Quest for a Computerised Semantics.
Adrian R. Leslie · UVic’s Research and Learning Repository (University of Victoria) · 1973
The objective was to colligate the various strands of research in the literature of computational linguistics that have to do with the computational treatment of semantic content so as to encode it into a computerised dictionary. In chapter 1 the course of mechanical translation (1947-1960) and quantitative linguistics is traced to demonstrate the limitations of computational linguistics without semantics. Chapter 2 covers linguistic research in the 1960s, which was essentially an offshoot of transformational grammar. In chapter 3, various classification schemes are examined as a body of experience from which to draw conclusions on the constraints to which the construction of a computerised dictionary is subject. Chapter 4 is a synthesis of all this data in the form of a model dictionary entry. In chapters 2 and 3 the approaches to semantics are of two types. In one, the semantic categories for each dictionary entry were in the form of unordered elements, and the means of applying them in text was placed within the realm of grammar, In the other type, syntagmatic relationships occurred between the encoded components of dictionary definitions just as they did between those of utterances in a text, The conclusion reached is that the latter type of approach provides firmer foundations upon which to set up a computerised dictionary, as it shows how information is structured in terms of its application in text. In the model structured entry the components of the definitions of a word are tagged according to their part of speech. The representation of a discourse then is determined by mapping out the dictionary definitions of words as they occur in text. The necessity for integrating semantic components into the structure of a dictionary was brought out at the sixth annual symposium on mechanical translation held by the National Research Council at Ottawa in April, 1972. Several teams apparently achieved some results through the parsing of texts by means of ad hoc dictionaries. Since the problems of cross-referencing and of the recognition of semantic grouping remain unsolved, the key to success even in mechanical translation lies in computerised lexicography. It seems unlikely that much more progress will be achieved without it. While the solution of these complex problems· is beyond the scope of this single thesis, a formulation of them is provided as a starting point for further research.