Detection of Topic Change in IRC Chat Logs

We attack the problem of topic segmentation in the domain of Internet Relay Chat logs. In this process, we examine the previous work in text segmentation using a variety of methods. After considering the pros and cons of the methods, we employ Text Tiling, pause detection, and latent semantic analysis because they did not require the usage of large pre-tagged corpora. With these systems in place, we consider the properties and problems that exist when considering the domain of internet chat. To this end, we examine our results and show them to be fair at best.