DOCSLIB.ORG

Dijkstra-WSA: a Graph-Based Approach to Word Sense Alignment

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Dijkstra-WSA: a Graph-Based Approach to Word Sense Alignment Dijkstra-WSA: A Graph-Based Approach to Word Sense Alignment Michael Matuschek ‡ and Iryna Gurevych †‡ Ubiquitous Knowledge Processing Lab (UKP-DIPF), German† Institute for Educational Research and Educational Information Schloßstr. 29, 60486 Frankfurt, Germany ‡Ubiquitous Knowledge Processing Lab (UKP-TUDA), Department of Computer Science, Technische Universitat¨ Darmstadt Hochschulstr. 10, 64289 Darmstadt, Germany http://www.ukp.tu-darmstadt.de Abstract best for all purposes, as different LSRs cover differ- ent words, senses and information types. In this paper, we present Dijkstra-WSA, a These considerations have sparked increasing re- novel graph-based algorithm for word sense search efforts in the area of word sense alignment alignment. We evaluate it on four different pairs of lexical-semantic resources with dif- (WSA). It has been shown that aligned resources ferent characteristics (WordNet-OmegaWiki, can indeed lead to better performance than using the WordNet-Wiktionary, GermaNet-Wiktionary resources individually. Examples include seman- and WordNet-Wikipedia) and show that it tic parsing using FrameNet (FN), WN, and VerbNet achieves competitive performance on 3 out (VN) (Shi and Mihalcea, 2005), word sense disam- of 4 datasets. Dijkstra-WSA outperforms the biguation using an alignment of WN and Wikipedia state of the art on every dataset if it is com- (WP) (Navigli and Ponzetto, 2012) and semantic bined with a back-off based on gloss similar- role labeling using a combination of PropBank, VN ity. We also demonstrate that Dijkstra-WSA is not only flexibly applicable to different re- and FN in the SemLink project (Palmer, 2009). sources but also highly parameterizable to op- Some of these approaches to WSA either rely heav- timize for precision or recall. ily on manual labor (e.g. Shi and Mihalcea (2005)) or on information which is only present in few resources such as the most frequent sense (MFS) 1 Introduction (Suchanek et al., 2008). This makes it difficult to Lexical-semantic resources (LSRs) are a corner- apply them to a larger set of resources. stone for many Natural Language Processing (NLP) In earlier work, we presented the large-scale re- applications such as word sense disambiguation source UBY (Gurevych et al., 2012). It contains (WSD) and information extraction. However, the nine resources in two languages which are mapped growing demand for large-scale resources in dif- to a uniform representation using the LMF standard ferent languages is hard to meet. The Princeton (Eckle-Kohler et al., 2012). They are thus struc- WordNet (WN) (Fellbaum, 1998) is widely used for turally interoperable. UBY contains pairwise sense English, but for most languages corresponding re- alignments between a subset of these resources, and sources are considerably smaller or missing. Col- this work also presented a framework for creat- laboratively constructed resources like Wiktionary ing alignments based on the similarity of glosses (WKT) and OmegaWiki (OW) provide a viable op- (Meyer and Gurevych, 2011). However, it is not tion for such cases and seem especially suitable clear to what extent this approach can be applied to for smaller languages (Matuschek et al., 2013), but resources which lack this kind of information (see there are still considerable gaps in coverage which Section 3). need to be filled. A related problem is that there In summary, aligning senses is a key requirement usually does not exist a single resource which works for semantic interoperability of LSRs to increase the 151 Transactions of the Association for Computational Linguistics, 1 (2013) 151–164. Action Editor: Patrick Pantel. Submitted 12/2012; Revised 2/2013; Published 5/2013. c 2013 Association for Computational Linguistics. Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/tacl_a_00217 by guest on 28 September 2021 coverage and effectiveness in NLP tasks. Still, exist- two LSRs. A pair of aligned senses denote the same ing efforts are mostly focused on specific types of re- meaning. E.g., the two senses of letter “The conven- sources (most often requiring glosses) or application tional characters of the alphabet used to represent scenarios. In this paper, we propose an approach to speech” and “A symbol in an alphabet, bookstave” alleviate this and present Dijkstra-WSA, a novel, ro- (taken from WN and WKT, respectively) are clearly bust algorithm for word sense alignment which is equivalent and should be aligned. applicable to a wide variety of resource pairs and languages. For the first time, we apply a graph-based 2.2 Evaluation Resources algorithm which works on full graph representations For the evaluation of Dijkstra-WSA, we align four of both resources to word sense alignment. This en- pairs of LSRs used in previous work, namely WN- ables us to take a more abstract perspective and re- OW (Gurevych et al., 2012), WN-WKT (Meyer and duce the problem of identifying equivalent senses to Gurevych, 2011), GN-WKT (Henrich et al., 2011) the problem of matching nodes in these graphs. Also and WN-WP (Niemann and Gurevych, 2011). Our for the first time, we comparatively evaluate a WSA goal is to cover resources with different character- algorithm on a variety of different datasets with dif- istics: Expert-built (WN, GN) and collaboratively ferent characteristics. constructed LSRs (WP, WKT, OW), resources in The key properties of Dijkstra-WSA are: different languages (English and German) and also Robustness The entities within the LSRs which resources with few sense descriptions (GN) or se- are to be aligned (usually senses or synsets) are mod- mantic relations (WKT). We contrastively discuss eled as nodes in the graph. These nodes are con- the results of the Dijkstra-WSA algorithm on these nected by an edge if they are semantically related. different datasets and relate the results to the prop- While, for instance, semantic relations lend them- erties of the LSRs involved. Moreover, using exist- selves very well to deriving edges, different possi- ing datasets ensures comparability to previous work bilities for graph construction are equally valid as which discusses only one dataset at a time. the algorithm is agnostic to the origin of the edges. WordNet (WN) (Fellbaum, 1998) is a lexical re- Language-independence No external resources source for the English language created at Princeton such as corpora or other dictionaries are needed; the University. The resource is organized in sets of syn- graph construction and alignment only rely on the onymous words (synsets) which are represented by information from the considered LSRs. glosses (sometimes accompanied by example sen- Flexibility The graph construction as well as the tences) and organized in a hierarchy. The latest ver- actual alignment are highly parameterizable to ac- sion 3.0 contains 117,659 synsets. commodate different requirements regarding preci- Wikipedia (WP) is a freely available, multilin- sion or recall. gual online encyclopedia. WP can be edited by ev- The rest of this paper is structured as follows: ery Web user, which causes rapid growth: By Febru- In Section 2 we give a precise problem description ary 2013 the English WP contained over 4,000,000 and introduce the resources covered in our experi- article pages. Each article usually describes a dis- ments, in Section 3 we discuss some related work, tinct concept, and articles are connected by hyper- while our graph-based algorithm Dijkstra-WSA is links within the article texts. presented in Section 4. We describe an evaluation Wiktionary (WKT) is the dictionary pendant to on four datasets with different properties, including WP. By February 2013 the English WKT contained an error analysis, in Section 5 and conclude in Sec- over 3,200,000 article pages, while the German edi- tion 6, pointing out directions for future work. tion contained over 200,000 ones. For each word, multiple senses can be encoded. Similar to WN, 2 Notation and Resources they are represented by a gloss and usage exam- ples. There also exist hyperlinks to synonyms, hy- 2.1 Problem Description pernyms, meronyms etc. The targets of these rela- A word sense alignment, or alignment for short, is tions are not senses, however, but merely lexemes formally defined as a list of pairs of senses from (i.e. the relations are not disambiguated). 