This paper examines the case for a graded notion of word meaning in two experiments, one which uses WordNet senses in a graded fashion, contrasted with the “winner takes all” annotation, and one which asks annotators to judge the similarity of two usages.

In the WSsim experiment, annotators rated the applicability of each WordNet 3.0 sense for a given target word occurrence.

In WordNet , they have 5, 7, and 4 senses, respectively.

WSsim is a word sense annotation task using WordNet senses.5 Unlike previous word sense annotation projects, we asked annotators to provide judgments on the applicability of every WordNet sense of the target lemma with the instruction: 6

3The SemCor dataset was produced alongside WordNet, so it can be expected to support the WordNet sense distinctions.

5WordNet 1.7.1 was used in the annotation of both SE-3 and SemCor; we used the more current WordNet 3.0 after verifying that the lemmas included in this experiment had the same senses listed in both versions.

In the first one, referred to as WSsim (Word Sense Similarity), annotators give graded ratings on the applicability of WordNet senses.

The first study additionally tests to what extent the judgments on WordNet senses fall into clearcut clusters, while the second study allows us to explore meaning similarity independently of any lexicon resource.

Reported inter-annotator agreement (ITA) for fine-grained word sense assignment tasks has ranged between 69% (Kilgarriff and Rosenzweig, 2000) for a lexical sample using the HECTOR dictionary and 78.6% using WordNet (Landes et al., 1998) in all-words annotation.

Although we use WordNet for the annotation, our study is not a study of WordNet per se.

We choose WordNet because it is sufficiently fine- grained to examine subtle differences in usage, and because traditionally annotated datasets exist to which we can compare our results.

Our rule-base yields comparable performance to WordNet while providing largely complementary information.

N0 Expansion 0.19 0.54 0.28 WikiBL 0.19 0.53 0.28 Sn0w400K 0.19 0.54 0.28 Lin 0.25 0.39 0.30 WordNet 0.30 0.47 0.37 Extraction Methods from Wikipedia:

WordNet + WikiAuJules+mce 0.35 0.47 0.40

Ponzetto and Strube (2007) identified the subsumption (ISA) relation from Wikipedia’s category tags, while in Yago (Suchanek et al., 2007) these tags, redirect links and WordNet were used to identify instances of 14 predefined specific semantic relations.

classifiers, whose training examples are derived automatically from WordNet .

They use these classifiers to suggest extensions to the WordNet hierarchy, the largest one consisting of 400K new links.

A prominent available resource is WordNet (Fellbaum, 1998), from which classical relations such as synonyms, hyponyms and some cases of meronyms may be used as LR rules.

An extension to WordNet was presented by (Snow et al., 2006).

The rule base utility was evaluated within two lexical expansion applications, yielding better results than other automatically constructed baselines and comparable results to WordNet .

This is in contrast to (Snow et al., 2005) which focused only on hyponymy and synonymy relations and could therefore extract positive and negative examples from WordNet .

The gold standards used in the evaluation are hypernym taxonomies extracted from WordNet and GDP (Open Directory Project), and meronym taxonomies extracted from WordNet .

In WordNet taxonomy extraction, we only use the word senses within a particular taxonomy to ensure no ambiguity.

In total, there are 100 hypernym taxonomies, 50 each extracted from WordNet3 and ODP4, and 50 meronym taxonomies from WordNetS .

It has been receiving increasing afienfion.because senuuunztaxononues,such as WordNet (Fellbaum, 1998), play an important role in solving knowledge-rich problems, including question answering (Harabagiu et a1., 2003) and textual entailment (Geffet and Dagan, 2005).

Pattern quality control is also investigated by using WordNet (Girju et al., 2006), graph structures built among terms (Widdows and Dorow, 2002; Kozareva et al., 2008), and pattern clusters (Davidov and Rappoport, 2008).

In order to obtain parallel training data for the translation models, we collected three different datasets: manually-tagged question reformulations and question-answer pairs from the WikiAnswers social Q&A site (Section 3.1), and glosses from WordNet , Wiktionary, Wikipedia and Simple Wikipedia (Section 3.2).

