0

I am looking at teaching a computer about grammar which could turn into teaching it sentence analysis and eventually how to formulate a response.

This is the baby steps. I am defining a noun. So far I define a noun as a word that can be in singular or plural. But a noun also have a gender.

Since I am teaching it English, neuter probably won't matter much. But I was just thinking:

How big of an impact does gender really have on how you construct a sentence? Not being a native speaker (It's my 2nd language but I am very fluent in it) I actually don't know if my program needs to know about genders to know Nouns.

Some clearance would be nice!

4

As you know, English does not have grammatical gender, but only natural gender.

Natural gender may apply in different languages to humans only, humans and "higher" animals etc, it may or may not include pronouns. English has all those kinds of natural gender.

(English also has a quirk where some things such as ships may be referred to with female pronouns but this is not grammatical gender).

The concept of neuter doesn't have much meaning in English other than contrasting the pronoun it with he, she, him, and her.

What you will find when implementing a human language on a computer is that you will have to deal with semantics and context as well as just syntax. Treating gender as part of syntax will lead you to problems but if you want to match pronouns in English then you will need to have some kind of semantic or contextual awareness too.

What I find is that when adding such semantics is that other factors then spring up, some of which are more important than gender, and usually they have semantic counterparts in some languages. Chief amongst these is animacy.

Some verbs have different meanings if the subject is animate. Sometimes just figuring which noun phrases are subject, direct object, and indirect object requires you to know which are animate.

Sometimes you may even want to distinguish between human and animal.

So my answer is, for English, either leave out gender if you want something simple, or if you want something complex, then don't just implement gender but also either animacy or some kind of semantics, ontology, "real-world" knowledge, etc.

6
  • So in my definition of a noun (as is this case, I'll get to verbs and adjectives at some point) you'd suggest animacy and semantics? Or semantics overall? – OmniOwl Jun 22 '13 at 1:11
  • Well you don't need a lexical category of gender for each noun in your lexicon but in NLP it can make sense to include semantic categories in your lexicon. Once you start trying to do semantics you find you can go infinitely deeper and will keep finding problems the current depth you have implemented can't handle. But with no semantics at all your results might not satisfy you. If you don't know much about the problem domain, trial and error may be the best way to learn. You can try reading CS/NLP textbooks but you may find a steep learning curve, jargon, prerequisite knowledge, etc. – hippietrail Jun 22 '13 at 1:16
  • Well I know what semantics Can be. But not entirely sure how I'd implement it. It would be pretty cool if the program could make it's own semantics given data analysis. But that's probably too advanced for me at the current stage. – OmniOwl Jun 22 '13 at 1:19
  • 1
    NLP is a notoriously difficult domain. CS less so, but some of the techniques in CS require very advanced (and very interesting) algorithms and data structures. I for one cannot read the math that inevitably is included in papers on CS techniques, which proves to be a barrier to exploring my fascination with them. One problem you will come up against in every direction is ambiguity. It's everywhere in natural language, it can't be avoided, and dealing with it is hard. – hippietrail Jun 22 '13 at 1:26
  • My answer to it would be to let the user specify ambiguity. Let the machine learn from the user. – OmniOwl Jun 22 '13 at 9:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.