The large amount of verbal data from common knowledge-elicitation methods suggests using the data directly for knowledge acquisition by means of sophisticated natural-language analyzers (NLAs). In this paper, we analyze the feasibility of such an approach theoretically and present a number of examples. In the theoretical part of the text we first provide a detailed analysis of the entities involved, i.e., the domains of expertise, the qualities of knowledge about domains, the properties of generic sentences and texts in natural languages, and the conclusions to be drawn from the limited expressiveness of formal representations. Then we discuss the processes of transforming knowledge into natural language and of transforming natural language into formal language. Since much can go wrong in both processes, we derive desired relations or validity criteria among the entities and strategies to meet the criteria. We believe that this broad theoretical framework can be used to analyze and compare existing attempts at directly using natural language for knowledge acquisition, and thus assess the present status of the field.
Note: The Institute of Electrical and Electronics Engineers, Incorporated is distributing this Article with permission of the International Business Machines Corporation (IBM) who is the exclusive owner. The recipient of this Article may not assign, sublicense, lease, rent or otherwise transfer, reproduce, prepare derivative works, publicly display or perform, or distribute the Article.