0 Wordnet (sense I): the natural satellite of the Earth.

o WordNet (Fellbaum, 1998).

Murdock and Croft (2005) created a first parallel corpus of synonym pairs extracted from WordNet , and an additional parallel corpus of English words translating to the same Arabic term in a parallel English-Arabic corpus.

Knowledge-based measures rely on lexical semantic resources such as WordNet and comprise path length based measures (Rada et al., 1989) and concept vector based measures (Qiu and Frei, 1993).

The method consists in representing words as a concept vector, where concepts correspond to WordNet synsets, Wikipedia article titles or Wiktionary entry names.

glosses in WordNet , the full article or the first paragraph of the article in Wikipedia or the full contents of a Wiktionary entry.

WordNet .26 .46 Wikipedia .27 .03 WikipediaFirst .30 .38 Wiktionary .39 .58 Translation probabilities

(2007) compute the semantic similarity using WordNet .

Its definition is much larger than that of WordNet .

We ap—p1y standard DCL in Run 2, where concepts are determined according to their definitions in WordNet (Ye et a1., 2007).

When we merge diverse words having similar semantic according to WordNet concepts , we obtain 873 concepts per article on average in Run 2.

Even though the reduction of total concepts is limited, these new wiki concepts will group the terms that cannot be detected by WordNet .

Its results are better than other external resources such as WordNet , Gazetteers and Google’s define operator, especially for definition QA (Lita et al., 2004).

Most of these concepts are multi—word terms, whereas WordNet has only 50,000 plus multi—word terms.

Any term could appear in the definition of a concept if necessary, while the total vocabulary existing in WordNet’s glossary definition is less than 2000.

html#wn) to WordNet (Miller, 1995) for lemmatization information.

We also tried ablating the WordNet relations, and observed that the “identical-word” feature hurt the model the most.

10This is accomplished by eliminating lines 12 and 13 from the definition of pm and redefining pword to be the unigram word distribution estimated from the Gigaword corpus, as in G0, without the help of WordNet .

Because dependency syntax is still only a crude approximation to semantic structure, we augment the model with a lexical semantics component, based on WordNet (Miller, 1995), that models how words are probabilistically altered in generating a paraphrase.

WordNet relation(s) The model next chooses a lexical semantics relation between 3360-) and the yet-to-be-chosen word ti (line 12).

(2007),6 we employ a 14-feature log-linear model over all logically possible combinations of the 14 WordNet relations (Miller, 1995).7 Similarly to Eq.

14, we normalize this log-linear model based on the set of relations that are nonempty in WordNet for the word 3360-).

Snow (Snow et al., 2006) has extended the WordNet 2.1 by adding thousands of entries (synsets) at a relatively high precision.

They have made several versions of extended WordNet available4.

For the experimental comparison, we focused on leaf semantic classes from the extended WordNet that have many hypernyms, so that a meaningful comparison could be made: specifically, we selected nouns that have at least three hypernyms, such that the hypernyms are the leaf nodes in the hypernym hierarchy of WordNet .

We also compare ASIA on twelve additional benchmarks to the extended Wordnet 2.1 produced by Snow et al (Snow et al., 2006), and show that for these twelve sets, ASIA produces more than five times as many set instances with much higher precision (98% versus 70%).

We examine a number of variations of the method, including the addition of WordNet , partial matching, or removing relation labels from the dependencies.

This makes sense if we take into consideration that WordNet lists all possible synonyms for all possible senses of a word, and so, given a great number of cross-sentence comparisons in multi-sentence summaries, there is an increased risk of spurious matches between words which, despite being potentially synonymous in certain contexts, are not equivalent in the text.

Three such variants are the baseline DEPEVAL(summ), the WordNet version DEPEVAL(summ) wn, and the version with removed relation labels DEPEVAL(summ) norel.

The new implementation of BE presented at the TAC 2008 workshop (Tratz and Hovy, 2008) introduces transformations for dependencies in order to increase the number of matches among elements that are semantically similar yet differ in terms of syntactic structure and/or lexical choices, and adds WordNet for synonym matching.

