RANLP vol. IV

• Volume information (plain text)
• Ordering information
• BibTeX entry for the entire volume [BibTeX entries for all chapters]
• Table of contents [PDF]
• List of authors and titles, abstracts and individual BibTeX entries

Volume Information

↑

Ordering Information

↑

BibTeX Entry

@BOOK{Nicolov-Bontcheva-Angelova-Mitkov'2007
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,TITLE     = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,PUBLISHER = {John Benjamins}
  ,YEAR      = {2007}
  ,VOLUME    = {292}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,ISBN      = {978 90 272 4807 7}
}

Table of Contents

I.  COMPUTATION FOR LINGUISTICS

II.  INFORMATION EXTRACTION & INDEXING

Florian Seydoux & Jean-Cedric Chappelier
Semantic indexing using minimum redundancy cut in ontologies   25

→

Niraj Aswani, Valentin Tablan, Kalina Bontcheva & Hamish Cunningham
Indexing and querying linguistic metadata and document content   35

→

Irina Matveeva, Gina-Anne Levow, Ayman Farahat & Christiaan Royer
Term representation with Generalized Latent Semantic Analysis   45

→

III.  PARSING

Sandra Kübler
How does treebank annotation influence parsing? Or how not to compare apples and oranges   79

→

Laura Alonso, Joan Antoni Capilla, Irene Castellon, Ana Fernandez-Montraveta & Gloria Vazquez
The SenSem project: Syntactico-semantic annotation of sentences in Spanish   89

→

IV.  ANAPHORA & REFERRING EXPRESSIONS

Erhard W. Hinrichs, Katja Filippova & Holger Wunsch
A data-driven approach to pronominal anaphora resolution for German   115

→

V.  CLASSIFICATION

Filip Ginter, Sampo Pyysalo & Tapio Salakoski
Document classification using semantic networks with an adaptive similarity measure   137

→

Rada Mihalcea & Samer Hassan
Text summarization for improved text classification   147

→

Caroline Sporleder & Alex Lascarides
Exploiting linguistic cues to classify rhetorical relations   157

→

VI.  TEXTUAL ENTAILMENT & QUESTION ANSWERING

Jorg Tiedemann
A genetic algorithm for optimising information retrieval with linguistic features in question answering   177

→

Vasile Rus & Arthur Graesser
Lexico-syntactic subsumption for textual entailment   187

→

Courtney Corley, Andras Csomai & Rada Mihalcea
A knowledge-based approach to text-to-text similarity   197

→

VII.  ONTOLOGIES

Eduard Barbu & Verginica Barbu Mititelu
Automatic building of Wordnets   217

→

VIII.  MACHINE TRANSLATION

Victoria Arranz, Elisabet Comelles & David Farwell
Multi-perspective evaluation of the FAME speech-to-speech translation system for Catalan, English and Spanish   237

→

Daniel Varga, Peter Halacsy, Andras Kornai, Viktor Nagy, Laszlo Nemeth & Viktor Tron
Parallel corpora for medium density languages   247

→

IX.  CORPORA

Anne De Roeck, Avik Sarkar & Paul H. Garthwaite
Even very frequent function words do not distribute homogeneously   267

→

Lucia Specia, Maria das Gra¸cas V. Nunes & Mark Stevenson
Exploiting parallel texts to produce a multilingual sense tagged corpus for word sense disambiguation   277

→

Francis Chantree, Alistair Willis, Adam Kilgarriff & Anne de Roeck
Detecting dangerous coordination ambiguities using word distribution   287

→

List and Addresses of Contributors   297

Index of Subjects and Terms   303

↑

List of Authors and Titles, Abstracts and Individual BibTeX Entries

I.  COMPUTATION FOR LINGUISTICS

Even now techniques are in common use in computational linguistics which could lead to important advances in pure linguistics, especially language acquisition and the study of language variation, if they were applied with intelligence and persistence. Reliable techniques for assaying similarities and differences among linguistic varieties are useful not only in dialectology and sociolinguistics, but could also be valuable in studies of first and second language learning and in the study of language contact. These techniques would be even more valuable if they indicated relative degrees of similarity, but also the source of deviation (contamination). Given the current tendency in linguistics to wish to confront the data of language use more directly, techniques are needed which can handle large amounts of noisy data and extract reliable measures from them. The current focus in Computational Linguistics on useful applications is a very good thing, but some further attention to the linguistic use of computational techniques would be very rewarding.

