Towards strategies for processing relationships between multiple relation participants in knowledge patterns

Knowledge patterns are an effective tool for automatically or semi-automatically locating specific types of information — such as conceptual relations — in text corpora. However, pattern-based approaches are vulnerable to a number of types of variation; one of these is the expression of multiple participants in a single occurrence of a relation. Despite the challenges posed by this phenomenon, however, such cases may be particularly rich in useful information about the principal relation expressed and/or others involving the relation participants. Strategies that allow for formal evaluation and processing of such cases can enable pattern-based applications to capitalize on this information. This article will present a description, in English and French, of the types of relation occurrences in which multiple participants in CAUSE–EFFECT and ASSOCIATION relations are named, and the information that each can offer in addition to these primary relations. In addition, some strategies and challenges for processing these cases automatically will be discussed, and the phenomena as observed in the two languages will be briefly compared.