying many combinations of scenarios of WSD,

QE and relevance feedback.

Based on this survey, we propose a study of a

combined approach for WSD and QE tasks, using on

possibilistic networks and applied on an extracted

co-occurrence graph.

3 A POSSIBILISTIC APPROACH

FOR COMBINED WSD & SQE

Our approach combines automatic QE, WSD and

pseudo relevance feedback. For the first two tasks,

we need to compute the similarity between queries’

terms (in the case of expansion) or between terms

and senses (in the case of disambiguation). In this

paper, we opted for co-occurrence graphs extracted

from corpora to model contextual and similarity

links. Nevertheless, our implementation of similarity

calculus is generic enough to be used with other

types of graphs (e.g. dictionary graphs in (Elayeb et

al., 2011)).

ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence

154

3.1 Graph-based Knowledge

Representation

Our approach is based on possibilistic networks for

WSD and QE. In fact, we consider, for building the

co-occurrence graph, that two nodes are related if

they exist in the same sentence. The edges are bi-

oriented and weighted by the normalized co-

occurrence frequency of the related terms. On the

other hand, ambiguous words are related with their

appropriate senses in the dictionary.

We consider the different components as follow:

 T: the set of terms in the corpus.

 S: the set of senses in the dictionary.

 A node t

is related to a node t

if t

and t

co-

occur in the same sentence; where {t

T}.

 A node t

is related to a node s

if t

is an

ambiguous term and s

represents a sense of

; where {t

T} and {s

S}.

3.2 Graph-based Possibilistic

Similarity

To compute terms similarity in both QE and WSD

tasks, we based our approach on the possibilistic

theory introduced by (Zadeh, 1978) and developed

by several authors (Dubois and Prade, 2011; Dubois

and Prade, 2012). We adapted the possibilistic

model architecture of (Elayeb et al., 2011) to be

applied on co-occurrence graphs. We define the

Degree of Possibilistic Relevance (DPR) for each

co-occurrence graph’ node n

given a query Q = (t

, …, t

) by:

)Q|n()Q|n()n(DPR





(1)

Where (n

|Q) and N(n

|Q) represent

respectively the possibility and necessity measures

(Elayeb et al., 2009). The former allows to reject the

non-relevant nodes (those who are not close to the

context of the query and may not be used to expand

or disambiguate it). The latter is used to reinforce

the relevance of the most important nodes. The two

measures are computed as follows:

Tjj1

jTj1j

nft...nft

)n|(t...)n|(tQ)|n(









(2)

)]1(...)1[(1Q)|(

1 Tjjj

nnnN





(3)

Where nft

represents the normalized frequency of

query terms in the co-occurrence graph:

)tf(max

nft

kjk



(4)

In this formula tf

is the weight of the edge relating

the nodes t

and n

(i.e. the number of times the two

nodes co-occur).

And:

10ij

nft)

nCN

(Logn 

(5)

Where:

 nCN = total number of nodes in the co-

occurrence graph related to the query terms;

 nN

= number of nodes related to the term t

Using the Log function (such as in TF-IDF)

allows to compute the discriminative power of the

query terms. Thus, we select the graph nodes which

are closest to the most discriminative items of the

contextual information represented in the query.

3.3 Query Treatment Process

The process in Figure 1 presents the different

resources used in the WSD task, QE and pseudo

relevance feedback.

Starting from an initial query, the QE module is

executed to generate an expanded query. In the case

of ambiguous terms, the WSD module is used before

applying QE. Thus, the best sense node having the

greater possibilistic score is selected and the terms

existing in its definition are used for expanding the

original query.

Figure 1: Query expansion using WSD process.

For both QE and WSD processes, the co-

occurrence graph is used to achieve possibilistic

ImprovingQueryExpansionbyAutomaticQueryDisambiguationinIntelligentInformationRetrieval

155

calculus. The expanded query is matched with

documents to achieve results.

A pseudo relevance feedback is applied at the

end of the process by extracting the most significant

terms from the top first returned documents.The

whole process may be iterated.

