Furthermore, the learned classes are not directly linked to existing resources such as WordNet (Fellbaum, 1998) or Wikipedia.

0 We propose SPred, a novel approach which harvests predicates from Wikipedia and generalizes them by leveraging core concepts from WordNet .

As explained below, we assume the set C to be made up of representative synsets from WordNet .

We perform this in two substeps: we first link all our disambiguated arguments to WordNet (Section 3.3.1) and then leverage the WordNet taxonomy to populate the semantic classes in 0 (Section 3.3.2).

3.3.1 Linking to WordNet

All these systems incorporated lexical semantics features derived from WordNet and named entity features.

Features used in the experiments can be categorized into six types: identical word matching (I), lemma matching (L), WordNet (WN), enhanced Lexical Semantics (LS), Named Entity matching (NE) and Answer type checking (Ans).

Arguably the most common source of word relations, WordNet (WN) provides the primitive features of whether two words could belong to the same synset in WordNet , could be antonyms and whether one is a hypernym of the other.

Although sets of synonyms can be easily found in thesauri or WordNet synsets, such resources typically cover only strict synonyms.

Traditionally, WordNet taxonomy is the linguistic resource for identifying hypernyms and hy-ponyms, applied broadly to many NLP problems.

However, WordNet has a number of well-known limitations including its rather limited or skewed concept distribution and the lack of the coverage of the IsA relation (Song et al., 2011).

Although lexical semantic information derived from WordNet has been used in some of these approaches, the research has mainly focused on modeling the mapping between the syntactic structures of questions and sentences, produced from syntactic analysis.

Structured resources such as WordNet offer a convenient hierarchical means for converging on a common ground for comparison, but offer little support for the divergent thinking that is needed to creatively view one concept as another.

These lateral views complement the vertical views of WordNet , and support a system for idea exploration called Thesaurus Rex.

We show also how Thesaurus Rex supports a novel, generative similarity measure for WordNet .

WordNet’s taxonomic organization of noun-senses and verb-senses — in which very general categories are successively divided into increasingly informative subcategories or instance-level ideas — allows us to gauge the overlap in information content, and thus of meaning, of two lexical concepts.

Wu & Palmer (1994) use the depth of a lexical concept in the WordNet hierarchy as such a proxy, and thereby estimate the similarity of two lexical concepts as twice the depth of their LCS divided by the sum of their individual depths.

Rather, when using Resnick’s metric (or that of Lin, or Jiang and Conrath) for measuring the similarity of lexical concepts in WordNet, one can use the category structure of WordNet itself to estimate information content.

This reliance on the consensus viewpoint explains why WordNet (Fellbaum, 1998) has proven so useful as a basis for computational measures of lexico-semantic similarity

Using WordNet , for instance, a similarity measure can vertically converge on a common superordinate category of both inputs, and generate a single numeric result based on their distance to, and the information content of, this common generalization.

Though WordNet is ideally structured to support vertical, convergent reasoning, its comprehensive nature means it can also be used as a solid foundation for building a more lateral and divergent model of similarity.

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).

Expensive feature engineering based on WordNet senses has been shown to be useful for document level sentiment classification.

WordNet is a byproduct of such an analysis.

In WordNet , paradigms are manually generated based on the principles of lexical and semantic relationship among words (Fellbaum, 1998).

WordNets are primarily used to address the problem of word sense disambiguation.

As our sense inventory, we use WordNet 3.0 (Fellbaum, 1998).

The WordNet ontology provides a rich network structure of semantic relatedness, connecting senses directly with their hypemyms, and providing information on semantically similar senses by virtue of their nearby locality in the network.

To extend beyond a single sense, the random walk may be initialized and restarted from a set of senses (seed nodes), rather than just one; this multi-seed walk produces a multinomial distribution over all the senses in WordNet with higher probability assigned to senses that are frequently visited from the seeds.

Additionally, because the texts often contain named entities which are not present in WordNet , we incorporated the similarity values produced by four string-based measures, which were used by other teams in the STS task: (1) longest common substring which takes into account the length of the longest overlapping contiguous sequence of characters (substring) across two strings (Gusfield, 1997), (2) longest common subsequence which, instead, finds the longest overlapping subsequence of two strings (Allison and Dix, 1986), (3) Greedy String Tiling which allows reordering in strings (Wise, 1993), and (4) the character/word n-gram similarity proposed by Barron-Cedefio et al.

o Explicit Semantic Analysis (Gabrilovich and Markovitch, 2007) where the high-dimensional vectors are obtained on WordNet , Wikipedia and Wiktionary.

WordNet (Fellbaum, 1998), by far the most widely used resource, specifies relations such as hyponymy, derivation, and entailment that can be used for semantic inference (Budanitsky and Hirst, 2006).

WordNet has also been exploited to automatically generate a training set for a hyponym classifier (Snow et al., 2005), and we make a similar use of WordNet in Section 5.1.

Recently, Szpektor and Dagan (2009) presented the resource Argument-mapped WordNet, providing entailment relations for predicates in WordNet .

The first step is preprocessing: We use a large corpus and WordNet to train an entailment classifier that estimates the likelihood that one propositional template entails another.

We describe a procedure for learning an entailment classifier, given a corpus and a lexicographic resource ( WordNet ).

Last, we use WordNet to automatically generate a training set and train a classifier.

They align senses in WordNet to Wikipedia entries in a supervised setting using semantic similarity measures.

The PPR measure (Agirre and Soroa, 2009) maps the glosses of the two senses to a semantic vector space spanned up by WordNet synsets and then compares them using the chi-square measure.

The semantic vectors ppr are computed using the personalized PageRank algorithm on the WordNet graph.

(2008) map FrameNet frames to WordNet synsets based on the embedding of FrameNet lemmas in WordNet .

They use Multi-WordNet, an English-Italian wordnet , to induce an Italian FrameNet lexicon with 15,000 entries.

To create MapNet, Tonelli and Pianta (2009) align FrameNet senses with WordNet synsets by exploiting the textual similarity of their glosses.

In recent years, the increasing availability of large-scale, rich semantic knowledge sources (such as Wikipedia and WordNet ) creates new opportunities to enhance the named entity disambiguation by developing algorithms which can exploit these knowledge sources at best.

Fortunately, in recent years, due to the evolution of Web (e.g., the Web 2.0 and the Semantic Web) and many research efforts for the construction of knowledge bases, there is an increasing availability of large-scale knowledge sources, such as Wikipedia and WordNet .

The key point of our method is a reliable semantic relatedness measure between concepts (including WordNet concepts, NEs and Wikipedia concepts), called Structural Semantic Relatedness, which can capture both the explicit semantic relations between concepts and the implicit semantic knowledge embedded in graphs and networks.

We extract three types of semantic relations (semantic relatedness between Wikipedia concepts, lexical relatedness between WordNet concepts and social relatedness between NEs) correspondingly from three knowledge sources: Wikipedia, WordNet and NE Co-occurrence Corpus.

WordNet 3.02 (Fellbaum et al., 1998), a lexical knowledge source includes over 110,000 WordNet concepts (word senses about English words).

Various lexical relations are recorded between WordNet concepts, such as hyponyms, holonym and synonym.

(2004) construct a network based on WordNet synonyms and then use the shortest paths between any given word and the words ’good’ and ’bad’ to determine word polarity.