152 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/tacl_a_00217 by guest on 28 September 2021 LSRs P /R/F1/Acc. Approach Meyer and Gurevych (2011) WN-WKT 0.67/0.65/0.66/0.91 Gloss similarity + Machine learning Niemann and Gurevych (2011) WN-WP 0.78/0.78/0.78/0.95 Gloss similarity + Machine learning Henrich et al. (2011) GN-WKT 0.84/0.85/0.84/0.94 Pseudo-gloss overlap de Melo and Weikum (2010) WN-WP 0.86/NA/NA/NA Gloss/article overlap Laparra et al. (2010) FN-WN 0.79/0.79/0.79/NA Dijkstra-SSI+ (WSD algorithm) Navigli (2009) WN 0.64/0.64/0.64/NA Graph-based WSD of WN glosses Ponzetto and Navigli (2009) WN-WP NA/NA/NA/0.81 Graph-based, only for WP categories Navigli and Ponzetto (2012) WN-WP 0.81/0.75/0.78/0.83 Graph-based WSA using WN relations Table 1: Summary of various approaches to WSA. “NA” stands for “Not Available”. OmegaWiki (OW) is a freely editable online glosses, which are not present in every case (e.g. for dictionary like WKT. However, there do not ex- VN). Moreover, as it involves supervised machine ist distinct language editions as OW is organized learning, it requires the initial effort of manually an- in language-independent concepts (“Defined Mean- notating a sufficient amount of training data. Hen- ings”) to which lexicalizations in various languages rich et al. (2011) use a similar approach for align- are attached. These can be considered as multilin- ing GN and WKT.
Load more

Recommended publications
  • Exploring and Visualizing Wordnet Data with Germanet Rover
    Exploring and Visualizing Wordnet Data with GermaNet Rover Marie Hinrichs, Richard Lawrence, Erhard Hinrichs University of Tübingen, Germany www.clarin-d.net GERMANET AND ROVER 1 www.clarin-d.net GermaNet: a Wordnet for German A wordnet groups synonyms into synsets and represents relations between synsets. The hypernym relation forms a hierarChiCal graph structure. 2 www.clarin-d.net Rover: a web appliCation for GermaNet Rover displays the data in GermaNet in an interaCtive interfaCe designed for researChers. It offers: • advanced searChing for synsets • visualizing the hypernym graph • CalCulating synsets’ semantiC relatedness via graph-based measures 3 www.clarin-d.net SYNSET SEARCH 4 www.clarin-d.net Synset SearCh Overview Try it: https://weblicht.sfs.uni-tuebingen.de/rover/search 5 www.clarin-d.net SearCh Options Advanced searChes for • search by regular synsets: expression and edit distance • restrict results by grammatical category, semantic class, and orthographic variant 6 www.clarin-d.net Results List A summary of each synset in the search results includes: • words in the synset • its semantic class • associated Wiktionary definitions • summary of conceptual relations to other synsets 7 www.clarin-d.net Conceptual Relations Selecting a synset displays details about its conceptual relations: • a network diagram of the synset’s hypernyms • related synsets, displayed as navigation buttons 8 www.clarin-d.net LexiCal Units and Relations More details about the words in the seleCted synset: • lexically-related words, displayed as navigation buttons • Interlingual index reCords (pointers to Princeton WordNet) • examples and assoCiated frame types • deComposition of Compound nouns 9 www.clarin-d.net SEMANTIC RELATEDNESS 10 www.clarin-d.net SemantiC Relatedness Overview Try it: https://weblicht.sfs.uni-tuebingen.de/rover/semrel 11 www.clarin-d.net SemantiC Relatedness Measures Rover supports six graph-based measures of semantic relatedness between pairs of synsets: 1.
