Is there any influence of Language Technologies and Computational Linguistics on linguistics?
Hello,
Since the end of the 70ies we have linguistic theories that are formal enough and worked out in enough detail to be implementable on computers. Examples are Tree Adjoing Grammar, Lexical Functional Grammar, Generalized Phrase Structure Grammar, Head-Driven Phrase Structure Grammar, and Categorial Grammar (actually around since 1935). See my Grammar Theory book for an overview of implementations of all these theories (and also for remarks on Chomskian theories):
https://www.researchgate.net/publication/258328909_Grammatiktheorie
This is a document in English containing the frameworks and the implementations (I know of):
http://hpsg.fu-berlin.de/~stefan/PS/implementations.pdf
If you look at the names of those who worked on the implementations, you will notice that they also play important roles in theoretical linguistics.
Computer implementations are useful to verify the consistency of developed theories. If you have computer implementations you can check your grammar against thousands and thousands of test suite examples (grammatical and ungrammatical ones). After changing a theory one can do a test run and find out quickly that the change was not as brilliant as one first thought.
Implementations that have both syntax and semantics can be used both for analysis and for generation. That is, you can use the grammars to generate utterances for a certain meaning.
Those who work on implementations will notice that their grammars license bizarre word sequences, sequences no theoretical linguist could have thought of. By discovering such sentences one discovers that the grammar is not constrained enough and that further constraints have to be added. This directly feeds back into theoretical linguistics.
Those who work with real data and try to increase the coverage of their linguistically motivated deep grammars will run into phenomena that have been neglected in main stream linguistics, since they are not relevant for the study of Universal Grammar, since they do not contribute `leading ideas' (Chomsky, 1981) or for whatever reasons. Since grammar is a very complex topic and everything is connected, all phenomena interact, it is good to have some help by machines. It is like doping, but legal.
You can find some remarks on complexity of linguistic theories, on computer implementations, and theory verification in a draft of a paper on our CoreGram project:
https://www.researchgate.net/publication/258338696_The_CoreGram_Project_Theoretical_Linguistics_Theory_Development_and_Verification
Comments welcome!
Best wishes
Stefan (Müller)
Book Grammatiktheorie
Article The CoreGram Project: Theoretical Linguistics, Theory Develo...
I agree with Lionel Nicolas' point of view. However, I think that there are at least a series og goals of applied) computational linguistics which have (or could have) a direct influence on language teaching and/or linguistic research. I think of, for instance, corpus linguistics, (computer assisted) lexicography, (computer assisted) terminology, also (computer assisted) language learning, text comprehension/generation, etc.
I have been on both ends of this. Trying to teach a dummy like a computer to handle natural language requires precision in explanation. Therefore, the theoretical benefit is to keep researchers completely explicit in their theories. One cannot teach a computer to kind of do something sort of some way or other. Directions (programs) must be unequivocal.
Hello,
Since the end of the 70ies we have linguistic theories that are formal enough and worked out in enough detail to be implementable on computers. Examples are Tree Adjoing Grammar, Lexical Functional Grammar, Generalized Phrase Structure Grammar, Head-Driven Phrase Structure Grammar, and Categorial Grammar (actually around since 1935). See my Grammar Theory book for an overview of implementations of all these theories (and also for remarks on Chomskian theories):
https://www.researchgate.net/publication/258328909_Grammatiktheorie
This is a document in English containing the frameworks and the implementations (I know of):
http://hpsg.fu-berlin.de/~stefan/PS/implementations.pdf
If you look at the names of those who worked on the implementations, you will notice that they also play important roles in theoretical linguistics.
Computer implementations are useful to verify the consistency of developed theories. If you have computer implementations you can check your grammar against thousands and thousands of test suite examples (grammatical and ungrammatical ones). After changing a theory one can do a test run and find out quickly that the change was not as brilliant as one first thought.
Implementations that have both syntax and semantics can be used both for analysis and for generation. That is, you can use the grammars to generate utterances for a certain meaning.
Those who work on implementations will notice that their grammars license bizarre word sequences, sequences no theoretical linguist could have thought of. By discovering such sentences one discovers that the grammar is not constrained enough and that further constraints have to be added. This directly feeds back into theoretical linguistics.
Those who work with real data and try to increase the coverage of their linguistically motivated deep grammars will run into phenomena that have been neglected in main stream linguistics, since they are not relevant for the study of Universal Grammar, since they do not contribute `leading ideas' (Chomsky, 1981) or for whatever reasons. Since grammar is a very complex topic and everything is connected, all phenomena interact, it is good to have some help by machines. It is like doping, but legal.
You can find some remarks on complexity of linguistic theories, on computer implementations, and theory verification in a draft of a paper on our CoreGram project:
https://www.researchgate.net/publication/258338696_The_CoreGram_Project_Theoretical_Linguistics_Theory_Development_and_Verification
Comments welcome!
Best wishes
Stefan (Müller)
Book Grammatiktheorie
Article The CoreGram Project: Theoretical Linguistics, Theory Develo...
- Badges
- Science topic
- Similar topics
- Linguistics