’ good’ and ’bad’ themselves are closely related in WordNet with a 5-long sequence “good, sound, heavy, big, bad”.

Hu and Liu (2004) use WordNet synonyms and antonyms to predict the polarity of words.

One such important source is WordNet (Miller, 1995).

WordNet is a large lexical database of English.

The simplest approach is to connect words that occur in the same WordNet synset.

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 .

In addition, by virtue of its WordNet and FrameNet annotations, MASC will be linked to parallel WordNets and FrameNets in languages other than English, thus creating a global resource for multilingual technologies, including machine translation.

The MASC project is itself producing annotations for portions of the corpus for WordNet senses and FrameNet frames and frame elements.

4This includes WordNet sense annotations, which are not listed in Table 2 because they are not applied to full texts; see Section 3.1 for a description of the WordNet sense annotations in MASC.

3.1 WordNet Sense Annotations

The data were annotated with coarse-grained senses which were obtained by clustering senses from the WordNet 2.1 sense inventory based on the procedure proposed by Navigli (2006).

Previous approaches using topic models for sense disambiguation either embed topic features in a supervised model (Cai et al., 2007) or rely heavily on the structure of hierarchical lexicons such as WordNet (Boyd-Graber et al., 2007).

(2007) enhance the basic LDA algorithm by incorporating WordNet senses as an additional latent variable.

Instead of generating words directly from a topic, each topic is associated with a random walk through the WordNet hierarchy which generates the observed word.

iosyncracies in the hierarchical structure of WordNet can harm performance.

These paraphrases can be taken from an existing resource such as WordNet (Miller, 1995) or supplied by the user (see Section 4).

In Model I and Model 11, the sense paraphrases are obtained from WordNet , and both the context and the sense paraphrases are treated as documents, 0 2 dc and s 2 d8.

WordNet is a fairly rich resource which provides detailed information about word senses (glosses, example sentences, synsets, semantic relations between senses, etc.).

The resource is automatically constructed by means of a methodology that integrates lexicographic and encyclopedic knowledge from WordNet and Wikipedia.

Concepts and relations in BabelNet are harvested from the largest available semantic lexicon of English, WordNet , and a wide-coverage collaboratively edited encyclopedia, the English Wikipedia (Section 3.1).

We collect (a) from WordNet , all available word senses (as concepts) and all the semantic pointers between synsets (as relations); (b) from Wikipedia, all encyclopedic entries (i.e.

their concepts in common) by establishing a mapping between Wikipedia pages and WordNet senses (Section 3.2).

A pioneering endeavor was WordNet (Fellbaum, 1998), a computational lexicon of English based on psycholinguistic theories.

Wikipedia represents the perfect complement to WordNet , as it provides multilingual lexical knowledge of a mostly encyclopedic nature.

But while a great deal of work has been recently devoted to the automatic extraction of structured information from Wikipedia (Wu and Weld, 2007; Ponzetto and Strube, 2007; Suchanek et al., 2008; Medelyan et al., 2009, inter alia), the knowledge extracted is organized in a looser way than in a computational lexicon such as WordNet .

WordNet .

The most popular lexical knowledge resource in the field of NLP is certainly WordNet , a computational lexicon of the English language.

We create an open multilingual wordnet with large wordnets for over 26 languages and smaller ones for 57 languages.

It is made by combining wordnets with open li-cences, data from Wiktionary and the Unicode Common Locale Data Repository.

One of the many attractions of the semantic network WordNet (Fellbaum, 1998), is that there are numerous wordnets being built for different languages.

There are, in addition, many projects for groups of languages: Euro WordNet (Vossen, 1998), BalkaNet (Tufis et al., 2004), Asian Wordnet (Charoenporn et al., 2008) and more.

Although there are over 60 languages for which wordnets exist in some state of development (Fellbaum and Vossen, 2012, 316), less than half of these have released any data, and for those that have, the data is often not freely accessible (Bond and Paik, 2012).

It is a linked structure of wordnets of 18 different Indian languages, Universal Word dictionary and the Suggested Upper Merged Ontology (SUMO).

Past couple of decades have shown an immense growth in the development of lexical resources such as wordnet , Wikipedia, ontologies etc.

In this paper we present IndoNet, a lexical resource created by merging wordnets of 18 dif-

Suggested Upper Merged Ontology (SUMO) is the largest freely available ontology which is linked to the entire English WordNet (Niles and Pease, 2003).

Over the years wordnet has emerged as the most widely used lexical resource.

Though most of the wordnets are built by following the standards laid by English Wordnet (Fellbaum, 1998), their conceptualizations differ because of the differences in lexicalization of concepts across languages.

Wordnets are available in following Indian languages: Assamese, Bodo, Bengali, English, Gujarati, Hindi, Kashmiri, Konkani, Kannada, Malayalam, Ma-nipuri, Marathi, Nepali, Punjabi, Sanskrit, Tamil, Telugu and Urdu.

HAHAcronym is mainly based on lexical substitution via semantic field opposition, rhyme, rhythm and semantic relations such as antonyms retrieved from WordNet (Stark and Riesenfeld, 1998) for adjectives.

To further increase the size of the ingredient list, we utilize another resource called WordNet (Miller, 1995), which is a large lexical database for English.

In WordNet , nouns, verbs, adjectives and adverbs are grouped into sets of cognitive synonyms called synsets.

Each synset in WordNet expresses a different concept and they are connected to each other with lexical, semantic and conceptual relations.

A generic, lightweight resource like WordNet can provide these relations, or a richer ontology can be used if one is available (e.g.

But ad-hoc categories do not replace natural kinds; rather, they supplement an existing system of more-or-less rigid categories, such as the categories found in WordNet .

member of the category named by C. AC can denote a fixed category in a resource like WordNet or even Wikipedia; thus, Afruit matches any member of {apple, orange, pear, lemon} and Aanimal any member of {dog, cat, mouse, deer, fox}.

and @ as category builders to a handcrafted gold standard like WordNet .

Other researchers have likewise used WordNet as a gold standard for categorization experiments, and we replicate here the experimental setup of Almuhareb and Poesio (2004, 2005), which is designed to measure the effectiveness of web-acquired conceptual descriptions.

Almuhareb and Poesio choose 214 English nouns from 13 of WordNet’s upper-level semantic categories, and proceed to harvest property values for these concepts from the web using the Hearst-like pattern “alanlthe * C islwas”.

Techniques vary, from the use of stemmers and morphological analysis to the use of thesauri (such as WordNet ; see Fellbaum, 1998; Voorhees, 1998) to pad a query with synonyms, to the use of statistical analysis to identify more appropriate context-sensitive associations and near-synonyms (e.g.

Hearst (1992) shows how a pattern like “Xs and other Ys” can be used to construct more fluid, context-specific taxonomies than those provided by WordNet (e.g.

Our system combines multiple preexisting sources of similarity data (a standard thesaurus, WordNet , contextual similarity), enabling it to capture many types of similarity groups (“synonyms of crash,” “types of car,” etc.).

We consider three similarity data sources: the Moby thesaurus1 , WordNet (Fellbaum, 1998), and distributional similarity based on a large corpus of text (Lin, 1998).

Unlike some other thesauri (such as WordNet and thesaurus.com), entries are not broken down by word sense.

