FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W46651708> ?p ?o ?g. }

• W46651708 abstract "From Social Networks To Distributional Properties: A Comparative Study On Computing Semantic Relatedness Ulli Waltinger (ulli marc.waltinger@uni-bielefeld.de) Text Technology, Bielefeld University Irene Cramer (irene.cramer@udo.edu) Faculty of Cultural Studies, TU Dortmund University Tonio Wandmacher (tonio.wandmacher@uni-osnabrueck.de) Institute of Cognitive Science, University of Osnabr¨uck Abstract Introduction behaving similarly. For example they can show a strong asso- ciative relationship (e.g. ball - goal), and they can be related across different linguistic categories (e.g. milk - white, dog - bark). With respect to the automatic computation of SR, how- ever, many research questions remain unanswered. As stated above, many algorithms were presented in the past decade, but thorough evaluations and comparisons of their ability to capture SR in a human-like manner are still rare. In this work we therefore present a study comparing various semantic re- latedness measures. We evaluate sixteen different algorithms involving four different resources based on a human judge- ment experiment, and we analyze the algorithms from a the- oretical and practical point of view. The paper is organized as follows: the subsequent section describes two works rep- resenting the methodological basis for our study. The various semantic relatedness measures employed in our experiment are described in Section Semantic Relatedness Measures. The experimental setup as well as the results obtained are pre- sented in Section Evaluation. The computation of semantic relatedness (SR) has become an important task in many NLP applications such as spelling error detection, automatic summarization, word sense dis- ambiguation, and information extraction. In recent years a large variety of approaches in computing SR has been pro- posed. However, algorithms and results differ depending on resources and experimental setup. It is obvious that SR plays a crucial role in the lexical retrieval of humans. In various priming experiments it could be shown that semantically re- lated terms influence the semantic processing of one another. For example, if ”bread” is primed by ”butter” it is recog- nized more quickly. Moreover, many theories of memory are based on the notion of SR. The spreading activation theory of (Collins & Loftus, 1975) for example groups lexical items according to their SR in a conceptual graph. Similar ideas can be found in the ACT theory of Anderson (Anderson, 1983). The question that arises for us is, how this kind of relatedness can be determined by automatic means. In the literature the notion of SR is often confounded with semantic similarity; there is however a clear distinction between these notions. Two terms are semantically similar if they behave similarly in a given context and if they share some aspects of meaning (e.g. in the case of synonyms or hypernyms). On the other hand two terms can be semantically strongly related without The task of estimating SR between two given lexical items can be performed by humans in an effortless and intuitive manner. However, this notion is very difficult to formalize from a psycholinguistic or computational point of view. In terms of an evaluation of SR algorithms, most commonly human judgement experiments are conducted. The perfor- mance of an SR measure is determined by directly compar- ing the automatic computed results with those gained from the human judgements via correlation. As a most prominent example Budanitsky and Hirst (Budanitsky & Hirst, 2006) presented a comparison of five semantic relatedness mea- sures for the English language. They recommended a three- level evaluation including theoretical examination, compari- son with human judgements and evaluation with respect to a given NLP-application. The measures were evaluated on two different data sets: The first data set was compiled by Ruben- stein and Goodenough (Rubenstein & Goodenough, 1965); it contained 65 word-pairs. The second set, containing 30 word pairs, was compiled by Miller and Charles (Miller & Charles, 1991). For each of the five measures, Budanitsky and Hirst reported correlation coefficients between 0.78 and 0.83. Boyd-Graber et al. (Boyd-Graber, Fellbaum, Osher- In recent years a variety of approaches in computing seman- tic relatedness have been proposed. However, the algorithms and resources employed differ strongly, as well as the results obtained under different experimental conditions. This article investigates the quality of various semantic relatedness mea- sures in a comparative study. We conducted an extensive ex- periment using a broad variety of measures operating on so- cial networks, lexical-semantic nets and co-occurrence in text corpora. For two sample data sets we obtained human relat- edness judgements which were compared to the estimates of the automated measures. We also analyzed the algorithms im- plemented and resources employed from a theoretical point of view, and we examined several practical issues, such as run time and coverage. While the performance of all measures is still mediocre, we could observe that in terms of of coverage and correlation distributional measures operating on controlled corpora perform best. Keywords: Semantic Relatedness; Semantic Similarity; Hu- man Judgement; Social Networks; WordNet; LSA; Related Work" @default.
• W46651708 created "2016-06-24" @default.
• W46651708 creator A5019881745 @default.
• W46651708 creator A5059629985 @default.
