| 摘要 |
PURPOSE:To use not domain knowledge but a topic expansion mode and linguistic knowledge for conversation data to recognize a topic by divisionally subjecting the conversation data to basic expansion and semantic expansion and integrating topics in basic expansion and semantic expansion. CONSTITUTION:Input interactive data 110 is subjected to morpheme analysis processing. This processing segments the character string of inputted interactive data 110 into individual words to obtain a work string and identifies the part of speech of each word, conjugations of inflective words, etc. That is, inputted interactive data 110 is subjected to topic structure recognition preprocessing 120 to obtain a block, and this block is subjected to basic expansion processing 130 and semantic expansion processing 140 to determine the topic level sections independently of each other. Topics is basic expansion and semantic expansion are integrated in integration processing 150 based on topic levels determined by basic expansion processing 130 and semantic expansion processing 140, thus determining a topic structure 160. |