Unlike some other thesauri (including WordNet and thesaurus.com), the entries are not broken down by word sense or part of speech.

degree of Wordnet polysemy for polysemous words

Table 4: Average degree of Wordnet polysemy per category in the 2 domains for Hindi

degree of Wordnet polysemy for polysemous words

This is achieved with the help of a novel synset-aligned multilingual dictionary which facilitates the projection of parameters learned from the Wordnet and annotated corpus of L1 to L2.

Wordnet-dependent parameters depend on the structure of the Wordnet whereas the Corpus-dependent parameters depend on various statistics learned from a sense marked corpora.

Both the tasks of (a) constructing a Wordnet from scratch and (b) collecting sense marked corpora for multiple languages are tedious and expensive.

(2009) observed that by projecting relations from the Wordnet of a language and by projecting corpus statistics from the sense marked corpora of the language to those of the target language, the efi‘ort required in constructing semantic graphs for multiple Wordnets and collecting sense marked corpora for multiple languages can be avoided or reduced.

They showed that it is possible to project the parameters learned from the annotation work of one language to another language provided aligned Wordnets for the two languages are available.

However, they do not address situations where two resource deprived languages have aligned Wordnets but neither has sufficient annotated data.

The selectional preferences are expressed in terms of semantic concepts from WordNet , rather than a set of individual words.

4. selecting a WordNet concept r for each argument slot, e.g.

and Light (1999) and turn WordNet into a Hidden Markov model (HMM).

RRR consists of 20,081 training and 3,097 test quadruples of the form (v, n1 , p, n2) , where the attachment decision is either v or n l. The best published results over RRR are those of Stetina and Nagao (1997), who employ WordNet sense predictions from an unsupervised WSD method within a decision tree classifier.

(2005) experimented with first-sense and hypernym features from HowNet and CiLin (both WordNets for Chinese) in a generative parse model applied to the Chinese Penn Treebank.

We experimented with a range of semantic representations, all of which are based on WordNet 2.1.

As mentioned above, words in WordNet are organised into sets of synonyms, called synsets.

Note that these are the two extremes of semantic granularity in WordNet , and we plan to experiment with intermediate representation levels in future research (c.f.

Our choice for this work was the WordNet 2.1 lexical database, in which synonyms are grouped into synsets, which are then linked via an ISA hierarchy.

WordNet contains other types of relations such as meronymy, but we did not use them in this research.

mallet, square and steel-wool pad are also descendants of TOOL in WordNet , none of which would conventionally be used as the manner adjunct of cat).

We explore several models for semantic representation, based around WordNet (Fellbaum, 1998).

Our NR classification evaluation strictly follows the ACL SemEval-07 Task 4 datasets and protocol, obtaining an f-score of 70.6, as opposed to 64.8 of the best previous work that did not use the manually provided WordNet sense disambiguation tags.

Nouns in this pair were manually labeled with their corresponding WordNet 3 labels and the web queries used to

The 15 submitted systems were assigned into 4 categories according to whether they use the WordNet and Query tags (some systems were assigned to more than a single category, since they reported experiments in several settings).

In our evaluation we do not utilize WordNet or Query tags, hence we compare ourselves with the corresponding group (A), containing 6 systems.

To improve results, some systems utilize additional manually constructed semantic resources such as WordNet (WN) (Beamer et al., 2007).

Furthermore, usage of such resources frequently requires disambiguation and connection of the data to the resource (word sense disambiguation in the case of WordNet ).

We evaluated our algorithm on SemEval-07 Task 4 data, showing superior results over participating algorithms that did not utilize WordNet disambiguation tags.

Many relation classification algorithms utilize WordNet .

Among the 15 systems presented by the 14 SemEval teams, some utilized the manually provided WordNet tags for the dataset pairs (e.g., (Beamer et al., 2007)).

Method P R F Acc Unsupervised clustering (4.3.3) 64.5 61.3 62.0 64.5 Cluster Labeling (4.3.1) 65.1 69.0 67.2 68.5 HITS Features (4.3.2) 69.1 70.6 70.6 70.1 Best Task 4 (no WordNet) 66.1 66.7 64.8 66.0 Best Task 4 (with WordNet ) 79.7 69.8 72.4 76.3

Table 1 shows our results, along with the best Task 4 result not using WordNet labels (Costello, 2007).

Our approach leverages a similarity measure that enables the structural comparison of senses across lexical resources, achieving state-of-the-art performance on the task of aligning WordNet to three different collaborative resources: Wikipedia, Wiktionary and OmegaWiki.

To enable a comparison with the state of the art, we followed Matuschek and Gurevych (2013) and performed an alignment of WordNet synsets (WN) to three different collaboratively-constructed resources: Wikipedia

As mentioned in Section 2.1.1, we build the WN graph by including all the synsets and semantic relations defined in WordNet (e.g., hypernymy and meronymy) and further populate the relation set by connecting a synset to all the other synsets that appear in its disambiguated gloss.

Notable examples are WordNet , Wikipedia and, more recently, collaboratively-curated resources such as OmegaWiki and Wiktionary (Hovy et al., 2013).

When a lexical resource can be viewed as a semantic graph, as with WordNet or Wikipedia, this limit can be overcome by means of alignment algorithms that exploit the network structure to determine the similarity of concept pairs.

We report state-of-the-art performance when aligning WordNet to Wikipedia, OmegaWiki and Wiktionary.

A good example is WordNet , which has been exploited as a semantic network in dozens of NLP tasks (Fellbaum, 1998).

For instance, WordNet can be readily represented as an undirected graph G whose nodes are synsets and edges are modeled after the relations between synsets defined in WordNet (e. g., hypernymy, meronymy, etc.

Any semantic network with a dense relational structure, providing good coverage of the words appearing in the definitions, is a suitable candidate for H. For this purpose we used the WordNet (Fellbaum, 1998) graph which was further enriched by connecting

As an example, assume we are given two semantic signatures computed for two concepts in WordNet and Wiktionary.

We conduct a thorough evaluation of the proposed methodology both manually as well as through comparison with WordNet .

Remarkably, in 44% cases the birth of a novel sense is attested by WordNet, while in 46% cases and 43% cases split and join are respectively confirmed by WordNet .

6.2 Automated evaluation with WordNet

We chose WordNet for automated evaluation because not only does it have a wide coverage of word senses but also it is being maintained and updated regularly to incorporate new senses.

For our evaluation, we developed an aligner to align the word clusters obtained with WordNet senses.

Remarkably, comparison with the English WordNet indicates that in 44% cases, as identified by our algorithm, there has been a birth of a completely novel sense, in 46% cases a new sense has split off from an older sense and in 43% cases two or more older senses have merged in to form a new sense.

A few approaches suggested by (Bond et al., 2009; Paakko' and Linden, 2012) attempt to augment WordNet synsets primarily using methods of annotation.

Although handcrafted lexical resources, such as WordNet, could provide more reliable related terms, previous studies showed that query expansion using only WordNet leads to very limited performance improvement.

Intuitively, compared with co-occurrence-based thesauri, handcrafted thesauri, such as WordNet , could provide more reliable terms for query expansion.

However, previous studies failed to show any significant gain in retrieval performance when queries are expanded with terms selected from WordNet (Voorhees, 1994; Stairmand, 1997).

In this paper, we study several term similarity functions that exploit various information from two lexical resources, i.e., WordNet

Although the use of WordNet in query expansion has been studied by various researchers, the improvement of retrieval performance is often limited.