Since our method, presented in this paper, also uses the reranking parser, as well as WordNet , it would be interesting to compare both methods directly in terms of the performance of the dependency extraction procedure.

The correlations are listed for the following versions of our method: pm - partial matching for dependencies; wn - WordNet ; pred - matching predicate-only dependencies; norel - ignoring dependency relation label; one - counting a match only once irrespective of how many instances of

This paper presents a set of Bayesian methods for automatically extending the WORDNET ontology with new concepts and annotating existing concepts with generic property fields, or attributes.

We base our approach on Latent Dirichlet Allocation and evaluate along two dimensions: (l) the precision of the ranked lists of attributes, and (2) the quality of the attribute assignments to WORDNET concepts.

This paper introduced a set of methods based on Latent Dirichlet Allocation (LDA) for jointly extending the WORDNET ontology and annotating its concepts with attributes (see Figure 4 for the end result).

We present a Bayesian approach for simultaneously extending IsA hierarchies such as those found in WORDNET (WN) (Fellbaum, 1998) with additional concepts, and annotating the resulting concept graph with attributes, i.e., generic property fields shared by instances of that concept.

We use WORDNET 3.0 as the specific test ontology for our annotation procedure, and evalu-

A large body of previous work exists on extending WORDNET with additional concepts and instances (Snow et al., 2006; Suchanek et al., 2007); these methods do not address attributes directly.

The evaluation set is derived from WordNet in a semi-supervised way.

We proposed a semi-supervised way to extract non-compositional MWEs from WordNet .

Thirdly, we propose a semi-supervised approach for extracting non-compositional MWEs from WordNet , to decrease annotation cost.

Given a MWE, a set of queries is created: All synonyms of the M WE extracted from WordNet are collectedl.

For each of the 52, 217 MWEs of WordNet 3.0 (Miller, 1995) we collected:

We have empirically shown how automatically generated selectional preferences, using WordNet and distributional similarity measures, are able to effectively generalize lexical features and, thus, improve classification performance in a large-scale argument classification task on the CoNLL-2005 dataset.

This happens more often with WordNet based models, which have a lower word coverage compared to distributional similarity—based models.

Resnik (1993) proposed to model selectional preferences using semantic classes from WordNet in order to tackle ambiguity issues in syntax (noun-compounds, coordination, PP-attachment).

The performance loss on recall can be explained by the worse lexical coverage of WordNet when compared to automatically generated thesauri.

Examples of words missing in WordNet include abbreviations (e.g., Inc., Corp.) and brand names (e.g., Texaco, Sony).

Such methods include, but by no means limited to, semantic similarities between word pairs using lexical resources such as WordNet (Miller, 1995) and data-driven methods with various topic-dependent term weighting schemes on labeled corpus with topics such as MPQA.

Also, the distance between words in the local context or in the thesaurus-like dictionaries such as WordNet may be approximated as such measure.

It consists of WordNet synsets, where each synset is assigned three probability scores that add up to 1: positive, negative, and objective.

These scores are assigned at sense level (synsets in WordNet ), and we use the following equations to assess the sentiment scores at the word level.

The NER module is based on a combination of user defined rules based on Lesk word disambiguation (Lesk, 1988), WordNet (Miller, 1995) lookups, and many user-defined dictionary lookups, e.g.

—When words don’t match we attempt matching synonyms in WordNet for most common senses.

—Verb match statistics using WordNet’s cause and entailment relations.

The first is that the lack of alignments for two function words is unproblematic; the second is that the alignment between fact and reality, which is established on the basis of WordNet similarity, is indeed licensed in the current context.

Banerjee and Lavie (2005) and Chan and Ng (2008) use WordNet , and Zhou et a1.

The computation of these scores make extensive use of about ten lexical similarity resources, including WordNet , InfoMap, and Dekang Lin’s thesaurus.

In the first phase, it extracts all synonyms from a thesaurus, such as WordNet , for the words to be substituted.

They paraphrase a sentence s by replacing its words with WordNet synonyms, so that s can be more similar in wording to another sentence s’.

has also been proposed in (Zhou et al., 2006), which uses paraphrase phrases like our PT—l instead of WordNet synonyms.