@INCOLLECTION{Nerbonne'2007
  ,AUTHOR    = {John Nerbonne}
  ,TITLE     = {{L}inguistic {C}hallenges for {C}omputationalists}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {1-16}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

II.  INFORMATION EXTRACTION & INDEXING

This chapter examines some current issues in natural language processing from the vantage point of information extraction (IE), and so gives some perspective on what is needed to make ie more successful. By ie we mean the identification of important types of relations and events in unstructured text. IE provides a nice reference point because it can benefit from a wide range of technologies, including techniques for analyzing sentence structure, for identifying syntactic and semantic paraphrases, and for discovering the information structure of domains.

@INCOLLECTION{Grishman'2007
  ,AUTHOR    = {Ralph Grishman}
  ,TITLE     = {{NLP}: {A}n {I}nformation {E}xtraction {P}erspective}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {17-23}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Seydoux, Florian & Jean-Cedric Chappelier. 2007. "Semantic Indexing Using Minimum Redundancy Cut in Ontologies". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 25-34. Amsterdam & Philadelphia: John Benjamins.

This paper presents a new method that improves semantic indexing while reducing the number of indexing terms. Indexing terms are determined using a minimum redundancy cut in a hierarchy of conceptual hypernyms provided by an ontology (e.g., WordNet, EDR). The results of some information retrieval experiments carried out on several standard document collections using the WordNet and EDR ontologies are presented, illustrating the benefit of the method.

@INCOLLECTION{Seydoux-Chappelier'2007
  ,AUTHOR    = {Florian Seydoux and Jean-C\'{e}dric Chappelier}
  ,TITLE     = {{S}emantic {I}ndexing {U}sing {M}inimum {R}edundancy {C}ut in {O}ntologies}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {25-34}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Aswani, Niraj, Valentin Tablan, Kalina Bontcheva & Hamish Cunningham. 2007. "Indexing and Querying Linguistic Metadata and Document Content". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 35-44. Amsterdam & Philadelphia: John Benjamins.

The need for efficient corpus indexing and querying arises frequently both in machine learning-based and human-engineered natural language processing systems. This paper presents the ANNIC system, which can index documents not only by content, but also by their linguististic annotations and features. It also enables users to formulate versatile queries mixing keywords and linguistic information. The result consists of the matching texts in the corpus, displayed within the context of linguistic annotations (not just text, as is customary for KWIC systems). The data is displayed in a graphical user interface, which facilitates its exploration and the discovery of new patterns, which can in turn be tested by launching new annic queries.

@INCOLLECTION{Aswani-Tablan-Bontcheva-Cunningham'2007
  ,AUTHOR    = {Niraj Aswani and Valentin Tablan and Kalina Bontcheva and Hamish Cunningham}
  ,TITLE     = {{I}ndexing and {Q}uerying {L}inguistic {M}etadata and {D}ocument {C}ontent}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {35-44}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Matveeva, Irina, Gina-Anne Levow, Ayman Farahat & Christiaan Royer. 2007. "Term Representation with Generalized Latent Semantic Analysis". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 45-54. Amsterdam & Philadelphia: John Benjamins.

Representation of term-document relations is very important for document indexing and retrieval. In this paper, we present Generalized Latent Semantic Analysis as a framework for computing semantically motivated term and document vectors. Our focus on term vectors is motivated by the recent success of co-occurrence based measures of semantic similarity obtained from very large corpora. Our experiments demonstrate that glsa term vectors efficiently capture semantic relations between terms and outperform related approaches on the synonymy test.

@INCOLLECTION{Matveeva-Levow-Farahat-Royer'2007
  ,AUTHOR    = {Irina Matveeva and Gina-Anne Levow and Ayman Farahat and Christiaan Royer}
  ,TITLE     = {{T}erm {R}epresentation with {G}eneralized {L}atent {S}emantic {A}nalysis}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {45-54}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

III.  PARSING

This paper develops a general framework for machine learning based dependency parsing based on a pipeline approach, where a task is decomposed into several sequential stages. To overcome the error accumulation problem of pipeline models, we propose two natural principles for pipeline frameworks: (i) make local decisions as reliable as possible, and (ii) reduce the number of sequential decisions made. We develop an algorithm that provably satisfies these principles and show that the proposed principles support several algorithmic choices that improve the dependency parsing accuracy significantly. We present state of the art experimental results for English and several other languages.