Voorhees (Voorhees, 1994) expanded queries using a combination of synonyms, hypemyms and hyponyms manually selected from WordNet , and achieved limited improvement (i.e., around —2% to

Stairmand (Stairmand, 1997) used WordNet for query expansion, but they concluded that the improvement was restricted by the coverage of the WordNet and no empirical results were reported.

We examine the differences in content between the 1911 and 1987 versions of Roget’s, and we test both versions with each other and WordNet on problems such as synonym identification and word relatedness.

We also present a novel method for measuring sentence relatedness that can be implemented in either version of Roget’s or in WordNet .

Although the 1987 version of the Thesaurus is better, we show that the 1911 version performs surprisingly well and that often the differences between the versions of R0-get’s and WordNet are not statistically significant.

We compare two versions, the 1987 and 1911 editions of the Thesaurus with each other and with WordNet 3.0.

Roget’s Thesaurus has a unique structure, quite different from WordNet , of which the NLP community has yet to take full advantage.

In this paper we demonstrate that although the 1911 version of the Thesaurus is very old, it can give results comparable to systems that use WordNet or newer versions of Roget’s Thesaurus.

WordNet ?

Table 1: Class labels found in WordNet in original form, or found in WordNet after removal of leading words, or not found in WordNet at all

Accuracy of Class Labels: Built over many years of manual construction efforts, lexical gold standards such as WordNet (Fellbaum, 1998) provide wide-coverage upper ontologies of the English language.

WordNet (Fellbaum, 2010) has been used in numerous natural language processing tasks such as word sense disambiguation and information extraction with considerable success.

Kurdish is a less-resourced language for which, among other resources, no wordnet has been built yet.

1. highlighted the main challenges in building a wordnet for the Kurdish language (including its inherent diversity and morphological complexity),

2. built the first prototype of KurdNet, the Kurdish WordNet (see a summary below), and

There are two well-known models for building wordnets for a language (Vossen, 1998):

Typically, word frequency distributions are estimated with respect to a sense-tagged corpus such as SemCor (Miller et al., 1993), a 220,000 word corpus tagged with WordNet (Fellbaum, 1998) senses.

The distributional similarity scores of the nearest neighbours are associated with the respective target word senses using a WordNet similarity measure, such as those proposed by J iang and Conrath (1997) and Banerjee and Pedersen (2002).

(2004b) to remove low-frequency senses from WordNet , we focus on finding senses that are unattested in the corpus on the premise that, given accurate disambiguation, rare senses in a corpus contribute to correct interpretation.

For the purposes of this research, the choice of Macmillan is significant in that it is a conventional dictionary with sense definitions and examples, but no linking between senses.11 In terms of the original research which gave rise to the sense-tagged dataset, Macmillan was chosen over WordNet for reasons including: (l) the well-documented difficulties of sense tagging with fine-grained WordNet senses (Palmer et al., 2004; Navigli et al., 2007); (2) the regular update cycle of Macmillan (meaning it contains many recently-emerged senses); and (3) the finding in a preliminary sense-tagging task that it better captured Twitter usages than WordNet (and also OntoNotes: Hovy et al.

The average sense ambiguity of the 20 target nouns in Macmillan is 5.6 (but 12.3 in WordNet ).

We first notice that, despite the coarser-grained senses of Macmillan as compared to WordNet , the upper bound WSD accuracy using Macmillan is comparable to that of the WordNet-based datasets over the balanced BNC, and quite a bit lower than that of the two domain corpora of Koeling et al.

the WordNet hierarchy).

For each domain, annotators were asked to sense-annotate a random selection of sentences for each of 40 target nouns, based on WordNet v1.7.

For each dataset, we use HDP to induce topics for each target lemma, compute the similarity between the topics and the WordNet senses (Equation (1)), and rank the senses based on the prevalence scores (Equation (2)).

It is important to bear in mind that MKWC in these experiments makes use of full-text parsing in calculating the distributional similarity thesaurus, and the WordNet graph structure in calculating the similarity between associated words and different senses.

Almuhareb and Poesio (2004) used as their experimental basis a sampling of 214 English nouns from 13 of WordNet’s upper-level semantic categories, and proceeded to harvest adjectival features for these noun-concepts from the web using the textual pattern “[a | an | the] * C [is | was]”.

Veale and Hao (2007) use the Google API in conjunction with Princeton WordNet (Fellbaum, 1998) as the basis of their harvesting system.

They first extracted a list of antonymous adjectives, such as “hot” or “cold”, from WordNet , the intuition being that explicit similes will tend to exploit properties that occupy an exemplary point on a scale.

To harvest a comparable body of Chinese similes from the web, we also use the Google API, in conjunction with both WordNet and HowNet (Dong and Dong, 2006).

(1999), in which each of the textual glosses in WordNet (Fellbaum, 1998) is linguistically analyzed to yield a sense-tagged logical form, is an example of the former approach.

Almuhareb and Poesio go on to demonstrate that the values and attributes that are found for word-concepts on the web yield a sufficiently rich representation for these word-concepts to be automatically clustered into a form resembling that assigned by WordNet (see Fellbaum, 1998).

In the case of English similes, Veale and Hao (2007) describe how two English similes “as A as N1” and “as A as N2” will be mutually disambiguating if N1 and N2 are synonyms in WordNet, or if some sense of N1 is a hypernym or hyponym of some sense of N2 in WordNet .

For instance, though HowNet has a much shallower hierarchical organization than WordNet , it compensates by encapsulating the meaning of different word senses using simple logical formulae of semantic primitives, or sememes, that are derived from the meaning of common Chinese characters.

WordNet and HowNet thus offer two complementary levels or granularities of generalization that can be exploited as the context demands.

Manually constructing a Wordnet is a difficult task, needing years of experts’ time.

As a first step to automatically construct full Wordnets, we propose approaches to generate Wordnet synsets for languages both resource-rich and resource-poor, using publicly available Wordnets , a machine translator and/or a single bilingual dictionary.

Our algorithms translate synsets of existing Wordnets to a target language T, then apply a ranking method on the translation candidates to find best translations in T. Our approaches are applicable to any language which has at least one existing bilingual dictionary translating from English to it.

Wordnets are intricate and substantive repositories of lexical knowledge and have become important resources for computational processing of natural languages and for information retrieval.

Good quality Wordnets are available only for a few "resource-rich" languages such as English and Japanese.

Published approaches to automatically build new Wordnets are manual or semiautomatic and can be used only for languages that already possess some lexical resources.

The key aspect of our method is that it is the first unified approach that assigns the polarity of both word- and sense-level connotations, exploiting the innate bipartite graph structure encoded in WordNet .

We have introduced a novel formulation of lexicon induction operating over both words and senses, by exploiting the innate structure between the words and senses as encoded in WordNet .

7Because senses in WordNet can be tricky to understand, care should be taken in designing the task so that the Turkers will focus only on the corresponding sense of a word.

Therefore, we provided the part of speech tag, the WordNet gloss of the selected sense, and a few examples as given in WordNet .

We introduce ConnotationWordNet, a connotation lexicon over the network of words in conjunction with senses, as defined in WordNet .

For example, consider “abound”, for which lexicographers of WordNet prescribe two different senses:

Especially if we look up the WordNet entry for “bristle”, there are noticeably more negatively connotative words involved in its gloss and examples.

