Anaphora Resolution Based on Dynamic Context Model in Database-Oriented Discourse

概要

論文の詳細を見る
This paper extends the previous work which used Context Model as a means for resolving Anaphora Problem in Database-Oriented Discourse. When a discourse is concerned with the information of more than one object, a focus traversal in a discourse causes the system to keep account of many context models which need to be kept for further unifying. This paper, therefore, introduces Dynamic Context Model, i. e., the using of context models which can be dynamically changed when a shift in focus occurs. Using context models to provide the correct referent is still sufficient in the case of each referent described in a discourse be distinguishably classified according to its attributes. However, when the candidates of referent have common attributes, the misidentification of the intended referent, called referent failure, can occur. In such a case, the system should be able to recover the misinterpretation of anaphora. As a possible approach to handle this problem, this type of an ambiguous anaphora will be treated as a nondeterministically interpretable anaphora which must be reinterpreted if the referent failure is found. Accordingly, this paper also proposes an algorithmic theory of anaphora resolution based on Dynamic Context Model which covers the referent failure problem. That is, it will embody the method of dynamically anaphora interpreting which consists of the process of capturing the intended referent and the process of reinterpreting the anaphora. To accomplish this task, the memory organization is needed. Those are CM stack, PCM list, and REUNIFY stack.
社団法人人工知能学会の論文
1990-11-01

Anaphora Resolution Based on Dynamic Context Model in Database-Oriented Discourse

スポンサーリンク

概要

著者

関連論文

スポンサーリンク