In order to perform pseudo relevance feedback

based on the document collection, we used the Bo1

(Bose-Einstein 1) pseudo relevance feedback

method implemented in the Terrier information

retrieval platform (Ounis et al., 2005). The default

settings are specified as follows: the number of

terms to expand a query is set to 10 and the number

of top-ranked documents from which these terms are

extracted is limited to 3 documents.

3.4 Illustrative Example

We consider in this example an excerpt of an

ambiguous query:

Les règles d'orthographe et de

ponctuation pour la langue allemande

ont été considérablement simplifiées

Which may be translated as follows:

The rules of spelling and

punctuation for the German language

has been considerably simplified

The query is tokenized and lemmatized ignoring

stop words (like pronouns, articles, etc.) as follow:

règle (rule), orthographe

(spelling), ponctuation

(punctuation), langue (language),

allemand (German), considérable

(considerable), simple (simple)

The output query contains the ambiguous word

“simple” (simple). So the WSD is executed and the

sense having the best possibilistic score from

ROMANSEVAL dictionary is selected (in this

example we consider the sense AII1):

[…]

AI2 Qui n'est formé que […]

AI3 Qui suffit à soi seul […]

AII1 Qui est facile à comprendre […]

Translated as:

[…]

AI2 Which is formed only by[…]

AI3 Sufficient to itself alone […]

AII1 That is easy to understand […]

The corresponding terms in the definition are

injected in query:

règle (rule), orthographe

(spelling), ponctuation

(punctuation), langue (language),

allemand (German), considérable

(considerable), simple (simple),

facile (easy), comprendre

(understand)

Afterwards, the disambiguated query is

processed to be expanded by the QE module.

4 EXPERIMENTAL RESULTS

In order to study the impact of WSD on QE in

French language, we used two test collections to

experiment our approach: CLEF-2003 and

ROMANSEVAL.

In all our experiments, we focused only on

queries from CLEF-2003 test collection which

contains ambiguous terms included in

ROMANSEVAL test collection. The two test

collections are presented in the following sub-

sections.

4.1 CLEF-2003 Test Collection

We used series of standard tests from the Cross-

Language Evaluation Forum (CLEF). It provides

necessary tools for the evaluation of information

retrieval systems on large corpora including a set of

documents, a set of queries and the list of relevant

documents for each query.

Each query is represented in the XML format by

a title containing its terms, a description and a

detailed narrative text. The CLEF-2003 collection

for French language is composed of Le Monde94,

ATS94, and ATS95 sub-collections forming 57 test

queries and more than 300 MB of data (Braschler

and Peters, 2004).

4.2 ROMANSEVAL Test Collection

For the WSD task, we used the ROMANSEVAL

standard test collection which provides the necessary

resources for WSD including a set of documents and

a list of test sentences containing ambiguous words.

A set of 60 ambiguous words distributed on three

grammatical categories (20 nouns, 20 adjectives, 20

verbs) were annotated by 6 members in accordance

with the senses. Each word occurrence may have

one or several labels of sense or none (Segond,

2000).

ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence

156

4.3 Experimental Setup

The query sub-set used for experiments is composed

of 15 queries containing ambiguous words from the

ROMANSEVAL test collection.

In the first step, we studied in sub-section 4.4 the

impact of QE, as a separated process, on the IR

performance. Then, WSD is experimented apart in

sub-section 4.5 to evaluate the disambiguating

process. The impact of WSD on QE is experimented

in sub-section 4.6.

We used the Terrier experimental platform for IR

to evaluate our system (Ounis et al., 2006). Two

common IR measures where used: (6) The precision

measured by the ratio of relevant documents

retrieved to the number of documents retrieved and

(7) The recall presenting the ratio of relevant

documents retrieved to the number of relevant

documents in the collection.

Precision 

#relevantretrieveddocuments

#retrieveddocuments

(6)

Recall 

#relevantretrieveddocuments



relevantdocumentsinthecollection



(7)

In this paper our experiments are limited to the

using of the Okapi (BM25) matching model already

available in Terrier platform. But, we plan in the

future to experiment our approach via the

possibilistic matching model proposed by (Elayeb et

al., 2009) in order to compare results to those

obtained via Okapi. In fact, our goal is to approve

that our approach is generic as it is independent of

the used matching model.