@INCOLLECTION{Chang-Do-Roth'2007
  ,AUTHOR    = {Ming-Wei Chang and Quang Do and Dan Roth}
  ,TITLE     = {{M}ultilingual {D}ependency {P}arsing: {A} {P}ipeline {A}pproach}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {55-78}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Kübler, Sandra. 2007. "How does Treebank Annotation Influence Parsing? Or How Not to Compare Apples and Oranges". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 79-88. Amsterdam & Philadelphia: John Benjamins.

In this paper, we will investigate the influence which different decisions in the annotation schemes of treebanks have on parsing. The investigation uses the comparison of similar treebanks of German, Negra and TuBa-D/Z, which are subsequently modified to allow a comparison of the differences. The results show that some modification have a negative effect while others have a positive influence.

@INCOLLECTION{Kuebler'2007
  ,AUTHOR    = {Sandra K\"{u}bler}
  ,TITLE     = {{H}ow {D}oes {T}reebank {A}nnotation {I}nfluence {P}arsing? {O}r {H}ow {N}ot to {C}ompare {A}pples and {O}ranges}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {79-88}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Alonso, Laura, Joan Antoni Capilla, Irene Castellon, Ana Fernandez-Montraveta & Gloria Vazquez. 2007. "The SenSem Project: Syntactico-Semantic Annotation of Sentences in Spanish". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 89-98. Amsterdam & Philadelphia: John Benjamins.

This paper presents SenSem, a project that aims to systematize the behavior of verbs in Spanish at the lexical, syntactic and semantic level. As part of the project, two resources are being built: a corpus where sentences are associated to their syntactico-semantic interpretation and a lexicon where each verb sense is linked to the corresponding annotated examples in the corpus. We also discuss some tendencies that can be observed in the current state of development.

@INCOLLECTION{Alonso-Capilla-Castellon-Fernandez-Montraveta-Vazquez'2007
  ,AUTHOR    = {Laura Alonso and Joan Antoni Capilla and Irene Castell\'{o}n and Ana Fern\'{a}ndez-Montraveta and Gloria V\'{a}zquez}
  ,TITLE     = {{T}he {SenSem} {P}roject: {S}yntactico-{S}emantic {A}nnotation of {S}entences in {S}panish}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {89-98}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

IV.  ANAPHORA & REFERRING EXPRESSIONS

The generation of referring expressions—noun phrases which speakers use to identify intended referents for their hearers—is the most well-established and agreed-upon subtask in natural language generation. This paper surveys the developments in the field that have brought us to this point, and characterises the state of the art in the area; we also point to unsolved problems and new directions waiting to be pursued.

@INCOLLECTION{Dale'2007
  ,AUTHOR    = {Robert Dale}
  ,TITLE     = {{G}enerating {R}eferring {E}xpressions: {P}ast, {P}resent and {F}uture}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {99-114}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Hinrichs, Erhard W., Katja Filippova & Holger Wunsch. 2007. "A Data-Driven Approach to Pronominal Anaphora Resolution for German". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 115-124. Amsterdam & Philadelphia: John Benjamins.

This paper reports on a hybrid architecture for computational anaphora resolution (CAR) of German that combines a rule-based pre-filtering component with a memory-based resolution module (using the Tilburg Memory Based Learner - TiMBL). The Car experiments performed on the TuBa-D/Z treebank data corroborate the importance of modelling aspects of discourse structure for robust, data-driven anaphora resolution. The best result with an F-measure of 0.734 achieved by these experiments outperforms the results reported by Schiehlen (2004), the only other study of German Car that is based on newspaper treebank data.

@INCOLLECTION{Hinrichs-Filippova-Wunsch'2007
  ,AUTHOR    = {Erhard W. Hinrichs and Katja Filippova and Holger Wunsch}
  ,TITLE     = {{A} {D}ata-{D}riven {A}pproach to {P}ronominal {A}naphora {R}esolution for {G}erman}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {115-124}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

V.  CLASSIFICATION

This paper shows that in the context of spam classification of weblogs, segmenting the urls beyond the standard punctuation is helpful. Many spam urls contain phrases in which the words are glued together in order to avoid spam filtering techniques based on unigrams and punctuation segmentation. An approach for segmenting tokens into words forming a phrase is proposed and validated. The accuracy of the implemented system is better than that of humans (78% vs. 76%).

