Author Identification in Chatlogs using Formal Concept Analysis
Linda van der Knaap, F.A. Grootjen · Belgium-Netherlands Conference on Artificial Intelligence · 2007
In this project a new approach to author identification in chatlogs is investigated. The method is based on Formal Concept Analysis (FCA): a classification technique that is able to classify users of a chat-channel based on the words they posted. The concepts produced by FCA are scored using various evaluation functions. The results of this explorative survey are most promising. Although identification is far from perfect and many tests still need to be done, FCA certainly has potential as a method for author identification in chat texts.