LEARNING GRAMMATICAL RULES FROM EXAMPLES USING A CREDIT ASSIGNMENT ALGORITHM

@inproceedings{Bianchi1996LEARNINGGR,
  title={LEARNING GRAMMATICAL RULES FROM EXAMPLES USING A CREDIT ASSIGNMENT ALGORITHM},
  author={Dario Bianchi},
  year={1996}
}
A credit assignment algorithm is proposed for inferring a context free grammar from a set of sample sentences. Only one copy of the grammar is maintained as a population of rules. A strength, associated to each rule, takes in account its usefulness in correctly classify the test sentences. Some results are reported for the language of correctly nested and balanced brackets and for a natural language fragment. In the case of complex grammars the system may be used for grammar refinement… CONTINUE READING

Similar Papers

Loading similar papers…