World Knowledge and Natural Language Analysis (review)

This volume introduces research conducted within the OMNIS (Danish acronym for ‘World Knowledge and Natural Language System’) project, initiated and carried out from September 1994 to November 1998 by the Department of Computational Linguistics, Copenhagen Business School. The main aim of the project was to investigate methods and principles of representing world knowledge in natural languages and to apply this knowledge in natural language systems, i.e. natural-language-based databases and translation systems. The six papers in this volume were presented at the final OMNIS workshop held in November 1998 in Copenhagen.

The first paper, ‘The project “Computer-aided ontology structuring” (CAOS)’ by Bodil Nistrup Madsen, Hanne Erdman Thomsen and Carl Vikner, introduces the computational system CAOS, designed to support a terminologist in the construction of ontologies. The authors outline the process of semiautomatic construction of an ontology within the system and discuss the problems involved with this work.

The second paper, ‘Typed feature specifications for establishing terminological equivalence relations’ by Hanne Erdman Thomsen, explores the feasibility of automatically establishing equivalence relations between concepts in two languages in the process of a translation database creation. Thomsen describes various types of equivalence, investigates how to rank candidate equivalents, and introduces an algorithm for finding the closest equivalent.

The third paper, ‘Can you be more specific? World knowledge and the analysis of specificity’ by Stig W. Jørgensen, analyzes the notion of specificity and defines a special kind of specificity closely connected to his notion of referring expressions. Jørgensen also shows the role of world knowledge in the recognition of NPs as referring expressions.

The fourth paper, ‘What are proper names and how do we identify them?’ by Mette Nelson, addresses the problem of recognition and automatic identification of proper names by computers. Seven prototypical structural properties of proper names are introduced by Nelson based on the differences between common nouns and proper names as well as on the internal structure of proper names. Nelson shows that these properties in combination with capital letters are sufficient to automatically identify proper names in natural language texts.

The two papers concluding the volume deal with the concept of locations. In her paper ‘Properties of motion verbs’, Christina Philp describes some properties of Danish motion verbs in combination with prepositional phrases. She shows that the status of the prepositional phrase following a motion verb can be decisive with respect to the meaning of the motion verb. Philp also investigates abstract motion verbs and demonstrates that these do not carry the same properties as other motion verbs since they have no indication of manner of motion. The final paper, ‘Locations: On the meaning of phrases like i Paris (in Paris) and i fœngsel (in prison)’ by Finn Sørensen, examines the meaning of nonreferring expressions vs. that of referring expressions, used in both cases with prepositions. Sørensen demonstrates that locations in Danish are constructed on the basis of material and abstract entities.

Despite the technical nature of the problems under discussion, the book is written in simple language and is not limited to the range of problems dealt with in computational linguistics only. Because of the philosophical and linguistic importance of the concepts of ontology, proper names, specificity, locations, etc., the issues raised in the papers are of wide crosslinguistic application and will be undoubtedly of interest to linguists working in areas other than natural language processing.