GHH at SemEval-2018 Task 10: Discovering Discriminative Attributes in Distributional Semantics

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 947–952 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics GHH at SemEval-2018 Task 10: Discovering Discriminative Attributes in Distributional Semantics Mohammed Attia Google Inc. New York City NY, 10011 attia@google.com Younes Samih Dept. of Computational Linguistics Heinrich Heine University, D¨usseldorf, Germany samih@phil.hhu.de Manaal Faruqui Google Inc. New York City NY, 10011 mfaruqui@google.com Wolfgang Maier Independent Researcher T¨ubingen, Germany wolfgang.maier@gmail.com Abstract This paper describes our system submission to the SemEval 2018 Task 10 on Capturing Discriminative Attributes. Given two con- cepts and an attribute, the task is to determine whether the attribute is semantically related to one concept and not the other. In this work we assume that discriminative attributes can be detected by discovering the association (or lack of association) between a pair of words. The hypothesis we test in this contribution is whether the semantic difference between two pairs of concepts can be treated in terms of measuring the distance between words in a vector space, or can simply be obtained as a by-product of word co-occurrence counts. 1 Introduction Equipped with their cognitive skills, encyclope- dic knowledge and linguistic competence, humans generally can identify the lexical association or se- mantic relation between two words or concepts with relative ease. However, building a com- putational model for identifying fine-grained se- mantic relations (such as synonymy, antonymy, hyponymy, or hypernymy, meronymy, holonymy, metonymy, containment or causality) or even de- tecting binary relatedness has proven to be a chal- lenging task. Efforts to model semantic representation com- putationally are generally classified into statistical and knowledge-driven semantics. This classifica- tion depends on whether the assumption is that hu- man knowledge is encapsulated in language man- ifestation or that explicit manual encoding of this knowledge is needed. The statistical approach to the encoding of semantic relations is referred to as “distributional semantics” or “distributed word representations” (Speer et al., 2017), and its the- oretical appeal stems from the fact that it gives practical application to the Firthian dictum “You shall know a word by the company it keeps” (Firth, 1957) which has become commonsense wisdom in lexical semantics. Features of the statistical model are extracted from unstructured data, such as words embeddings, n-gram counts, or directly from raw data. The basic idea with word embeddings is to formulate semantic relations in arithmetic fash- ion by creating a vector space in which words with similar contextual embeddings have closer vectors distance (Hinton et al., 1986; Rumelhart et al., 1986; Elman, 1990; Bengio et al., 2003; Kann and Schtze, 2008; Mikolov et al., 2013c). The public availability of word embedding train- ing programs such as word2vec (Mikolov et al., 2013a) and GloVe (Pennington et al., 2014) al- lowed researchers to create models with different parameters and dimensionality sizes for different purposes including capturing semantic relations (Gladkova et al., 2016; Attia et al., 2016). The Google n-gram corpus (Brants and Franz, 2006) is a collection of English word n-grams and their observed …

Content truncated to 500 words