4.4 Evaluating QE Approach

We compare in Table 1 different QE scenarios based

on co-occurrence possibilistic graph (CooQE) built

from the ROMANSEVAL Test Collection.

Ogilvie et al. (2009) studied the number of

expansion terms to use in automatic QE through

eight IR systems. The results show that the number

of expansion terms that optimizes mean average

precision varies widely across systems and topic

sets. For many topics, ten or fewer expansion terms

provided the best average precision according to the

experiments of (Ogilvie et al., 2009).This

assumption is studied for the French language as

follows.

The number of expansion terms in Table 1 was

varied from N div 4 terms to N terms where N

represents the number of terms in the original query.

These numbers for expansion terms are chosen

by considering that the narrative part of test queries

is long (more than 10 terms). Applying QE on such

long queries as detailed by (Pinto and Pérez-

sanjulián, 2008) may produce noisy and non-

interpretable results. Thus, we fixed the quarter of

query terms as minimum scenario to have significant

expansion results.

The last two columns of table 1 present the MAP

measure, which is the mean of the average precision

scores for each query and the exact precision (R-

Precision), which is the precision at rank R; where R

is the total number of relevant documents (Manning

et al., 2008). Baseline results, applied on reference

initial queries without QE, are also presented in

Table1.

Table 1: Query expansion results.

Method

Number of

terms for QE

MAP R-precision

baseline - 0.5487 0.5174

CooQE

N 0.4180 0.4043

N div 2 0.4700 0.4633

N div 4 0.5083 0.4742

The experimented results show a decrease in IR

performance when applying the QE process

proportionally to the number of expansion terms in

both MAP and R-precision measures.

According to the Recall-Precision curve

presented in Figure 2, the results for the three QE

scenarios are not satisfying in comparison with the

default baseline results.

However, we can affirm that QE (mainly for N

div 4 scenario) is better than the baseline at high

recall levels (initially better at retrieving the relevant

documents).

Figure 2: Recall-Precision curve for QE.

These results are affected by the ambiguity of the

queries and the difficulty of distinguishing the right

sense for the ambiguous terms. In fact, the longer the

query is the worst IR performance results are.

ImprovingQueryExpansionbyAutomaticQueryDisambiguationinIntelligentInformationRetrieval

157

4.5 Evaluating the Possibilistic WSD

Approach

In this section, we experiment the efficiency of

WSD using the possibilistic approach described in

Section 3. We consider only the sense having the

best DPR score according to the possibilistic co-

occurrence graph-based calculus.

Afterwards, we performed expert-based

evaluation for the relevance of the selected sense

according to the original query and tagged it by three

degrees of relevance: 1 (relevant), 0 (partially

relevant) or -1 (not relevant).

After applying WSD on the 15 sub-test

ambiguous queries, we identified 5 relevant senses

and 4 senses as not relevant (cf. Table2).

Table 2: Evaluating WSD approach.

#Relevant

senses

#Partially

relevant senses

#Not relevant

senses

5 6 4

This evaluation was conducted manually for the

lack of ambiguous contexts’ tagging of

ROMANSEVAL words according to CLEF-2003

collection’s queries.

4.6 Combining WSD and QE

Approaches

The final set of experimentations consists in

applying WSD on queries before expanding terms.

This task may help in selecting the best sense for

ambiguous words before applying a QE process

aiming to reduce noise.

Therefore, the terms composing the selected

sense are injected in the query and a QE process is

then applied (WSD_QE test).

We also applied the pseudo relevance feedback

technique in our experiments at the end of

disambiguation and expansion chain (WSD_QE_RF

test).

For all the expanded queries in Figure 3 (adding

N terms), Figure 4 (adding N div 2 terms) and Figure 5

(

adding N div 4 terms), the WSD applied alone after

possibilistic QE has a minor enhancement in

comparison with the results of QE without WSD.

Nevertheless, the two experiments results (i.e.

WSD_QE and QE) are above the reference baseline.

However, when combining pseudo relevance

feedback with QE and WSD, we observe better IR

performance especially for a limited number of

expansion terms (cf. Figure 4 and Figure 5).

Figure 3: Recall-Precision curve by adding N terms for

each ambiguous query with and without WSD.