@INCOLLECTION{Nicolov-Salvetti'2007
  ,AUTHOR    = {Nicolas Nicolov and Franco Salvetti}
  ,TITLE     = {{E}fficient {S}pam {A}nalysis for {W}eblogs through {URL} {S}egmentation}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,PAGES     = {125-136}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Ginter, Filip, Sampo Pyysalo & Tapio Salakoski. 2007. "Document Classification Using Semantic Networks with an Adaptive Similarity Measure". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 137-146. Amsterdam & Philadelphia: John Benjamins.

We consider supervised document classification where a semantic network is used to augment document features with their hypernyms. A novel document representation is introduced in which the contribution of the hypernyms to document similarity is determined by semantic network edge weights. We argue that the optimal edge weights are not a static property of the semantic network, but should rather be adapted to the given classification task. To determine the optimal weights, we introduce an efficient gradient descent method driven by the misclassifications of the k-nearest neighbors (kNN) classifier. The method iteratively adjusts the weights, increasing or decreasing the similarity of documents depending on their classes. We thoroughly evaluate the method using the kNN classifier and show it to statistically significantly outperform the bag-of-words representation as well as the more advanced hypernym density representation of Scott & Matwin (1998).

@INCOLLECTION{Ginter-Pyysalo-Salakoski'2007
  ,AUTHOR    = {Filip Ginter and Sampo Pyysalo and Tapio Salakoski}
  ,TITLE     = {{D}ocument {C}lassification {U}sing {S}emantic {N}etworks with an {A}daptive {S}imilarity {M}easure}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {137-146}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Mihalcea, Rada & Samer Hassan. 2007. "Text Summarization for Improved Text Classification". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 147-156. Amsterdam & Philadelphia: John Benjamins.

This paper explores the possible benefits of the interaction between text summarization and text classification. Through experiments performed on standard data sets, we show that techniques for extractive summarization can be effectively combined with classification methods, resulting in improved performance in a text categorization task. Moreover, comparative results suggest that the synergy between text summarization and text classification can be regarded as a new task-oriented evaluation testbed for automatic summarization.

@INCOLLECTION{Mihalcea-Hassan'2007
  ,AUTHOR    = {Rada Mihalcea and Samer Hassan}
  ,TITLE     = {{T}ext {S}ummarization for {I}mproved {T}ext {C}lassification}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {147-156}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Sporleder, Caroline & Alex Lascarides. 2007. "Exploiting Linguistic Cues to Classify Rhetorical Relations". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 157-166. Amsterdam & Philadelphia: John Benjamins.

We propose a method for automatically identifying rhetorical relations. We use supervised machine learning but exploit discourse markers to automatically label training data. Our models draw on a variety of linguistic cues to distinguish between relations. We show that these feature-rich models outperform the previously suggested bigram models by more than 20%, at least for small training sets. Our approach is therefore better suited to deal with relations for which it is difficult to automatically label a lot of training data because they are rarely signalled by unambiguous discourse markers.

@INCOLLECTION{Sporleder-Lascarides'2007
  ,AUTHOR    = {Caroline Sporleder and Alex Lascarides}
  ,TITLE     = {{E}xploiting {L}inguistic {C}ues to {C}lassify {R}hetorical {R}elations}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {157-166}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

VI.  TEXTUAL ENTAILMENT & QUESTION ANSWERING

This paper addresses Textual Entailment (i.e., recognizing that the meaning of a text entails the meaning of another text) using a Tree Edit Distance algorithm between the syntactic trees of the two texts. It also compares the contribution of lexical resources for recognizing textual entailment in an experiment carried on over the pascal-rte dataset.

@INCOLLECTION{Kouylekov-Magnini'2007
  ,AUTHOR    = {Milen Kouylekov and Bernardo Magnini}
  ,TITLE     = {{T}ree {E}dit {D}istance for {T}extual {E}ntailment}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {167-176}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Tiedemann, Jorg. 2007. "A Genetic Algorithm for Optimising Information Retrieval with Linguistic Features in Question Answering". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 177-186. Amsterdam & Philadelphia: John Benjamins.

