| 摘要 |
PURPOSE: To automatically add the semantic limitation to the grammar of a natural language by generalizing the positive/negative examples related to each grammatical rule generated by evaluating the justice of a partial phrase structure to an input sentence string by means of a lexical hierarchy. CONSTITUTION: When a morpheme is analyzed and an input sentence added with a correct syntax analysis tree and a grammatical rule are inputted to a phrase structure evaluation part 1, the part 1 compares the grammatical rule with the analysis tree to decide a proper or improper partial structure. An example generation part 2 sorts the proper and improper partial structures as the positive and negative examples respectively for each corresponding grammatical rule. An example generalization part 3 extracts the areas to which the positive example is corresponded in common out of a lexical hierarchy that prescribes a generalized lexical hierarchical structure. Then the part 3 converts the vocabularies included in the positive example into the general vocabularies and outputs a grammatical rule which is added with the semantic limitation. A grammatical rule selection part 4 selects the optimum one of those grammatical rules that are generated at the part 3. |
