Natural Language Generation

1
Natural Language Generation

• Background information
• Tim Havinga

2
Overview of the presentation

• What is NLG?
• What stages are there in NLG?

3
What is NLG?

• Natural Language Generation
• Definition?
• NLG is linguistic manipulation of data
• NLG manipulates (linguistically) deeper
  information to produce shallower information
• Where does it begin?

4
Where does NLG begin?

• 1st definition by McDonald (1993)
• the first point where a speaker must appeal to
  her knowledge of language as she begins the
  process of carrying out some action through the
  use of language
• General model by Reiter and Dale (2000)
• ltk, c, u, dgt
• Prior knowledge
• Knowledge Acquisition (KA)

5
ltk, c, u, dgt

• Weatherreporter example
• k database with meteorological data values
• c comm. goal (summiriseMonth(m))
• Conclusions
• Map non-linguistic data to linguistic data
• Summarize a month
• Produce a text to communicate the result

6
Prior Knowledge

• Domain Knowledge
• Communication Knowledge
• Domain Communication Knowledge

7
Knowledge Acquisition (KA)

• Acquiring Correct Knowledge for NLG (Reiter and
  Robertson)
• Acquisition methods
• Directly Asking Experts for Knowledge
• Structured Expert-Oriented KA
• Think-Aloud Protocols
• Corpus Analysis
• Expert Revision

8
Table of findings regarding different KA
approaches
9
What stages are there in NLG?

• Proposed (Reiter and Dale, 2000)
• Document Planning
• Microplanning
• Surface Realization

10
(No Transcript)