| 摘要 |
PROBLEM TO BE SOLVED: To provide a document summarizing device which can obtain a high precision summary from a title and its body like a newspaper article. SOLUTION: A title extraction part 9 extracts the title and body from a document inputted from a document input part 1 and a document analysis part 5 decomposes the title into words and the body into sentences and words. A context vector generation part 6 generates the context vectors of the title and the sentences in the body. A context vector comparison part 7 compares the context vectors of the title and the respective sentences of the body with each other to calculate the distances between the context vectors. Sentences close to the title are generated and used as an important part of the document. Thus, the input document is analyzed by using the context vectors to extract the important part of good quality through an easy process without assuming a specific document form and a specific context. |