    [Show full text]
  • Calculating Semantic Relatedness with Germanet
    Organismus, Lebewesen ‘organism, being’ ... ... Haustier ... ‘pet’ Baum Katze Hund ‘tree’ ‘cat’ ‘dog’ Calculating Semantic Relatedness with GermaNet Verena Henrich, Düsseldorf, 19. Februar 2015 Semantic Relatedness Measures – Overview • Path-based measures - Leacock and Chodorow (1998) - Wu and Palmer (1994) - Hirst and St-Onge (1998) - Path • Information-content-based measures - Resnik (1995) - Jiang and Conrath (1997) - Lin (1998) • Gloss-based measures: several variants of Lesk (1986) - Glosses from GermaNet and/or Wiktionary - Lexical fields from GermaNet and/or Wiktionary Semantic Relatedness Measures – Terminology • Relatedness/similarity is GNROOT GermaNet’s root node calculated between the two semantic concepts A (synset nodes) s1 and s2 B C D Haustier LCS(s1, s2) = lowest common subsumer ‘pet’ of s1 and s2 synset s1 Katze Hund synset s2 ‘cat’ ‘dog’ Semantic Relatedness Measures – Terminology GNROOT A B C depth(s2) = 6 length of the shortest path D from synset s2 to GNROOT LCS Haustier ‘pet’ s1 s2 Katze Hund ‘cat’ ‘dog’ Semantic Relatedness Measures – Terminology GNROOT A B C D length(s1, s2) = 2 LCS shortest path Haustier between s1 and s2 ‘pet’ s1 s2 Katze Hund ‘cat’ ‘dog’ Path-based Relatedness Measures GNROOT A B Similarity by Wu and Palmer (1994): 2 · depth(LCS(s1, s2)) C sim (s1, s2) = wup depth(s1) + depth(s2) D LCS sim (cat, dog) = 0.83 Haustier wup ‘pet’ where • depth(LCS) = 5 s1 s2 • depth(cat) = 6 Katze Hund • depth(dog) = 6 ‘cat’ ‘dog’ Path-based Relatedness Measures GNROOT A B LCS simwup(cat, tree) = 0.5 Organismus,
    [Show full text]
  • The Germanet Editing Tool Verena Henrich, Erhard Hinrichs University of Tübingen, Department of Linguistics Wilhelmstr
    GernEdiT – The GermaNet Editing Tool Verena Henrich, Erhard Hinrichs University of Tübingen, Department of Linguistics Wilhelmstr. 19, 72074 Tübingen, Germany E-mail: [email protected], [email protected] Abstract This paper introduces GernEdiT (short for: GermaNet Editing Tool), a new graphical user interface for the lexicographers and developers of GermaNet, the German version of the Princeton WordNet. GermaNet is a lexical-semantic net that relates German nouns, verbs, and adjectives. Traditionally, lexicographic work for extending the coverage of GermaNet utilized the Princeton WordNet development environment of lexicographer files. Due to a complex data format and no opportunity of automatic consistency checks, this process was very error prone and time consuming. The GermaNet Editing Tool GernEdiT was developed to overcome these shortcomings. The main purposes of the GernEdiT tool are, besides supporting lexicographers to access, modify, and extend GermaNet data in an easy and adaptive way, as follows: Replace the standard editing tools by a more user-friendly tool, use a relational database as data storage, support export formats in the form of XML, and facilitate internal consistency and correctness of the linguistic resource. All these core functionalities of GernEdiT along with the main aspects of the underlying lexical resource GermaNet and its current database format are presented in this paper. 1 Introduction data such as appropriate linking of lexical units with The traditional development of GermaNet1 (Kunze and synsets, connectedness of the synset graph, and Lemnitzer, 2002) was based on lexicographer files. automatic closure among relations and their inverse These were originally developed for the English counterparts.
