Section 2 : Semantic interpretation and disambiguation

Commentary

Section Goals

• To introduce the concepts and methods used to represent and manipulate semantics and pragmatics information, generate natural language, and to perform disambiguation in natural language processing.

Learning Objectives

Learning Objective 1

• Discuss semantic interpretation and pragmatic interpretation based on first-order logic knowledge representation.
• Describe the principles underlying language generation with DCGs.
• Explain different kinds of ambiguity, and how to perform disambiguation.
• Explain the following concepts or terms:
  • Compositional semantics
  • Quasi-logical form
  • Semantic interpretation
  • Pragmatic interpretation
  • Indexical
  • Fluent
  • Lexical, syntactic, structural, and semantic ambiguity
  • Disambiguation

Objective Readings

Required readings:

Reading topics:

Semantic Interpretation, Pragmatic Interpretation, Language Generation, Disambiguation (see Sections 23.3 of AIMA3ed)

Pradhan, S. S., Ward, W., and Martin, J. H. (2008). Towards robust semantic role labeling. Computational Linguistics, 34(2).

Objective Questions

• What kinds of semantic information can be interpreted using the methods introduced in this section?
• Why do we say ambiguity is one of the most challenging tasks in NLP?

Objective Activities

• Explore some engineering semantic representation methods, such as FrameNet (https://framenet.icsi.berkeley.edu/fndrupal/ ), VerbNet, and PropBank, to see how they are helpful in semantic interpretation.
• Complete Exercise 23.13 of AIMA3ed.