The Descent of Hierarchy, and Selection in Relational Semantics

The Descent of Hierarchy, and Selection in Relational Semantics � Barbara Rosario SIMS UC Berkeley Berkeley, CA 94720 rosario@sims.berkeley.edu Marti A. Hearst SIMS UC Berkeley Berkeley, CA 94720 hearst@sims.berkeley.edu Charles Fillmore ICSI UC Berkeley Berkeley, CA 94720 fillmore@icsi.berkeley.edu Abstract In many types of technical texts, meaning is embedded in noun compounds. A language un- derstanding program needs to be able to inter- pret these in order to ascertain sentence mean- ing. We explore the possibility of using an ex- isting lexical hierarchy for the purpose of plac- ing words from a noun compound into cate- gories, and then using this category member- ship to determine the relation that holds be- tween the nouns. In this paper we present the results of an analysis of this method on two- word noun compounds from the biomedical do- main, obtaining classification accuracy of ap- proximately 90%. Since lexical hierarchies are not necessarily ideally suited for this task, we also pose the question: how far down the hi- erarchy must the algorithm descend before all the terms within the subhierarchy behave uni- formly with respect to the semantic relation in question? We find that the topmost levels of the hierarchy yield an accurate classification, thus providing an economic way of assigning rela- tions to noun compounds. 1 Introduction A major difficulty for the interpretation of sentences from technical texts is the complex structure of noun phrases and noun compounds. Consider, for example, this title, taken from a biomedical journal abstract: Open-labeled long-term study of the subcutaneous sumatriptan efficacy and tolerability in acute mi- graine treatment. An important step towards being able to interpret such technical sentences is to analyze the meaning of noun compounds, and noun phrases more generally. � With apologies to Charles Darwin. Interpretation of noun compounds (NCs) is highly de- pendent on lexical information. Thus we explore the use of a large corpus (Medline) and a large lexical hierarchy (MeSH, Medical Subject Headings) to determine the re- lations that hold between the words in noun compounds. Surprisingly, we find that we can simply use the juxta- position of category membership within the lexical hier- archy to determine the relation that holds between pairs of nouns. For example, for the NCs leg paresis, skin numbness, and hip pain, the first word of the NC falls into the MeSH A01 (Body Regions) category, and the second word falls into the C10 (Nervous System Diseases) cat- egory. From these we can declare that the relation that holds between the words is “located in”. Similarly, for influenza patients and aids survivors, the first word falls under C02 (Virus Diseases) and the second is found in M01.643 (Patients), yielding the “afflicted by” relation. Using this technique on a subpart of the category space, we obtain 90% accuracy overall. In some sense, this is a very old idea, dating back to the early days of semantic nets and semantic grammars. The critical difference now is that large lexical resources and corpora have become available, thus allowing some …

Content truncated to 500 words