    [Show full text]
  • Porting a Crowd-Sourced German Lexical Semantics Resource to Ontolex-Lemon
    Proceedings of eLex 2019 Porting a Crowd-Sourced German Lexical Semantics Resource to Ontolex-Lemon Thierry Declerck 1,2 , Melanie Siegel 3 1 German Research Center for Artificial Intelligence, Stuhlsatzenhausweg 3, 66123 Saarbrücken, Germany 2 Austrian Centre for Digital Humanities, Sonnenfelsgasse 19, 1010 Vienna, Austria 3 Darmstadt University of Applied Science, Max-Planck-Str. 2, 64807 Dieburg, Germany E-mail: [email protected], [email protected] Abstract In this paper we present our work consisting of mapping the recently created open source German lexical semantics resource “Open-de-WordNet” (OdeNet) into the OntoLex-Lemon format. OdeNet was originally created in order to be integrated in the Open Multilingual Wordnet initiative. One motivation for porting OdeNet to OntoLex-Lemon is to publish in the Linguistic Linked Open Data cloud this new WordNet-compliant resource for German. At the same time we can with the help of OntoLex-Lemon link the lemmas of OdeNet to full lexical descriptions and so extend the linguistic coverage of this new WordNet resource, as we did for French, Italian and Spanish wordnets included in the Open Multilingual Wordnet collection. As a side effect, the porting of OdeNet to OntoLex-Lemon helped in discovering some issues in the original data. Keywords: Open Multilingual Wordnet; OntoLex-Lemon; OdeNet; Lexical Semantics 1. Introduction Wordnets are well-established lexical resources with a wide range of applications in various Natural Language Processing (NLP) fields, like Machine Translation, Information Retrieval, Query Expansion, Document Classification, etc. (Morato et al., 2004). For more than twenty years they have been elaborately set up and maintained by hand, especially the original Princeton WordNet of English (PWN) (Fellbaum, 1998).
    [Show full text]
  • Textimager-VSD
    Wahed Hemati 3007 80 822 147 Computer science Text Technology Lab [email protected] Dissertation TextImager-VSD Large Scale Verb Sense Disambiguation and Named Entity Recognition in the Context of TextImager Abdul Wahed Hemati Frankfurt, 12.2019 Goethe University Frankfurt am Main Prof. Alexander Mehler Prof. Visvanathan Ramesh 1. Danksagung An dieser Stelle möchte ich mich bei all denjenigen bedanken, die mich während der Anfertigung dieser Dissertation unterstützt und motiviert haben. Die Arbeit wäre ohne Unterstützung einer Vielzahl von Leuten nicht möglich gewesen. Ganz besonders gilt dieser Dank Herrn Prof. Dr. Alexander Mehler, der meine Arbeit und somit auch mich betreut hat. Durch stetig kritisches Hinterfragen und konstruktive Kritik gab er mir wertvolle Hinweise mit auf den Weg. Er stand nicht nur mit wissenschaftlichem Rat und fachlichen Diskussionen zur Seite, sondern auch mit moralischer Unterstützung und Motivation. Die zahlreichen Gespräche auf intellektueller und persönlicher Ebene werden mir immer als bereichernder und konstruktiver Austausch in Erinnerung bleiben. Ich bin dankbar dafür, dass er mich stets sehr gut betreut und dazu gebracht hat, über meine Grenzen hinaus zu denken. Vielen Dank für die Geduld und Mühen. Ich danke Herrn Prof. Dr. Visvanathan Ramesh für die hilfsbereite und wis- senschaftliche Betreuung als Zweitgutachter. Ich bedanke mich für die Möglichkeit, diese Arbeit im Text-Technology-Lab der Goethe-Universität Frankfurt anfertigen zu können. Ich hatte die Möglichkeit, von fachlich kompetenten Kollegen und Kolleginnen beraten und betreut zu werden. Dadurch bin ich fachlich und zwischenmenschlich mit ihnen gewach- sen. Nicht zuletzt gebührt meinen Eltern der höchste Dank, die während des Lebens und vor allem während des Studiums immer an meiner Seite standen.
