| Abstract | This paper presents some statistical observations relevant to Japanese keyphrase extraction, as well as the details of the implementation of a keyphrase extraction algorithm (called Extractor) for Japanese documents. Parts of the algorithm include an efficient method of extracting the keyphrase candidates, a way to pinpoint the most probable keyphrases using contextual information, a technique to find the main ideas conveyed in the text, and a way to express those ideas using extracted phrases. Finally, a comparison with the English and French versions of Extractor will be presented. |
|---|