• W46651708 creator A5089108608 @default.
• W46651708 date "2009-01-01" @default.
• W46651708 modified "2023-09-23" @default.
• W46651708 title "From Social Networks To Distributional Properties: A Comparative Study On Computing Semantic Relatedness" @default.
• W46651708 cites W136846643 @default.
• W46651708 cites W138707325 @default.
• W46651708 cites W158057341 @default.
• W46651708 cites W1647729745 @default.
• W46651708 cites W1956559956 @default.
• W46651708 cites W1968317778 @default.
• W46651708 cites W2038721957 @default.
• W46651708 cites W204849900 @default.
• W46651708 cites W2080100102 @default.
• W46651708 cites W2082557450 @default.
• W46651708 cites W2100935296 @default.
• W46651708 cites W2102131037 @default.
• W46651708 cites W2108656749 @default.
• W46651708 cites W2117805756 @default.
• W46651708 cites W2120779048 @default.
• W46651708 cites W2129773034 @default.
• W46651708 cites W2133322356 @default.
• W46651708 cites W2135255848 @default.
• W46651708 cites W2136480620 @default.
• W46651708 cites W2136930489 @default.
• W46651708 cites W2140887277 @default.
• W46651708 cites W2147152072 @default.
• W46651708 cites W2165897980 @default.
• W46651708 cites W2397300033 @default.
• W46651708 cites W2404602200 @default.
• W46651708 cites W2534712034 @default.
• W46651708 hasPublicationYear "2009" @default.
• W46651708 type Work @default.
• W46651708 sameAs 46651708 @default.
• W46651708 citedByCount "6" @default.
• W46651708 countsByYear W466517082012 @default.
• W46651708 countsByYear W466517082013 @default.
• W46651708 countsByYear W466517082014 @default.
• W46651708 countsByYear W466517082017 @default.
• W46651708 crossrefType "proceedings-article" @default.
• W46651708 hasAuthorship W46651708A5019881745 @default.
• W46651708 hasAuthorship W46651708A5059629985 @default.
• W46651708 hasAuthorship W46651708A5089108608 @default.
• W46651708 hasConcept C130318100 @default.
• W46651708 hasConcept C154945302 @default.
• W46651708 hasConcept C170858558 @default.
• W46651708 hasConcept C204321447 @default.
• W46651708 hasConcept C23123220 @default.
• W46651708 hasConcept C2522767166 @default.
• W46651708 hasConcept C41008148 @default.
• W46651708 hasConceptScore W46651708C130318100 @default.
• W46651708 hasConceptScore W46651708C154945302 @default.
• W46651708 hasConceptScore W46651708C170858558 @default.
• W46651708 hasConceptScore W46651708C204321447 @default.
• W46651708 hasConceptScore W46651708C23123220 @default.
• W46651708 hasConceptScore W46651708C2522767166 @default.
• W46651708 hasConceptScore W46651708C41008148 @default.
• W46651708 hasIssue "31" @default.
• W46651708 hasLocation W466517081 @default.
• W46651708 hasOpenAccess W46651708 @default.
• W46651708 hasPrimaryLocation W466517081 @default.
• W46651708 hasRelatedWork W111662600 @default.
• W46651708 hasRelatedWork W1140107 @default.
• W46651708 hasRelatedWork W128594437 @default.
• W46651708 hasRelatedWork W1445288453 @default.
• W46651708 hasRelatedWork W155194117 @default.
• W46651708 hasRelatedWork W1987145075 @default.
• W46651708 hasRelatedWork W1989080863 @default.
• W46651708 hasRelatedWork W2032741002 @default.
• W46651708 hasRelatedWork W2070584861 @default.
• W46651708 hasRelatedWork W2087739686 @default.
• W46651708 hasRelatedWork W2123457740 @default.
• W46651708 hasRelatedWork W2151226364 @default.
• W46651708 hasRelatedWork W2160353247 @default.
• W46651708 hasRelatedWork W2296563821 @default.
• W46651708 hasRelatedWork W2551701198 @default.
• W46651708 hasRelatedWork W2777502251 @default.
• W46651708 hasRelatedWork W2785469404 @default.
• W46651708 hasRelatedWork W2972278525 @default.
• W46651708 hasRelatedWork W2991480621 @default.
• W46651708 hasRelatedWork W4866103 @default.
• W46651708 hasVolume "31" @default.
• W46651708 isParatext "false" @default.
• W46651708 isRetracted "false" @default.
• W46651708 magId "46651708" @default.
• W46651708 workType "article" @default.