Figure 4: Recall-Precision curve by adding N div 2 terms

for each ambiguous query with and without WSD.

Figure 5: Recall-Precision curve by adding N div 4 terms

for each ambiguous query with and without WSD.

According to the three scenarios, we can confirm

the positive performance impact of WSD on QE

mainly for the initial recall levels (<10%).

Combining relevance feedback with WSD and QE

contributed also in the enhancement of IR

performance.

The same positive impact of relevance feedback

ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence

158

was observed by (Paskalis and Khodra, 2011). We

also join the works interpretations of (Pinto and

Pérez-sanjulián, 2008) who studied the IR

performance according to short and long queries

which may generate noise while applying QE.

5 CONCLUSIONS

In this work, we present a possibilistic approach to

study the impact of Word Sense Disambiguation

(WSD) on Query Expansion (QE). The approach

was applied for the French language to verify many

query treatment scenarios, but it is also applicable to

other languages. As a first step, we prepared a co-

occurrence graph from the documents’ collection.

Then, this resource was used in the selection of

candidate sense/terms for both WSD and QE. Final

results confirmed that WSD is necessary in the IR

process overcome the ambiguity problem.

Furthermore, Pseudo Relevance Feedback plays

an important role in the combined WSD and QE

approach proposed in this paper. However, the

retrieval performance is decreased when using many

expansion terms. This fact is interpreted by the noise

effect issued from the co-occurrence graph resource.

As future perspectives of the current work, we

propose to compare the use of document knowledge

extraction (as presented in the current work by co-

occurrence graph presentation) to other external

resources such as dictionaries. We aim to study also

the effectiveness of possibilistic networks in query

disambiguation compared to other probabilistic

approaches such as the circuit-based calculus

(Elayeb et al., 2011). Finally, the graph-based query

treatment algorithms were implemented in a generic

manner which may be applied with other languages

such as English, Spanish and Arabic.

ACKNOWLEDGEMENTS

We are grateful to the Evaluations and Language

resources Distribution Agency (ELDA) which

kindly provided us the Le Monde94 and ATS94

document collections of the CLEF-2003 campaign.

REFERENCES

Banerjee, S., and Pedersen, T., 2002. An Adapted Lesk

Algorithm for Word Sense Disambiguation Using

WordNet. In Proceedings of the Third International

Conference on Computational Linguistics and

Intelligent Text Processing. CICLing’02. London, UK,

UK: Springer-Verlag, pp. 136–145.

Banerjee, S., and Pedersen, T., 2003. Extended Gloss

Overlaps as a Measure of Semantic Relatedness. In

Proceedings of the 8

International Joint Conference

on Artificial Intelligence. Acapulco, Mexico: Morgan

Kaufmann, pp. 805–810.

Braschler, M., and Peters, C., 2004. CLEF 2003

Methodology and Metrics. In C. Peters et al., eds.

Comparative Evaluation of Multilingual Information

Access Systems. LNCS 3237, Springer Berlin

Heidelberg, pp. 7–20.

Carpineto, C. and Romano, G., 2012. A Survey of

Automatic Query Expansion in Information Retrieval.

ACM Computing Surveys (CSUR), 44(1), pp.1:1–1:50.

Chifu, A.-G. and Ionescu, R.-T., 2012. Word sense

disambiguation to improve precision for ambiguous

queries. Central European Journal of Computer

Science, 2(4), pp.398–411.

Dubois, D. and Prade, H., 2012. Possibility Theory. In R.

A. M. Ph. D, ed. Computational Complexity. Springer

New York, pp. 2240–2252.

Dubois, D. and Prade, H., 2011. Possibility theory and its

application: Where do we stand. Mathware and Soft

Computing, 18(1), pp.18–31.

Elayeb, B., Evrard, F., Zaghdoud, M., and Ben Ahmed,

M., 2009. Towards An Intelligent Possibilistic Web

Information Retrieval using Multiagent System:

Interactive Technology and Smart Education, Special

issue: New learning support systems, 6(1), pp.40–59.

Elayeb, B., Bounhas, I., Ben Khiroun, O., Evrard, F., and

Bellamine Ben Saoud, N., 2011. Towards a

Possibilistic Information Retrieval System Using

Semantic Query Expansion: International Journal of

Intelligent Information Technologies, 7(4), pp.1–25.