In this chapter we describe a genetic algorithm applied to the optimisation of information retrieval (IR) within a Dutch question answering (QA) system. In our approach, information retrieval is used for passage retrieval to filter out relevant text passages for subsequent answer extraction modules in the QA system. We focus on the integration of linguistic features extracted from syntactic analyses of the texts in the retrieval component. For this we apply output of the Dutch dependency parser Alpino together with an off-the-shelf ir engine. Query parameters are optimised using a genetic algorithm to improve the overall performance of the QA system.

@INCOLLECTION{Tiedemann'2007
  ,AUTHOR    = {J\"{o}rg Tiedemann}
  ,TITLE     = {{A} {G}enetic {A}lgorithm for {O}ptimising {I}nformation {R}etrieval with {L}inguistic {F}eatures in {Q}uestion {A}nswering}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {177-186}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Rus, Vasile & Arthur Graesser. 2007. "Lexico-Syntactic Subsumption for Textual Entailment". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 187-196. Amsterdam & Philadelphia: John Benjamins.

We present here a graph-based approach to the task of textual entailment between a Text and Hypothesis. The approach takes into account the full lexico-syntactic context of both the Text and Hypothesis and relies heavily on the concept of subsumption. It starts with mapping the Text and Hypothesis into graph-structures where nodes represent concepts and edges represent lexico-syntactic relations among concepts. Based on a subsumption score between the Text-graph and Hypothesis-graph an entailment decision is made. A novel feature of our approach is the handling of negation. The results obtained from a standard entailment test data set are better than results reported by systems that use similar knowledge resources. We also found that a tf-idf approach performs close to chance for sentence-like Texts and Hypotheses.

@INCOLLECTION{Rus-Graesser'2007
  ,AUTHOR    = {Vasile Rus and Arthur Graesser}
  ,TITLE     = {{L}exico-{S}yntactic {S}ubsumption for {T}extual {E}ntailment}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {187-196}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Corley, Courtney, Andras Csomai & Rada Mihalcea. 2007. "A Knowledge-based Approach to Text-to-Text Similarity". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 197-206. Amsterdam & Philadelphia: John Benjamins.

In this paper, we present a knowledge-based method for measuring the semantic similarity of texts. Through experiments performed on two different applications: (1) paraphrase and entailment identification, and (2) word sense similarity, we show that this method outperforms the traditional text similarity metrics based on lexical matching.

@INCOLLECTION{Corley-Csomai-Mihalcea'2007
  ,AUTHOR    = {Courtney Corley and Andras Csomai and Rada Mihalcea}
  ,TITLE     = {{A} {K}nowledge-based {A}pproach to {T}ext-to-{T}ext {S}imilarity}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {197-206}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

VII.  ONTOLOGIES

This chapter describes a simple method to obtain semantic word classes from html documents. We previously showed that itemizations in html documents can contain semantically coherent word classes. However, not all the itemizations are semantically coherent. Our goal is to provide a simple method to extract only semantically coherent itemizations from html documents. Our new method can perform this task by obtaining hit counts from an existing search engine 2n times for an itemization consisting of n items.

@INCOLLECTION{Shinzato-Torisawa'2007
  ,AUTHOR    = {Keiji Shinzato and Kentaro Torisawa}
  ,TITLE     = {{A} {S}imple {WWW}-based {M}ethod for {S}emantic {W}ord {C}lass {A}cquisition}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {207-216}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Barbu, Eduard & Verginica Barbu Mititelu. 2007. "Automatic Building of Wordnets". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 217-226. Amsterdam & Philadelphia: John Benjamins.

This paper dealing with the automatic building of wordnets starts by stating the assumptions behind such enterprise and then presents a two-phase methodology for automatically building a target wordnet strictly aligned with an already available wordnet (source wordnet). In the first phase the synsets for the target language are automatically generated and mapped onto the source language synsets using a series of heuristics. In the second phase the salient relations that can be automatically imported are identified and the procedure for their import is explained. The idea behind all our heuristics will be stated, the heuristics employed will be presented and their success evaluated against a case study: automatically building a Romanian wordnet using the Princeton WordNet.