Another benefit of our approach is that for various WordNet relations (e.g., antonym relations), which are defined over synsets (not over words), we can add edges directly between corresponding synsets, rather than projecting (i.e., approximating) those relations over words.

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.

As regards recall, we note that in two cases (i.e., DBpedia returning page super-types from its upper taxonomy, YAGO linking categories to WordNet synsets) the generalizations are neither pages nor categories and that MENTA returns heterogeneous hypernyms as mixed sets of WordNet synsets, Wikipedia pages and categories.

MENTA seems to be the closest resource to ours, however, we remark that the hypernyms output by MENTA are very heterogeneous: 48% of answers are represented by a WordNet synset, 37% by Wikipedia categories and 15% are Wikipedia pages.

However, unlike the case with smaller manually-curated resources such as WordNet (Fellbaum, 1998), in many large automatically-created resources the taxonomical information is either missing, mixed across resources, e.g., linking Wikipedia categories to WordNet synsets as in YAGO, or coarse-grained, as in DBpedia whose hypernyms link to a small upper taxonomy.

(2005) provide a general vector-based method which, however, is incapable of linking pages which do not have a WordNet counterpart.

Higher coverage is provided by de Melo and Weikum (2010) thanks to the use of a set of effective heuristics, however, the approach also draws on WordNet and sense frequency information.

However, these methods do not link terms to existing knowledge resources such as WordNet , whereas those that explicitly link do so by adding new leaves to the existing taxonomy instead of acquiring wide-coverage taxonomies from scratch (Pan-tel and Ravichandran, 2004; Snow et al., 2006).

Other approaches, such as YAGO (Suchanek et al., 2008; Hoffart et al., 2013), yield a taxonomical backbone by linking Wikipedia categories to WordNet .

However, the categories are linked to the first, i.e., most frequent, sense of the category head in WordNet , involving only leaf categories in the linking.

This paper presents experiments with WordNet semantic classes to improve dependency parsing.

Base WordNet WordNet Clusters

WordNet .

(2011), based on WordNet 2.1.

Broadly speaking, we can classify the methods to incorporate semantic information into parsers in two: systems using static lexical semantic repositories, such as WordNet or similar ontologies (Agirre et al., 2008; Agirre et al., 2011; Fujita et al., 2010), and systems using dynamic semantic clusters automatically acquired from corpora (Koo et al., 2008; Suzuki et al., 2009).

0 Does semantic information in WordNet help

0 How does WordNet compare to automatically obtained information?

Broadly speaking, we can classify the attempts to add external knowledge to a parser in two sets: using large semantic repositories such as WordNet and approaches that use information automatically acquired from corpora.

The results showed a signi-cant improvement, giving the first results over both WordNet and the Penn Treebank (PTB) to show that semantics helps parsing.

(201 1) successfully introduced WordNet classes in a dependency parser, obtaining improvements on the full PTB using gold POS tags, trying different combinations of semantic classes.

To train the system, we extract substructures of WordNet and dis-criminatively learn to reproduce them, using adaptive subgradient stochastic optimization.

On the task of reproducing sub-hierarchies of WordNet , our approach achieves a 51% error reduction over a chance baseline, including a 15% error reduction due to the non-hypernym—factored sibling features.

We considered two distinct experimental setups, one that illustrates the general performance of our model by reproducing various medium-sized WordNet domains, and another that facilitates comparison to previous work by reproducing the much larger animal subtree provided by Kozareva and Hovy (2010).

General setup: In order to test the accuracy of structured prediction on medium-sized full-domain taxonomies, we extracted from WordNet 3.0 all bottomed-out full subtrees which had a tree-height of 3 (i.e., 4 nodes from root to leaf), and contained (10, 50] terms.11 This gives us 761 non-overlapping trees, which we partition into

To project WordNet synsets to terms, we used the first (most frequent) term in each synset.

However, currently available taxonomies such as WordNet are incomplete in coverage (Pennacchiotti and Pantel, 2006; Hovy et al., 2009), unavailable in many domains and languages, and

Figure 1: An excerpt of WordNet’s vertebrates taxonomy.

First, on the task of recreating fragments of WordNet , we achieve a 51% error reduction on ancestor-based F1 over a chance baseline, including a 15% error reduction due to the non-hypernym-factored sibling features.

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

Table 2 shows the results of the evaluation of our initial thesaurus, achieved by comparing the selected semantic neighbors with two complementary reference resources: WordNet 3.0 synonyms (Miller, 1990) [W], which characterize a semantic similarity based on paradigmatic relations, and the Moby thesaurus (Ward, 1996) [M], which gathers a larger set of types of relations and is more representative of semantic relatedness3.

WordNet provides a restricted number of synonyms for each noun while the Moby thesaurus contains for each entry a large number of synonyms and similar words.

As a consequence, the precisions at different cutoffs have a significantly higher value with Moby as reference than with WordNet as reference.

The second approach makes use of a less structured source of knowledge about words such as the definitions of classical dictionaries or the glosses of WordNet .

WordNet’s glosses were used to support Lesk-like measures in (Banerjee and Pedersen, 2003) and more recently, measures were also defined from Wikipedia or Wiktionaries (Gabrilovich and

where: infll and inflg are inflected variants of nounl and noung generated using the Java WordNet Libraryl; THAT is a complementizer and can be that, which, or who; and * stands for 0 or more (up to 8) instances of Google’s star operator.

Finally, we lemmatize the main verb using WordNet’s morphological analyzer Morphy (Fellbaum, 1998).

(2005) apply both classic (SVM and decision trees) and novel supervised models (semantic scattering and iterative semantic specialization), using WordNet , word sense disambiguation, and a set of linguistic features.

Their approach is highly resource intensive (uses WordNet , CoreLex and Moby’s thesaurus), and is quite sensitive to the seed set of verbs: on a collection of 453 examples and 19 relations, they achieved 52.6% accuracy with 84 seed verbs, but only 46.7% with 57 seed verbs.

We further experimented with the SemEval’07 task 4 dataset (Girju et al., 2007), where each example consists of a sentence, a target semantic relation, two nominals to be judged on whether they are in that relation, manually annotated WordNet senses, and the Web query used to obtain the sentence:

WordNet(el) = "vessel%l:06:OO::", WordNet(e2) = "tool%l:O6:OO::", Content—Container(e2, el) = "true", Query = "contents of the * were a"

The SemEval competition defines four types of systems, depending on whether the manually annotated WordNet senses and the Google query are used: A (WordNet=no, Query=no), B (WordNet=yes, Query=no), C (WordNet=no, Query=yes), and D (WordNet=yes, Query=yes).

(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.

Finally, we report on the quality of a large database of Wordnet-based preferences obtained after manually associating our topics with Wordnet classes (Section 4.4).

We then automatically filtered out any rules which contained a negation, or for which the antecedent and consequent contained a pair of antonyms found in WordNet (this left us with 85 rules).

Table 3: Top 10 and Bottom 10 ranked inference rules ranked by LDA-SPafter automatically filtering out negations and antonyms (using WordNet ).

Resnik (1996) presented the earliest work in this area, describing an information-theoretic approach that inferred selectional preferences based on the WordNet hypernym hierarchy.

This avoids problems like WordNet’s poor coverage of proper nouns and is shown to improve performance.

WordNet ), or automatically generated (Pantel, 2003).

