| 摘要 |
<P>PROBLEM TO BE SOLVED: To accurately detect a spam description from text data existing on the Internet. Ž<P>SOLUTION: A spam/event detection device includes: a data cleansing means 22 which removes spam description from accumulated text data to be detected by use of a spam description rule; a data extraction means 24 which classifies the data with no spam description into a plurality of data by a predetermined method; time-series burst detection means 311, 312 to 31n which detect time-series bursts of the data with no spam description respectively in association with the plurality of classified data; burst feature extraction means 321, 322 to 32n which respectively extract features of each detected burst period; and spam/event feature determination means 331, 332 to 33n which determines spam description or event description respectively based on the extracted features. The spam description rule is rewritten based on span information determined as a spam description. Ž<P>COPYRIGHT: (C)2009,JPO&INPIT Ž |
