Configuring Domain Knowledge for Natural Language Understanding

Knowledge-based configuration has been used for numerous applications including natural language processing (NLP). By formalising property grammars as a configuration problem, it has been shown that configuration can provide a flexible, nondeterministic, method of parsing natural language. However, it focuses only on syntactic parsing. In contrast, configuration is usually performed using knowledge about a domain and is semantic in nature. Therefore, we argue that configuration has the potential to be used, not only for syntactic processing, but for the semantic processing of natural language, effectively supporting Natural Language Understanding (NLU). In this paper, we propose an approach to NLP that applies configuration to the (partial) domain model evoked by the processing of a sentence. This has the benefit of ensuring the meaning of the sentence is consistent with the existing domain knowledge. Moreover, it allows the dynamic incorporation of domain knowledge in the configuration model as the text is processed. We demonstrate the approach on a business specification based on the Semantics of Business Vocabulary and Rules.