Sophie (soph) wrote in compling,

Building an AI dictionary?


This is going to be a big post, and I apologise for that. tl;dr version: I've always had a dream to build a huge dictionary for AI use, but I'm not sure where to start in actually getting one going.

I've always had a dream but to work on a giant database detailing, in a computer-readable fashion, words and their relationship to one another. Pretty much every word in the database would be described by its relationship to other words, rather than relying on knowledge that the computer already has - in other words, the AI equivalent of a dictionary.

Of course, it can't be entirely self-complete; there will need to be concepts that the AI knows about natively, and for the basic words describing those there could be a link there. But for the most part I'd love the whole thing to be self-describing.

The idea is to build something that the AIs of the future can use, and I believe that with such a dictionary, we would be able to have some extremely realistic AIs, since it would be able to not only generate its own sentences just by 'thinking' about whast it wants to communicate (and finding precisely the right word for it), but it would also be able to parse other people's sentences and have a good idea what they really *mean*.

Of course, I don't expect such a project to ever be finished or even nearly finished within my lifetime, even if (as I was hoping) there were multiple people involved. There'd need to be some way to pass the poject down to future generations.

Anyway, yeah, that's always been my dream, and I really want to do something about it, but I'm not sure where to start. Does anyone have any ideas?
  • Post a new comment


    default userpic
    When you submit the form an invisible reCAPTCHA check will be performed.
    You must follow the Privacy Policy and Google Terms of use.
>I've always had a dream to build a huge dictionary for AI use, but I'm not sure where to start in actually getting one going.

Give it an Internet access :)
Ah, and access to dictionaries like Oxford thesaurus. Or some defining dictionary. There are more than 100 000 words in English, though, to read a dictionary one may need no less than 2000 the most frequent English word.

To start with, 500 may be enough.

Then, initial baby's vocabulary is gained...perhaps, by the process of association. Something similar to OCR perhaps. So, it seems to be pretty simple at first sight.

I have a sort of a "project" like that, to try to emulate some sort of a linguistic understanding without even providing a proper language model. I seem to be free now and i have to give it a try, though i'm never a coder, just a linguist. Not even that experienced.

Ah, this community seem to be dead, by the way.
Do you know about WordNet?