    [Show full text]
  • Tools for Germanet
    Exploring and Navigating: Tools for GermaNet Marc Finthammer, Irene Cramer Faculty of Cultural Studies Dortmund University of Technology, Germany marc.fi[email protected], [email protected] Abstract GermaNet is regarded to be a valuable resource for many German NLP applications, corpus research, and teaching. This demo presents three GUI-based tools meant to facilitate the exploration of and navigation through GermaNet. The GermaNet Explorer exhibits various retrieval, sort, filter and visualization functions for words/synsets and also provides an insight into the modeling of GermaNet’s semantic relations as well as its representation as a graph. The GermaNet-Measure-API and GermaNet Pathfinder offer methods for the calcu- lation of semantic relatedness based on GermaNet as a resource and the visualization of (semantic) paths between words/synsets. The GermaNet-Measure-API furthermore features a flexible interface, which facilitates the integration of all relatedness measures provided into user-defined applications. We have already used the three tools in our research on thematic chaining and thematic indexing, as a tool for the manual annotation of lexical chains, and as a resource in our courses on corpus linguistics and semantics. 1. Motivation 2. GermaNet Explorer Many researchers working with GermaNet have the same GermaNet (Lemnitzer and Kunze, 2002), the German experience: they lose their way in the rich, complex equivalent of WordNet (Fellbaum, 1998), represents a valu- structure of its XML-representation. In order to solve able lexical-semantic resource for numerous German natu- this problem, we implemented the GermaNet Explorer, ral language processing (NLP) applications. However, in of which a screenshot is shown in Figure 1.
    [Show full text]
  • Arxiv:2106.00055V1 [Cs.CL] 31 May 2021
    More than just Frequency? Demasking Unsupervised Hypernymy Prediction Methods Thomas Bott, Dominik Schlechtweg and Sabine Schulte im Walde Institute for Natural Language Processing, University of Stuttgart, Germany fbottts,schlecdk,[email protected] Abstract which word in a pair of words is the hypernym and which is the hyponym). The target subtask This paper presents a comparison of unsuper- of the current study is hypernymy prediction: we vised methods of hypernymy prediction (i.e., to predict which word in a pair of words such perform a comparative analysis of a class of ap- as fish–cod is the hypernym and which the hy- proaches commonly refered to as unsupervised hy- ponym). Most importantly, we demonstrate pernymy methods (Weeds et al., 2004; Kotlerman across datasets for English and for German et al., 2010; Clarke, 2012; Lenci and Benotto, 2012; that the predictions of three methods (Weeds- Santus et al., 2014). These methods all rely on the Prec, invCL, SLQS Row) strongly overlap and distributional hypothesis (Harris, 1954; Firth, 1957) are highly correlated with frequency-based that words which are similar in meaning also occur predictions. In contrast, the second-order in similar linguistic distributions. In this vein, they method SLQS shows an overall lower accu- racy but makes correct predictions where the exploit asymmetries in distributional vector space others go wrong. Our study once more con- representations, in order to contrast hypernym and firms the general need to check the frequency hyponym vectors. bias of a computational method in order to While these unsupervised hypernymy prediction identify frequency-(un)related effects.
    [Show full text]
  • Standardizing Wordnets in the ISO Standard LMF: Wordnet-LMF for Germanet
    Standardizing Wordnets in the ISO Standard LMF: Wordnet-LMF for GermaNet Verena Henrich Erhard Hinrichs Universtiy of Tübingen Universtiy of Tübingen Department of Linguistics Department of Linguistics verena.henrich@uni- erhard.hinrichs@uni- tuebingen.de tuebingen.de tions, (annotated) text corpora, and dictionar- Abstract ies. It is fair to say that it has become common practice among developers of new linguistic It has been recognized for quite some resources to consult TEI guidelines and ISO time that sustainable data formats play standards in order to develop standard- an important role in the development conformant encoding schemes that serve as an and curation of linguistic resources. interchange format and that can be docu- The purpose of this paper is to show mented and validated by Document Type how GermaNet, the German version of Definitions (DTD) and XML schemata. the Princeton WordNet, can be con- However, for resources that were developed verted to the Lexical Markup Frame- prior to or largely in parallel with the emerging work (LMF), a published ISO standard acceptance of markup languages and of emerg- (ISO-24613) for encoding lexical re- ing encoding standards, the situation is far sources. The conversion builds on more heterogeneous. A wide variety of legacy Wordnet-LMF, which has been pro- formats exists, many of which have persisted posed in the context of the EU due to existing user communities and the KYOTO project as an LMF format for availability of tools that can process only such wordnets. The present paper proposes a idiosyncratic formats. The development of number of crucial modifications and a wordnets for a large number of languages is a set of extensions to Wordnet-LMF that typical example of a type of linguistic re- are needed for conversion of wordnets source, where legacy formats still persist as a in general and for conversion of Ger- de facto standard.
