Integrated Linguistic Processing with Constraints and Preferences
The CHORUS project is concerned with the semantic processing of natural language utterances, i.e. determining their meaning.
A specialty of human language understanding is the ability to deal with missing or ambiguous information. Incomplete information is very common in realistic situations of communication between humans; sentences in spoken language for example might be pronounced imprecise, they might be ambiguous or even grammatically wrong. Humans are in most situations able to react in an appropriate manner, even when the meaning of the perceived utterance is not clear, or they can reconstruct the missing information by using knowledge about the context.
People often prefer a particular reading among the many possible readings. This gives them additional information by which to resolve the remaining underspecfication. But preference may lead to wrong conclusions which have to be revised later on. Systems for natural language processing, too, have to be prepared to cope with this kind of imperfect input. Therefore the goal of the CHORUS project is to model this ability of human beings on a computer by modelling linguistic processing with preferences and constraints.