Ben Khiroun, O., Elayeb, B., Bounhas, I., Evrard, F., and

Bellamine Ben Saoud, N., 2012. A Possibilistic

Approach for Automatic Word Sense Disambiguation.

In Proceedings of the 24

Conference on

Computational Linguistics and Speech Processing

(ROCLING). Taiwan, pp. 261–275.

Krovetz, R., 1997. Homonymy and polysemy in

information retrieval. In Proceedings of the eighth

conference on European chapter of the Association for

Computational Linguistics. Stroudsburg, PA, USA:

Association for Computational Linguistics, pp. 72–79.

Liu, S., Yu, C., and Meng, W., 2005. Word sense

disambiguation in queries. In Proceedings of the 14

ACM international conference on Information and

knowledge management. New York, NY, USA: ACM,

pp. 525–532.

Manning, C. D., Raghavan, P., and Schütze, H., 2008.

Introduction to Information Retrieval, New York, NY,

USA: Cambridge University Press.

Miller, G. A. et al., 1990. Introduction to WordNet: An

On-line Lexical Database. International Journal of

Lexicography, 3(4), pp.235–244.

Navigli, R., 2009. Word sense disambiguation: A survey.

ACM Computing Surveys (CSUR), 41(2), pp.10:1–

ImprovingQueryExpansionbyAutomaticQueryDisambiguationinIntelligentInformationRetrieval

159

10:69.

Ogilvie, P., Voorhees, E., and Callan, J., 2009. On the

number of terms used in automatic query expansion.

Information Retrieval, 12(6), pp.666–679.

Ounis, I., Gianni, A., Vassilis, P., Ben, H., Craig, M., and

Douglas, J., 2005. Terrier Information Retrieval

Platform. In Proceedings of the 27

European

Conference on IR Research, LNCS 3408, Springer

Berlin Heidelberg, pp. 517-519.

Ounis, I., Gianni, A., Vassilis, P., Ben, H., Craig, M., and

Christina, L., 2006. Terrier: A High Performance and

Scalable Information Retrieval Platform. In

Proceedings of ACM SIGIR’06 Workshop on Open

Source Information Retrieval (OSIR). Seattle,

Washington, USA, pp. 18–25.

Paskalis, F. B. D., and Khodra, M. L., 2011. Word sense

disambiguation in information retrieval using query

expansion. In Proceedings of the IEEE-2011

International Conference on Electrical Engineering

and Informatics. Bandung, Indonesia, pp. 1-6.

Pinto, F. J., and Pérez-sanjulián, C. F., 2008. Automatic

query expansion and word sense disambiguation with

long and short queries using WordNet under vector

model. Actas de los Talleres de las Jornadas de

Ingeniería del Software y Bases de Datos, 2(2), pp.17–

23.

Rocchio, J., 1971. Relevance Feedback in Information

Retrieval. In The SMART Retrieval System. Prentice-

Hall, Englewood Cliffs, New Jersey, USA, pp. 313–

323.

Sanderson, M., 1994. Word sense disambiguation and

information retrieval. In Proceedings of the 17

annual international ACM SIGIR conference on

Research and development in information retrieval.

New York, NY, USA: Springer-Verlag New York,

Inc., pp. 142–151.

Segond, F., 2000. Framework and results for French.

Computers and the humanities, 34(1-2), pp.49–60.

Stokoe, C., Oakes, M.P., and Tait, J., 2003. Word sense

disambiguation in information retrieval revisited. In

Proceedings of the 26

annual international ACM

SIGIR conference on Research and development in

informaion retrieval. New York, NY, USA: ACM, pp.

159–166.

Zadeh, L., 1978. Fuzzy sets as a basis for a theory of

possibility. Fuzzy Sets and Systems, 1(1), pp.3–28.

Zhong, Z., and Ng, H. T., 2012. Word sense

disambiguation improves information retrieval. In

Proceedings of the 50

Annual Meeting of the

Association for Computational Linguistics: Long

Papers - Volume 1. Stroudsburg, PA, USA:

Association for Computational Linguistics (ACL), pp.

273–282.

ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence

160