Talking Points are a set of characteristics of concepts belonging to source and target domains and related facts about the world which the authors acquire automatically from WordNet and from the web.

They mine WordNet (Fellbaum, 1998) for the examples of systematic polysemy, which allows to capture metonymic and metaphorical relations.

The authors search for nodes that are relatively high up in the WordNet hierarchy and that share a set of common word forms among their descendants.

They use hyponymy relation in WordNet and word bigram counts to predict metaphors at a sentence level.

A number of metaphorical senses are included in WordNet , however without any accompanying semantic annotation.

5EuroWordNet is a multilingual database with wordnets for several European languages (Dutch, Italian, Spanish, German, French, Czech and Estonian).

The wordnets are structured in the same way as the Princeton WordNet for English.

WordNet (Lonneker and Eilts, 2004), would undoubtedly provide a new platform for experiments and enable researchers to directly compare their results.

mantic categories originating in WordNet .

7Supersenses are called “lexicographer classes” in WordNet documentation (Fellbaum, 1998), http: / /worolnet .

English adjectives do not, as yet, have a similar high-level semantic partitioning in WordNet , thus we use a 13-class taxonomy of adjective supersenses constructed by Tsvetkov et al.

WordNet lacks coarse-grained semantic categories for adjectives.

For example, the top-level classes in GermaNet include: adj.feeling (e.g., willing, pleasant, cheerful); adj.sabstance (e.g., dry, ripe, creamy); adj.spatial (e.g., adjacent, gigantic).12 For each adjective type in WordNet , they produce a vector with a classifier posterior probabilities corresponding to degrees of membership of this word in one of the 13 semantic classes,13 similar to the feature vectors we build for nouns and verbs.

Consider an example related to projection of WordNet supersenses.

(2013) describe a Concrete Category Overlap algorithm, where co-occurrence statistics and Turney’s abstractness scores are used to determine WordNet supersenses that correspond to literal usage of a given adjective or verb.

To implement this idea, they extend MRC imageability scores to all dictionary words using links among WordNet supersenses (mostly hypernym and hyponym relations).

Because they heavily rely on WordNet and availability of imageability scores, their approach may not be applicable to low-resource languages.

This study presents a novel approach to the problem of system portability across different domains: a sentiment annotation system that integrates a corpus-based classifier trained on a small set of annotated in-domain data and a lexicon-based system trained on WordNet .

In this paper, we present a novel approach to the problem of system portability across different domains by developing a sentiment annotation system that integrates a corpus-based classifier with a lexicon-based system trained on WordNet .

The information contained in lexicographical sources, such as WordNet , reflects a lay person’s general knowledge about the world, while domain-specific knowledge can be acquired through classifier training on a small set of in-domain data.

The final, third part of the paper presents our system, composed of an ensemble of two classifiers —one trained on WordNet glosses and synsets and the other trained on a small in-domain training set.

A lexicon-based approach capitalizes on the fact that dictionaries, such as WordNet (Fellbaum, 1998), contain a comprehensive and domain-independent set of sentiment clues that exist in general English.

One of the limitations of general lexicons and dictionaries, such as WordNet (Fellbaum, 1998), as training sets for sentiment tagging systems is that they contain only definitions of individual words and, hence, only unigrams could be effectively learned from dictionary entries.

Since the structure of WordNet glosses is fairly different from that of other types of corpora, we developed a system that used the list of human-annotated adjectives from (Hatzivassiloglou and McKeown, 1997) as a seed list and then learned additional unigrams

For verb classes, in particular, we devised a method for classifying all the verbs and verb phrases in WordNet into the activity and state classes.

The experimental results show that verb and verb phrase classification method is reasonably accurate with 91% precision and 78% recall with manually constructed gold standard consisting of 80 verbs and 82% accuracy for a random sample of all the WordNet entries.

We collected all hyponyms of words “do” and “act”, from WordNet (Fellbaum, 1998).

Lastly, we removed all the verbs that are under the hierarchy of “move” from WordNet .

In addition, the lexicon we constructed for the baseline (i.e., using the WordNet ) contains more errors than our activity lexicon for activity verbs.

We consider all the verbs and verb phrases in WordNet (Fellbaum, 1998) which is the largest electronic lexical database.

Based on the query matrix in table 2, we issued queries for all the verbs and verb phrases from WordNet to a search engine.

We finally trained our model with the top 10 features and classified all WordNet verbs and verb phrases.

Since our work is specifically geared toward domain-independent experience detection, we attempted to maximize the coverage by using all the verbs in WordNet , as opposed to the verbs appearing in a particular domain-specific corpus (e.g., medicine domain) as done in the previous work.

Frequently, such resources are constructed through automatic mergers of complementary resources, such as WordNet and Wikipedia.

Semantic knowledge bases such as WordNet (Fellbaum, 1998), YAGO (Suchanek et al., 2007), and BabelNet (Navigli and Ponzetto, 2010) provide ontological structure that enables a wide range of tasks, such as measuring semantic relatedness (Budanitsky and Hirst, 2006) and similarity (Pilehvar et al., 2013), paraphrasing (Kauchak and Barzilay, 2006), and word sense disambiguation (Navigli and Ponzetto, 2012; Moro et al., 2014).

extend WordNet using distributional or structural features to identify novel semantic connections between concepts.

The recent advent of large semistructured resources has enabled the creation of new semantic knowledge bases (Medelyan et al., 2009; Hovy et al., 2013) through automatically merging WordNet and Wikipedia (Suchanek et al., 2007; Navigli and Ponzetto, 2010; Nie-mann and Gurevych, 2011).

Rzeniewicz and Szymanski (2013) extend WordNet with commonsense knowledge using a 20 Questions-like game.

Knowledge base As the reference knowledge base, we chose BabelNet2 (Navigli and Ponzetto, 2010), a large-scale multilingual semantic ontology created by automatically merging WordNet with other collaboratively-constructed resources such as Wikipedia and OmegaWiki.

First, by connecting WordNet synsets to Wikipedia pages, most synsets are associated with a set of pictures; while often noisy, these pictures sometimes illustrate the target concept and are an ideal case for validation.

Second, BabelNet contains the semantic relations from both WordNet and hyperlinks in Wikipedia; these relations are again an ideal case of validation, as not all hyperlinks connect semantically-related pages in Wikipedia.

Many of these templates utilize information from WordNet (Fell-baum, 1998).

0 WordNet link types (link type list) (e.g., attribute, hypernym, entailment)

0 Lexicographer filenames (lexnames)—top level categories used in WordNet (e.g., noun.body, verb.cognition)

The features make extensive use of WordNet .

Since the PDEP system enables exploration of features from WordNet , FrameNet, and VerbNet, we are able to make some assessment of these resources.

WordNet played a statistically significant role in the systems developed by Tratz (2011) and Srikumar and Roth (2013).

This includes the WordNet lexicographer’s file name (e.g., noun.time), synsets, and hypernyms.

We are examining the WordNet detour to FrameNet, as described in Burchardt et al.

Section 4 describes how we are able to investigate the relationship of WordNet , FrameNet, and VerbNet to this effort and how this examination of preposition behavior can be used in working with these resources.

The feature extraction rules are (1) word class (we), (2) part of speech (pos), (3) lemma (1), (4) word (w), (5) WordNet lexical name (In), (6) WordNet synonyms (s), (7) WordNet hypernyms (h), (8) whether the word is capitalized (c), and (9) affixes (af).

