Inference-based approach to NLU, in Ovchinnikova's terminology, makes use of logical deduction on a knowledge base to support parsing, for example reference resolution, interpretation of noun compounds and prepositional phrases, resolution of some kinds of syntactic and semantic ambiguity, metonymy resolution.
According to Bos (2009), some obstacles to inference-base NLU are:
- only few researchers feel comfortable in the intersection of these areas, and collaborations between researchers of these fields are rare
- inference-based approaches are still computationally expensive, which prevents their efficient application to large amounts of data
- there are still open research questions and problems in these approaches, which make showing a good performance on “real-life” text data difficult.
Deduction (automated reasoning)
- Main article: Deduction
- Main article: Abduction
An overview of early approaches: Winograd (1972).
- Tutorial on Natural Language Understanding with World Knowledge and Inference (KR 2014, Summer of Logic, Vienna)
- Ovchinnikova, Ekaterina. Integration of world knowledge for natural language understanding. Vol. 3. Springer, 2012.
- Bos, J. (2009). Applying automated deduction to natural language understanding, Journal of Applied Logic 7, 1, pp. 100–112.
- Winograd, T. (1972). Understanding Natural Language (Academic Press, Inc., Orlando, FL, USA).