Text classification is an information retrieval task for which texts are classified as either relevant or irrelevant to a topic. We have developed three algorithms that use information extraction to support text classification: the relevancy signatures algorithm , the augmented relevancy signatures algorithm , and the case-based text classification algorithm . These algorithms classify texts using linguistic expressions and contexts extracted from the texts. The difference between the algorithms is the amount of extracted information that they use. The relevancy signatures algorithm uses simple phrases to classify texts. This algorithm performs well in domains characterized by strong key phrases. The augmented relevancy signatures algorithm uses phrases and semantic features representing local context. This algorithm can make semantic distinctions that the relevancy signatures algorithm cannot. The third algorithm, the case-based text classification algorithm, uses sentence contexts to classify texts based on multiple pieces of extracted information. One of the main strengths of this approach is that the algorithms can be easily ported across domains when combined with AutoSlog to generate a good dictionary of extraction patterns. The text classification algorithms have been evaluated in three domains: terrorism, joint ventures, and microelectronics.