    [Show full text]
  • Computational Linguistics in Practice: Weblicht and Germanet Two Projects at the Department of Linguistics at the University of Tübingen
    Computational Linguistics in Practice: WebLicht and GermaNet Two Projects at the Department of Linguistics at the University of Tübingen Verena Henrich University of Tübingen Department of Linguistics November 30, 2010 Who I am: Verena Henrich • 2009: Master in Computer Science at h_da - Lecture about Natural Language Processing (NLP) - Two semesters in Iceland - Topic of master thesis about NLP • Since 2009: Researcher at the Department of General and Computational Linguistics at the University of Tübingen - First task: development of an editor for the German wordnet (GermaNet) - Further project that I will introduce today: WebLicht - PhD plans: word sense disambiguation with GermaNet 2 | Verena Henrich November 30, 2010 GermaNet – A German Wordnet 3 | Verena Henrich November 30, 2010 GermaNet: A German Wordnet • GermaNet is a lexical resource covering the German base vocabulary • It is a lexical semantic network • Belongs to the family of wordnets modeled after the Princeton WordNet for English • GermaNet is divided into 3 word categories: - Adjectives - Nouns - Verbs • Words are ordered according to their meaning 4 | Verena Henrich November 30, 2010 GermaNet: Lexical Units • Word meanings are represented by lexical units • A lexical unit specifies one form and one meaning (i.e. reading) of a word • Examples: - “Bank“ has 2 readings . Reading 1: [Bank, {Sitzbank}] (bench) . Reading 2: [Bank, {Geldinstitut}] (financial institution) - “Leiter” has 3 readings . Reading 1: [Leiter, {Steiggerät}] (ladder) . Reading 2: [Leiter, {Verantwortlicher,
    [Show full text]
  • Modelling and Processing Wordnets in OWL
    Lüngen, Harald Beißwenger, Michael Selzam, Bianca Storrer, Angelika Modelling and Processing Wordnets in OWL To appear in: Alexander Mehler, Kai-Uwe Kühnberger, Hennig Lobin, Harald Lüngen, Angelika Storrer & Andreas Witt (eds.): Modeling, Learning and Processing of Text Technological Data Structures. Dordrecht: Springer. PREPRINT VERSION Modelling and Processing Wordnets in OWL Harald Lüngen, Michael Beißwenger, Bianca Selzam and Angelika Storrer Abstract In this contribution, we discuss and compare alternative options of mod- elling the entities and relations of wordnet-like resources in the Web Ontology Lan- guage OWL. Based on different modelling options, we developed three models of representing wordnets in OWL, i.e. the instance model, the class model, and the metaclass model. These OWL models mainly differ with respect to the ontological status of lexical units (word senses) and the synsets. While in the instance model lexical units and synsets are represented as individuals, in the class model they are represented as classes; both model types can be encoded in the dialect OWL DL. As a third alternative, we developed a metaclass model in OWL FULL, in which lexical units and synsets are defined as metaclasses, the individuals of which are classes themselves. We apply the three OWL models to each of three wordnet-style resources: (1) a subset of the German wordnet GermaNet, (2) the wordnet-style do- main ontology TermNet, and (3) GermaTermNet, in which TermNet technical terms and GermaNet synsets are connected by means of a set of “plugin” relations. We report on the results of several experiments in which we evaluated the performance of querying and processing these different models: (1) A comparison of all three OWL models (class, instance, and metaclass model) of TermNet in the context of automatic text-to-hypertext conversion, (2) an investigation of the potential of the GermaTermNet resource by the example of a wordnet-based semantic relatedness calculation.
