Desiderata for Tagging with WordNet Synsets or MCCA Categories

Desiderata for Tagging with WordNet Synsets or MCCA Categories Kenneth C. Litkowski CL Research 20239 Lea Pond Place Gaithersburg, MD 20879 (Email: ken@tires.corn) (Web site: http://www.clres.com) 1 Abstract Minnesota Contextual Content Analysis (MCCA) is a technique for characterizing the concepts and themes occurring in text (sentences, paragraphs, interview transcripts, books). MCCA tags each word with a category and examines the distribution of categories against norms representing general usage of categories. MCCA also scores texts in terms of social contexts that are similar to different functions of language. Distributions can be analyzed using non-agglomerative clustering to characterize the concepts and themes. MCCA categories have been mapped to WordNet senses. The &fining characteristics that emerge from the mapping and the statistical techniques used in MCCA for analyzing concepts and themes suggest that tagging with WordNet synsets or MCCA categories may produce epiphenomenal results that are misleading. We suggest that WordNet synsets and MCCA categories be augmented with further lexical semantic information for use after text is tagged or categorized. We suggest that such information is useful not only for the primary purposes of disambiguation in parsing and text classification in content analysis and information retrieval, but also for tasks in corpus analysis, discourse analysis, and automatic text summarization. 2 Introduction Content analysis provides distributional methods for analyzing characteristics of textual material. Its roots are the same as computational linguistics (CL), but it has been largely ignored in CL until recently (Dunning, 1993; Carletta, 1996; Kilgarriff, 1996). One content analysis approach, Minnesota Contextual Content Analysis (MCCA) (McTavish & Pirro, 1990), in use for over 20 years and with a well-developed dictionary category system, contains analysis methods that provide insights into the use of WordNet (Miller, et al., 1990) for tagging. We describe the unique characteristics of MCCA, how its categories relate to WordNet synsets, the analysis methods used in MCCA to provide quantitative information about texts, what implications this has for the use of WordNet in tagging, and how these techniques may contribute to lexical semantic tagging. Specifically, we show that WordNet provides a backbone, but that additional lexical semantic information needs to be associated with WordNet synsets. We describe novel perspectives on how this information can be used in various NLP tasks. 3 Minnesota Contextual Content Analysis MCCA differs from other content analysis techniques in using a norm for examining the distribution of its categories in a given text. The 116 categories used in the dictionary to characterize words) like other content analysis category systems, are heuristic in nature. Each category has a name (e.g., activiO~, fellow feeling, about changing, human roles, expresaion arena). The distinguishing characteristic of MCCA is that the emphasis of each category is normed in two ways. Categories that are emphasized in a text (E-scores) are normed against expected general usage of categories based on the Brown corpus (Kucera & Francis, 1967). The second way is based on relative usage of categories expected in four broad institutional areas. The latter is based on some initial research and subsequent …

Content truncated to 500 words