Advantages of a Transformational Grammar for Question Answering

Abstract

A number of researchers in ar t i f ic ia l intel l igence, for example, Woods(1975, p.88 f f.), have asserted that trans-format ional grammars are not a satisfactory basis on which to construct natural language understanding systems, pr imari ly because of eff ic iency considerations. The evidence for such a claim is by no means strong, Pe t r i ck (1976) , and it can be argued that transfer of new theoretical insights into a lan-guage understanding system based on transformational gram-mar is fac i l i tated, Plath ( 1973). This note shows that a trans-format iona l parser can also s impl i fy problems of relat ing canonical representations of queries to data base representa-tions. Consider a data base consisting of a set of company names each wi th an associated list of employees. A natural question for such a data base is M How many people does company Y employ? " Our grammar produces an underlying tree structure whose bracketted terminal string is something l ike ( I) , f rom which a Knuth-style semantic interpreter prod-uces a LISP form like (2).

Cite

Text

Damerau. "Advantages of a Transformational Grammar for Question Answering." International Joint Conference on Artificial Intelligence, 1977.

Markdown

[Damerau. "Advantages of a Transformational Grammar for Question Answering." International Joint Conference on Artificial Intelligence, 1977.](https://mlanthology.org/ijcai/1977/damerau1977ijcai-advantages/)

BibTeX

@inproceedings{damerau1977ijcai-advantages,
  title     = {{Advantages of a Transformational Grammar for Question Answering}},
  author    = {Damerau, Fred},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year      = {1977},
  pages     = {192},
  url       = {https://mlanthology.org/ijcai/1977/damerau1977ijcai-advantages/}
}