摘要 |
PURPOSE:To improve the extraction rate of effective retrieval keywords by discriminating whether each of words constituting a composite word is a keyword or not. CONSTITUTION:A candidate word is extracted from a requested sentence described in natural Japanese language (1-1), and it is discriminated whether this word exists in a dictionary as one word or not (1-2). If it exists in the dictionary, it is discriminated whether this candidate word is a keyboard independently or not (1-7), and it is used as a retrieval keyword (1-9) if it is a keyword independently, but is it not used as a retrieval keyword (1-10) if it is not a keyword independently. If the candidate word does not exist in the dictionary, it is discriminated whether the candidate word is composite word or not (1-3), and it is regarded as an undiscriminatable word (1-8) if it is not a composite word. If it is a composite word, respective constituting words are looked up in the dictionary (1-4) and it is discriminated whether they are used in a determined word order or not (1-5), and the candidate word is used as a retrieval keyword (1-9) if they are used in said order, but it is discriminated whether the composite word includes an unregistered word or not (1-6) if they are not used in said order. The candidate word is not used as a retrieval keyword (1-10) if it does not include an unregistered word, but the candidate word is regarded as an undiscriminatable word (1-8) if it includes an unregistered word.
|