This paper presents ForumMiner, a tool for the automatic analysis of students' posts in asynchronous discussions. ForumMiner uses a text mining system to extract graphs from texts that are given to students as a basis for their discussion. These graphs contain the most relevant terms found in the texts, as well as the relationships between them. By comparing these graphs with the students' writings, and using other information such as number and structure of posts, the tool is able to assess the relevance of each student's contributions. A group of 18 teachers was consulted about the most important aspects to be considered in the analysis of posts, and the tool has been built according to these aspects. Results of an experiment involving the analysis of 682 posts showed that the average of the ratings given to these posts by teachers was very close to those provided by the automated tool. Such results demonstrate the tool's potential to assist teachers in evaluating student participation in asynchronous discussions.