| 摘要 |
PURPOSE:To correctly decide an anticipatory word by deciding extraction-object area from which candidate auticipatory words for a pronoun are extracted based upon a syntax structure of the sentence containing a pronoun and extracting candidate anticipatory words from the extraction-object area to identify based upon pronoun,s grammatical nature and semantic nature. CONSTITUTION:The English sentence under analysis is examined if it contains a pronoun (2). The process completes if no pronoun is found. If any pronoun is found, the sentence containing the pronoun is examined if it is a subordinate cluas which starts with such subordinate conjunctions as 'although'. 'after', 'if', etc., which may cause reverse reference/correspondence of a pronoun. If it does not occur, a candidate anticipatory object area is decided before the pronoun and is it occurs, it is decided before and after the pronoun. In a candidate anticipatory word extracting process, noun phrases are sequentially extracted, starting with the top of the extraction object area, or the top part of the sentence whose analysis has completed. Then, they are piled up in a stack and the grammatical and semantic nature obtained by syntax analysis (5) are also stored. |
