A Thesaurus-Based Semantic Classification of English Collocations

A Thesaurus-Based Semantic Classification of English Collocations Chung-chi Huang, Chiung-hui Tseng, Kate H. Kao, Jason S. Chang ISA, National Tsing Hua University {u901571, smilet, msgkate, jason.jschang}@gmail.com Abstract We propose a new method for organizing the numerous collocates into semantic thesaurus categories. The approach introduces a thesaurus-based semantic classification model automatically learning semantic relations for classifying adjective-noun (A-N) and verb-noun (V-N) collocations into different categories. Our model uses a random walk over weighted graph derived from WordNet semantic relation. We compute a semantic label stationary distribution via an iterative graphical algorithm. The performance for semantic cluster similarity and the conformity of semantic labels are both evaluated. The resulting semantic classification establishes as close consistency as human judgments. Moreover, our experimental results indicate that the thesaurus structure is successfully imposed to facilitate grasping concepts of collocations. It might improve the performance of the state-of-art collocation reference tools. Keywords: Collocations, Semantic classification, Semantic relations, Random walk algorithm, Meaning access index. 1. Introduction Submitting queries (e.g., a search keyword “beach” for a set of adjective collocates) to collocation reference tools typically return many collocates (e.g., collocate adjectives with a pivot word “beach”: “rocky”, “golden”, “beautiful”, “pebbly”, “splendid”, “crowded”, “superb”, etc.) extracted from a English corpus. Applications of automatic extraction of collocations such as TANGO (Jian, Chang & Chang, 2004) have been created to answer queries of collocation usage. Unfortunately, existing collocation reference tools sometimes present too much information in a batch for a single screen. With web corpus sizes rapidly growing, it is not uncommon to find thousands collocates for a query word. An effective reference tool might strike a balance between quantity and accessibility of information. To satisfy the need for presenting a digestible amount of information, a promising approach is to automatically partition words into various categories to support meaning access to search results and thus give a thesaurus index. Instead of generating a long list of collocates, a good, better presentation could be composed of clusters of collocates inserted into distinct semantic categories. We present a robust thesaurus-based classification model that automatically group collocates of a given pivot word focusing on: (1) the adjectives in adjective-noun pairs (A-N); (2) the verbs in verb-noun pairs (V-N); and (3) the nouns in verb-noun pairs (V-N) into semantically related classes. Our model has determined collocation pairs that learn the semantic labels automatically during random walk algorithm by applying an iterative graphical approach and partitions collocates for each collocation types (A-N, V-N and V-N mentioned above). At runtime, we start with collocates in question with a pivot word, which is to be assigned under a set of semantically related labels for the semantic classification. An automatic classification model is developed for collocates from a set of A-N and V-N collocations. A random walk algorithm is proposed to disambiguate word senses, assign semantic labels and partition collocates into meaningful groups. As part of our evaluation, two metrics are designed. We assess the performance of collocation clusters classified by a robust evaluation metric and evaluate the conformity …

Content truncated to 500 words