For features such as the WordNet lexical name, synonyms and hypernyms, the number of values may be much larger.

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%).

(Pennacchiotti and Pantel, 2006) proposed an algorithm for automatically ontologizing semantic relations into WordNet .

However, despite its high precision entries, WordNet’s limited coverage makes it impossible for relations whose arguments are not present in WordNet to be incorporated.

They mapped each argument of the relation into WordNet and identified the senses for which the relation holds.

Unfortunately, despite its very high precision entries, WordNet is known to have limited coverage, which makes it impossible for algorithms to map the content of a relation whose arguments are not present in WordNet .

To surmount this limitation, we do not use WordNet , but employ a different method of obtaining superclasses of a filler term: the inverse doubly-anchored patterns DAP‘1 (Hovy et al., 2009), which, given two arguments, harvests its supertypes from the source corpus.

Since our problem definition differs from available related work, and WordNet does not contain all harvested arguments as shown in (Hovy et al., 2009), it is not possible to make a direct comparison.

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-).

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.

Precisions are simply omitted because the difference to the recalls are always the number of failures on referring to WordNet by mislabeling of lemmata or POSs, which is always the same for the three methods.

To calculate sense similarities, we used the WordNet similarity package by Pedersen et al.

Those texts were parsed using RASP parser (Briscoe et al., 2006) version 3.1, to obtain grammatical relations for the distributional similarity, as well as to obtain lemmata and part-of-speech (POS) tags which are required to look up the sense inventory of WordNet .

(2010) used a WordNet pre-pruning.

Disambiguation is performed by considering only those candidate synsets that belong to the top-k largest connected components of the WordNet on domain corpus.

The cluster centers are located at the means of hypotheses including miscellaneous alternatives not intended, thus the estimated probability distribution is, roughly speaking, offset toward the center of WordNet , which is not what we want.

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

These features are typically obtained from external resources such as Wordnet (Miller, 1990).

The first baseline searches for each head noun in WordNet and labels the noun as category Ck, if it has a hypernym synset corresponding to that category.

We manually identified the WordNet synsets that, to the best of our ability, seem to most closely correspond

from dependency parse tree) along with computing similarity in semantic spaces (using WordNet ) clearly produces an improvement in the summarization quality (+1.4 improvement in ROUGE-l F-score).

For each pair of nodes (u,v) in the graph, we compute the semantic similarity score (using WordNet ) between every pair of dependency relation (rel: a, b) in u and v as: s(u,v) = Z WN(a,-,aj) >< WN(b,-,bj),

WN(w,—, wj) is defined as the WordNet similarity score between words 212,- and to]?

2There exists various semantic relatedness measures based on WordNet (Patwardhan and Pedersen, 2006).

For these approaches, lists of instances are typically collected from publicly-available resources such as WordNet or Wikipedia (Pasca and Van Durme, 2007;

1Reisinger and Pasca (2009) considered the related problem of finding the most appropriate class for each attribute; they take an existing ontology of concepts ( WordNet ) as a class hierarchy and use a Bayesian approach to decide “the correct level of abstraction for each attribute.”

These efforts focused exclusively on the meronymy relation as used in WordNet (Miller et al., 1990).

Experts can manually specify the attributes of entities, as in the WordNet project (Miller et al., 1990).

In many ways WordNet can be regarded as a collection of commonsense relationships.

We further showed that a weakly supervised heuristic, making use of WordNet sense ranks, can be significantly improved by incorporating information from our system.

WordNet (Fellbaum, 1998) senses of w apply to this occurrence of w.

The dataset contains ordinal judgments of the applicability of WordNet senses on a 5 point scale, ranging from completely difi‘er—ent to identical for eight different lemmas in 50 different sentential contexts.

We apply the same method as in Section 4.3: For each instance in the dataset, we compute the second-order vector of the target verb, contextually constrain it by the first-order vectors of the verb’s arguments, and compare the resulting vector to the vectors that represent the different WordNet senses of the verb.

A widely-used resource is WordNet (Fellbaum, 1998), where relations such as synonymy and hyponymy can be used to generate rules.

single pair of words that are WordNet antonyms (2) Predicates differing by a single word of negation (3) Predicates p(t1, t2) and p(t2, 751) where p is a transitive verb (e.g., beat) in VerbNet (Kipper-Schuler et al., 2000).

Given a lexicographic resource ( WordNet ) and a set of predicates with their instances, we perform the following three steps (see Table 1):

1) Training set generation We use WordNet to generate positive and negative examples, where each example is a pair of predicates.

For every predicate p(t1, t2) 6 P such that p is a single word, we extract from WordNet the set S of synonyms and direct hy-pernyms of p. For every p’ E S, if p’ (t1, t2) 6 P then p(t1, 752) —> p’ (t1, 752) is taken as a positive example.

Some have established concept hierarchies based on manually-built semantic resources such as WordNet (Miller, 1995).

Such hierarchies have good structures and high accuracy, but their coverage is limited to fine-grained concepts (e.g., “Ranunculaceae” is not included in WordNet .).

(2008) link the categories in Wikipedia onto WordNet .

In the WordNet hierarchy, senses are organized according to the “isa” relations.

(2006) provides a global optimization scheme for extending WordNet , which is different from the above-mentioned pairwise relationships identification methods.

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).

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:

For a given predicate q, the system firstly computes its distribution of argument semantic classes based on WordNet .

Clark and Weir (2002) suggest a hypothesis testing method by ascending the noun hierarchy of WordNet .

Cia-ramita and Johnson (2000) model WordNet as a Bayesian network to solve the “explain away” ambiguity.

In Section 3.3, we developed three ways to compute the weight of an edge in the sentence quotation graph, i.e., clue words, semantic similarity based on WordNet and cosine similarity.

The above experiments show that the widely used cosine similarity and the more sophisticated semantic similarity in WordNet are less accurate than the basic CWS in the summarization framework.

We explore three types of cohesion measures: (1) clue words that are based on stems, (2) semantic distance based on WordNet

3.3.2 Semantic Similarity Based on WordNet

We use the well-known lexical database WordNet to get the semantic similarity of two words.

Wherever applicable, we explore different syntactic and semantic representations of the textual content, e. g., extracting the dependency-based representation of the text or generalizing words to their WordNet supersenses (WNSS) (Ciaramita and Altun, 2006).

In all these representations we skip stop words and normalize all words to their WordNet lemmas.

Each word was morphologically simplified using the morphological functions of the WordNet library8.

These tags, defined by WordNet lexicographers, provide a broad semantic categorization for nouns and verbs and include labels for nouns such as food, animal, body and feeling, and for verbs labels such as communication, contact, and possession.

here, we use the product of sim(xu, Qi) and sim(:cv, {mm 62,} to estimate the possibility of being a context-answer pair for (u, v) , where sim(-, is the semantic similarity calculated on WordNet as described in Section 3.5.

The semantic similarity between words is computed based on Wu and Palmer’s measure (Wu and Palmer, 1994) using WordNet (Fellbaum, 1998).1 The similarity between contiguous sentences will be used to capture the dependency for CRFs.

- Similarity with the question using WordNet

CCG parser, WordNet

hypernyms, WordNet

head word, parser SVM hypernyms, WordNet

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.

Most established resources (e.g., WordNet ) represent only the main and widely accepted relationships such as hyper-nymy and meronymy.

