Parsing vs. Text Processing in the Analysis of Dictionary Definitions

PARSING VS. TEXT PROCESSING IN THE ANALYSIS OF DICTIONARY DEFINITIONS Thomas Ahlswede and Martha Evens Computer Science Dept. Illinois Institute of Technology Chicago, 11. 60616 312-567-5153 ABSTRACT We have analyzed definitions from Webster's Seventh New Collegiate Dictionary using Sager's Linguistic String Parser and again using basic UNIX text processing utilities such as grep and awk. Tiffs paper evaluates both procedures, compares their results, and discusses possible future lines of research exploiting and combining their respective strengths. Introduction As natural language systems grow more sophisticated, they need larger and more d~led lexicons. Efforts to automate the process of generating lexicons have been going on for years, and have often been combined with the analysis of machine-readable dictionaries. Since 1979, a group at HT under the leadership of Manha Evens has been using the machine-readable version of Webster' s Seventh New Collegiate Dictionary (W7) in text generation, information retrieval, and the theory of lexical- semantic relations. This paper describes some of our recent work in extracting semantic information from WT, primarily in the form of word pairs linked by lexical-semantic relations. We have used two methods: parsing definitions with Sager's Linguistic String Parser (LSP) and text processing with a combination of UNIX utilities and interactive editing. We will use the terms "parsing" and "text processing" here primarily with reference to our own use of the LSP and UNIX utilities respectively, but will also use them more broadly. "Parsing" in this more general sense will mean a computational technique of text analysis drawing on an extensive database of linguistic knowledge, e.g., the lexicon, syntax and/or semantics of English; "text processing" will refer to any computational technique that involves little or no such knowledge. This research is supported by National Science Foundation grant IST 87-03580. Our thanks also to the G & C Merriam Company for permission to use the dictionary tapes. Our model of the lexicon emphasizes lexical and semantic relations between words. Some of these relationships axe fan~iliar. Anyone who has used a dictionary or thesaurus has encountered synonymy, and perhaps also antonymy. W7 abounds in synonyms (the capitalized words in the examples below): (1) funny 1 la aj affording light mirth and laughter : AMUSING (2) funny 1 lb aj seeking or intended to amuse : FACETIOUS Our notation for dictionary definitions consists of: (1) the entry (word or phrase being defined); (2) the homograph number (multiple homographs are given sepmaw entries in W7); (3) the sense number, which may include a subsense letter and even a sub- subseuse number (e.g. 263); (4) the text of the definition. We commonly express a relation between words through a triple consisting of Wordl, Relation, Word2: (3) funny SYN amusing (4) funny SYN facetious A third relation, particularly important in W7 and in dictionaries generally, is taxonomy, the species-genus relation or (in artificial intelligence) the IS-A relation. Consider the entries: (5) dodecahedron 0 0 n a solid having 12 plane faces (6) build 1 1 vt to form by ordering and uniting materials... These …

Content truncated to 500 words