@INCOLLECTION{Barbu-Mititelu'2007
  ,AUTHOR    = {Eduard Barbu and Verginica Barbu Mititelu}
  ,TITLE     = {{A}utomatic {B}uilding of {W}ordnets}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {217-226}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

VIII.  MACHINE TRANSLATION

This paper addresses the problem of bilingual lexicon extraction and lexical transfer selection, in the framework of computer-aided and machine translation. The proposed method relies on parallel corpora, annotated at part of speech and lemma level. We first extract a bilingual lexicon using unsupervised statistical techniques. For each word with more than one translation candidates we build context vectors, in order to aid the selection of the contextually correct translation equivalent. The method achieves an overall precision of ca. 85% while the maximum recall reaches 75%.

@INCOLLECTION{Piperidis-Dimitrakis-Balta'2007
  ,AUTHOR    = {Stelios Piperidis and Panagiotis Dimitrakis and Irene Balta}
  ,TITLE     = {{L}exical {T}ransfer {S}election {U}sing {A}nnotated {P}arallel {C}orpora}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {227-236}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Arranz, Victoria, Elisabet Comelles & David Farwell. 2007. "Multi-Perspective Evaluation of the FAME Speech-to-Speech Translation System for Catalan, English and Spanish". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 237-246. Amsterdam & Philadelphia: John Benjamins.

This paper describes the final evaluation of the FAME interlinguabased speech-to-speech translation system for Catalan, English and Spanish. It is an extension of the already existing NESPOLE! System that translates between English, French, German and Italian. However, the FAME modules have now been integrated in an Open Agent Architecture platform. We describe three types of evaluation (task-oriented, performance-based and of user satisfaction) and present their results when applied to our system. We also compare the results of the system with those obtained by a stochastic translator developed independently within the FAME project.

@INCOLLECTION{Arranz-Comelles-Farwell'2007
  ,AUTHOR    = {Victoria Arranz and Elisabet Comelles and David Farwell}
  ,TITLE     = {{M}ulti-{P}erspective {E}valuation of the {FAME} {S}peech-to-{S}peech {T}ranslation {S}ystem for {C}atalan, {E}nglish and {S}panish}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {237-246}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Varga, Daniel, Peter Halacsy, Andras Kornai, Viktor Nagy, Laszlo Nemeth & Viktor Tron. 2007. "Parallel Corpora for Medium Density Languages". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 247-258. Amsterdam & Philadelphia: John Benjamins.

The choice of natural language technology appropriate for a given language is greatly impacted by ‘density’ (availability of digitally stored material). More than half of the world speaks medium density languages, yet many of the methods appropriate for high or low density languages yield suboptimal results when applied to the medium density case. In this paper we describe a general methodology for rapidly collecting, building, and aligning parallel corpora for medium density languages, illustrating our main points on the case of Hungarian, Romanian, and Slovenian. We also describe and evaluate the hybrid sentence alignment method we are using.

@INCOLLECTION{Varga-Halacsy-Kornai-Nagy-Nemeth-Tron'2007
  ,AUTHOR    = {D\'{a}niel Varga, P\'{e}ter Hal\'{a}csy, Andr\{a}s Kornai, Viktor Nagy, L\'{a}szl\'{o} N\'{e}meth & Viktor Tr\'{o}n}
  ,TITLE     = {{P}arallel {C}orpora for {M}edium {D}ensity {L}anguages}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {247-258}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

IX.  CORPORA

There is a rich literature documenting the fact that text and document properties affect the success of techniques in data driven Natural Language Processing and Information Retrieval. At the same time, there is little by way of a systematic investigation of the precise role played by data. This paper sets out the rationale for achieving a better understanding of the role of data, and points out some undesirable methodological, epistemological and practical consequences of not doing so. Some rough sparseness measures are used to show that standard datasets differ in important ways. A case is made for exploring the role of data by compiling dataset profiles. These would contain measures tailored to highlight inherent bias in a collection, reflecting its fitness to support a technique in the context of a task.

@INCOLLECTION{DeRoeck'2007
  ,AUTHOR    = {Anne De Roeck}
  ,TITLE     = {{T}he {R}ole of {D}ata in {NLP}: {T}he {C}ase for {D}ataset {P}rofiling}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {259-266}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

De Roeck, Anne, Avik Sarkar & Paul H. Garthwaite. 2007. Even very frequent function words do not distribute homogeneously   267 Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 267-276. Amsterdam & Philadelphia: John Benjamins.