There is a large body of related work that deals with discovery of basic relationship types represented in useful resources such as WordNet , including hyper-nymy (Hearst, 1992; Pantel et al., 2004; Snow et al., 2006), synonymy (Davidov and Rappoport, 2006; Widdows and Dorow, 2002) and meronymy (Berland and Charniak, 1999; Girju et al., 2006).

Several algorithms use manually-prepared resources, including WordNet (Moldovan et al., 2004; Costello et al., 2006) and Wikipedia (Strube and Ponzetto, 2006).

Evaluation for hypemymy and synonymy usually uses WordNet (Lin and Pantel, 2002; Widdows and Dorow, 2002; Davidov and Rappoport, 2006).

We use a Maximum Entropy (Berger et al., 1996) classifier with a large number of boolean features, some of which are novel (e. g., the inclusion of words from WordNet definitions).

The WordNet gloss terms had a surprisingly strong influence.

As far as we know, this is the first time that WordNet definition words have been used as features for noun compound interpretation.

Kim and Baldwin (2005) and Turney (2006) use nearest neighbor approaches based upon WordNet (Fellbaum, 1998) and Tumey’s Latent Relational Analysis, respectively.

Voorhees (1993) used the hyponymy (“ISA”) relation in WordNet (Miller, 1990) to disambiguate the polysemous nouns in a text.

(2004) tagged words with 25 root senses of nouns in WordNet .

Some researchers achieved improvements by expanding the disambiguated query words with synonyms and some other information from WordNet (Voorhees, 1994; Liu et al., 2004; Liu et al., 2005; Fang, 2008).

Hard constrains for WordNet relations:

In particular, we found that it becomes nearly impractical to run the ILP formulation including all words in WordNet plus all words in the argument position in Google Web IT.

Therefore we revise those hard constraints to encode various semantic relations ( WordNet and semantic coordination) more directly.

One is word co-occurrence (if word w and word wj occur in the same sentence or in the adjacent sentences, Sim(wi,wj) increases 1), and the other is WordNet (Miller, 1995) based similarity.

3 Total weight of words in the focus candidate using the WordNet similarity.

4 Max weight of words in the focus candidate using the WordNet similarity.

In (Brown et al., 2011), the so called dynamic dependency neighborhoods (DDN), i.e., the set of verbs that are typically collocated with a direct object, are shown to be more helpful than lexical information (e. g., WordNet ).

In supervised language learning, when few examples are available, DMs support cost-effective lexical generalizations, often outperforming knowledge based resources (such as WordNet , as in (Pantel et al., 2007)).

The contribution of (ii) is proportional to the lexical similarity of the tree lexical nodes, where the latter can be evaluated according to distributional models or also lexical resources, e. g., WordNet .

The SS algorithm has access to all the definitions in WordNet (WN).

sourceforgenet, WordNet : :QueryData

as WordNet (Fellbaum, 1998) (WN).

If the node is a preterminal node, we capture its lexical semantic by adding features indicating its WordNet sense information.

Specifically, the first WordNet sense of the terminal word, and all this sense’s hyponym senses will be added as features.

For example, WordNet senses {New Y0rk#], city#], district#],

In step 1, in order to make sure we select a diverse list of words, we consider three attributes of a word: frequency in a corpus, number of parts of speech, and number of synsets according to WordNet .

(2010), we use WordNet to first randomly select one synset of the first word, we then construct a set of words in various relations to the first word’s chosen synset, including hypemyms, hy-ponyms, holonyms, meronyms and attributes.

To capture interesting word pairs, we sample different senses of words using WordNet (Miller, 1995).

In addition, we introduced the new and useful WordNet , Aflect, Length and Negation feature categories.

1 WordNet 20.07 34.07 52.96 11.58 2 Verb Class 14.24 24.84 49.6 10.04 3 MPN 23.84 38.58 49.97 13.16 4 Modality 17.49 28.92 13.84 10.72 5 Polarity 16.46 26.36 65.15 11.58 6 Affect 18.62 31.59 59.8 13.37 7 8 9

WordNet Features: We define four features based on WordNet (Fellbaum, 1998) - Synonyms, Antonyms, Hypernyms and Hyponyms.

Many approaches to the problem use lexical taxonomies such as WordNet to identify the semantic classes that typically fill a particular argument slot for a predicate (Resnik, 1993; Clark and Weir, 2002; Schulte im Walde et al., 2008).

(2007) integrate a model of random walks on the WordNet graph into an LDA topic model to build an unsupervised word sense disambiguation system.

Reisinger and Pasca (2009) use LDA-like models to map automatically acquired attribute sets onto the WordNet hierarchy.

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.

WordNet WordNet (Miller, 1995) is a database of information (synonyms, hypernyms, etc.)

For each entry, WordNet provides a set of synonyms, hypernyms, etc.

Given two spans, we use WordNet to determine semantic similarity by measuring how many synonym (or other) edges are needed to link two

(2008) use WordNet to develop sentiment lexicons in which each word has a positive or negative value associated with it, representing its strength.

The algorithm begins with seed sets of positive, negative, and neutral terms, and then uses the synonym and antonym structure of WordNet to expand those initial sets and refine the relative strength values.

We also replace the negation and the adjective by the antonyms given in WordNet (using the first sense).

named-entities and WordNet Super-Senses4.

4Lexicographic files according to WordNet terminology.

VENSES++ (Tonelli and Delmonte, 2010) applied a rule based anaphora resolution procedure and semantic similarity between candidates and thematic roles using WordNet (Fellbaum, 1998).

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.

We use WordNet to generate candidate lemmas, and we also use a fuzzy match of a concept, defined to be a word in the sentence that has the longest string prefix match with that concept’s label, if the match length is Z 4.

WordNet lemmas and fuzzy matches are only used if the rule explicitly uses them.

Strips off trailing ‘-[0-9]+’ from the concept (for example run—01 —> run), and matches any exact matching word or WordNet lemma.

This algorithm generated a set of synonyms from WordNet and then used the SUBTLEX frequencies to find the most frequent synonym.

This measure is taken from WordNet (Fellbaum, 1998).

Synonym Count Also taken from WordNet , this is the number of potential synonyms with which a word could be replaced.

Distributional models that integrate the visual modality have been learned from texts and images (Feng and Lapata, 2010; Bruni et al., 2012b) or from ImageNet (Deng et al., 2009), e.g., by exploiting the fact that images in this database are hierarchically organized according to WordNet synsets (Leong and Mihalcea, 2011).

Images for the concepts in McRae et al.’s (2005) production norms were harvested from ImageNet (Deng et al., 2009), an ontology of images based on the nominal hierarchy of WordNet (Fellbaum, 1998).

ImageNet has more than 14 million images spanning 21K WordNet synsets.

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.

processing tools, e.g., syntactic parsers (Wiebe, 2000), information extraction (IE) tools (Riloff and Wiebe, 2003) or rich lexical resources such as WordNet (Esuli and Sebastiani, 2006).

Many researchers have explored using relations in WordNet (Miller, 1995), e. g., Esuli and Sabastiani (2006), Andreevskaia and Bergler (2006) for English, Rao and Ravichandran (2009) for Hindi and French, and Perez-Rosas et al.

There is also a mismatch between the formality of many language resources, such as WordNet , and the extremely informal language of social media.