Recently, collaborative discussions based on the participant generated documents, e.g., customer questionnaires, aviation reports and medical records, are required in various fields such as marketing, transport facilities and medical treatment, in order to share useful knowledge which is crucial to maintain various kind of securities, e.g., avoiding air-traffic accidents and malpractice. We introduce several techniques in natural language processing for extracting information from such text data and verify the validity of such techniques by using aviation documents as an example. We automatically and statistically extract from the documents related words that have not only taxonomical relations like synonyms but also thematic (non-taxonomical) relations including causal and entailment relations. These related words are useful for sharing information among participants. Moreover, we acquire domain-specific terms and phrases from the documents in order to pick up and share important topics from such reports.