    [Show full text]
  • A Comparative Evaluation of Word Sense Disambiguation Algorithms for German
    A Comparative Evaluation of Word Sense Disambiguation Algorithms for German Verena Henrich, Erhard Hinrichs University of Tubingen,¨ Department of Linguistics Wilhelmstr. 19, 72074 Tubingen,¨ Germany fverena.henrich,[email protected] Abstract The present paper explores a wide range of word sense disambiguation (WSD) algorithms for German. These WSD algorithms are based on a suite of semantic relatedness measures, including path-based, information-content-based, and gloss-based methods. Since the individual algorithms produce diverse results in terms of precision and thus complement each other well in terms of coverage, a set of combined algorithms is investigated and compared in performance to the individual algorithms. Among the single algorithms considered, a word overlap method derived from the Lesk algorithm that uses Wiktionary glosses and GermaNet lexical fields yields the best F-score of 56.36. This result is outperformed by a combined WSD algorithm that uses weighted majority voting and obtains an F-score of 63.59. The WSD experiments utilize the German wordnet GermaNet as a sense inventory as well as WebCAGe (short for: Web-Harvested Corpus Annotated with GermaNet Senses), a newly constructed, sense-annotated corpus for this language. The WSD experiments also confirm that WSD performance is lower for words with fine-grained sense distinctions compared to words with coarse-grained senses. Keywords: Word sense disambiguation, German, combined classifiers 1. Introduction methods, which have previously been shown to per- Word sense disambiguation (WSD) has been a very active form well for English WSD by Pedersen et al. (2005). area of research in computational linguistics. Most of the The remainder of this paper is structured as follows: Sec- work has focused on English.
    [Show full text]
  • Using Wiktionary for Computing Semantic Relatedness
    Proceedings of the Twenty-Third AAAI Conference on Artificial Intelligence (2008) Using Wiktionary for Computing Semantic Relatedness Torsten Zesch and Christof Muller¨ and Iryna Gurevych Ubiquitous Knowledge Processing Lab Computer Science Department Technische Universitat¨ Darmstadt, Hochschulstraße 10 D-64289 Darmstadt, Germany {zesch,mueller,gurevych} (at) tk.informatik.tu-darmstadt.de Abstract approach (Gabrilovich and Markovitch 2007). We gener- alize this approach by using concept representations like We introduce Wiktionary as an emerging lexical semantic re- source that can be used as a substitute for expert-made re- Wiktionary pseudo glosses, the first paragraph of Wikipedia sources in AI applications. We evaluate Wiktionary on the articles, English WordNet glosses, and GermaNet pseudo pervasive task of computing semantic relatedness for English glosses. Additionally, we study the effect of using shorter and German by means of correlation with human rankings but more precise textual representations by considering only and solving word choice problems. For the first time, we ap- the first paragraph of a Wikipedia article instead of the full ply a concept vector based measure to a set of different con- article text. cept representations like Wiktionary pseudo glosses, the first We compare the performance of Wiktionary with expert- paragraph of Wikipedia articles, English WordNet glosses, made wordnets, like Princeton WordNet and GermaNet and GermaNet pseudo glosses. We show that: (i) Wiktionary (Kunze 2004), and with Wikipedia as another collabora- is the best lexical semantic resource in the ranking task and tively constructed resource. In order to study the effects performs comparably to other resources in the word choice task, and (ii) the concept vector based approach yields the of the coverage of lexical semantic resources, we conduct best results on all datasets in both evaluations.
    [Show full text]