We have known for some time that content words have "bursty" distributions in text. In contrast, much of the literature assumes that function words are uninformative and they distribute homogeneously. We describe two sets of experiments showing that assumptions of homogeneity do not hold, even for the distribution of extremely frequent function words. In the first experiment, we investigate the behaviour of very frequent function words in the tipster collection by postulating a "homogeneity assumption", which we then defeat in a series of experiments based on the χ² test. Results show that it is statistically unreasonable to assume homogeneous term distributions within a corpus. We also found that document collections are not neutral with respect to the property of homogeneity, even for very frequent function words. In the second set of experiments, we model the gaps between successive occurrences of a particular term using a mixture of exponential distributions. Where the "homogeneity assumption" holds these gaps should be uniformly distributed across the entire corpus. Using the model we demonstrate that gaps are not uniformly distributed, and even very frequent terms occur in bursts.

@INCOLLECTION{DeRoeck-Sarkar-Garthwaite'2007
  ,AUTHOR    = {Anne De Roeck and Avik Sarkar and Paul H. Garthwaite}
  ,TITLE     = {{E}ven {V}ery {F}requent {F}unction {W}ords do not {D}istribute {H}omogeneously}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {267-276}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Specia, Lucia, Maria das Gracas V. Nunes & Mark Stevenson. 2007. "Exploiting Parallel Texts to Produce a Multilingual Sense Tagged Corpus for Word Sense Disambiguation". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 277-286. Amsterdam & Philadelphia: John Benjamins.

We describe an approach to the automatic creation of a sense tagged corpus intended to train a word sense disambiguation (WSD) system for English-Portuguese machine translation. The approach uses parallel corpora, translation dictionaries and a set of straightforward heuristics. In an evaluation with nine corpora containing 10 ambiguous verbs, the approach achieved an average precision of 94%, compared with 58% when a state of the art statistical alignment tool was used. The resulting corpus consists of 113,802 tagged sentences and can be readily used to train a supervised machine learning algorithm to build WSD models for use in machine translation systems.

@INCOLLECTION{Specia-Nunes-Stevenson'2007
  ,AUTHOR    = {Lucia Specia and Maria das Gra\c{c}as V. Nunes and Mark Stevenson}
  ,TITLE     = {{E}xploiting {P}arallel {T}exts to {P}roduce a {M}ultilingual {S}ense {T}agged {C}orpus for {W}ord {S}ense {D}isambiguation}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {277-286}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}

Chantree, Francis, Alistair Willis, Adam Kilgarriff & Anne de Roeck. 2007. "Detecting Dangerous Coordination Ambiguities Using Word Distribution". Recent Advances in Natural Language Processing IV ed. by N.Nicolov, K.Botcheva, G.Angelova & R.Mitkov (= Current Issues in Libguistic Theory, CILT 292), pp. 287-296. Amsterdam & Philadelphia: John Benjamins.

In this paper we present heuristics for resolving coordination ambiguities. We test the hypothesis that the most likely reading of a coordination can be predicted using word distribution information from a generic corpus. Our heuristics are based upon the relative frequency of the coordination in the corpus, the distributional similarity of the coordinated words, and the collocation frequency between the coordinated words and their modifiers. These heuristics have varying but useful predictive power. They also take into account our view that many ambiguities cannot be effectively disambiguated, since human perceptions vary widely.

@INCOLLECTION{Chantree-Willis-Kilgarriff-DeRoeck'2007
  ,AUTHOR    = {Francis Chantree and Alistair Willis and Adam Kilgarriff and Anne de Roeck}
  ,TITLE     = {{D}etecting {D}angerous {C}oordination {A}mbiguities {U}sing {W}ord {D}istribution}
  ,BOOKTITLE = {{R}ecent {A}dvances in {N}atural {L}anguage {P}rocessing {IV}: {S}elected {P}apers from {RANLP} 2005}
  ,SERIES    = {{C}urrent {I}ssues in {L}inguistic {T}heory}
  ,VOLUME    = {292}
  ,EDITOR    = {Nicolas Nicolov and Kalina Bontcheva and Galia Angelova and Ruslan Mitkov}
  ,PAGES     = {287-296}
  ,PUBLISHER = {John Benjamins}
  ,ADDRESS   = {Amsterdam \& Philadelphia}
  ,YEAR      = {2007}
}