Word and Text relation using python and NLP

Question

I have a word, according to that i want to find out whether the text is related to that word or not using python and nltk is it possible ?

For example I have a word called "phosphorous". I would like to find out that the particular text file is related to this word or not?

I cant use bag of words in nltk as I have only one word and no training data.

Any Suggestions?

Thanks in Advance.

Answer 1

Not without a corpus, no.

Look at it this way: can you, an intelligent being, tell whether 光 is related to 部屋に入った時電気をつけました without asking someone or something that actually knows Japanese (assuming you don't know Japanese; if you do, try with "svjetlo" and "Kad je ušao u sobu, upalio je lampu"). If you can't, how do you expect a computer to do it?

And another experiment - can you, an intelligent being, give me the algorithm by which you can teach a non-english-speaking person that "light" is related to "When he entered the room, he turned on the lamp"? Again, no.

tl;dr: You need training data, unless you significantly restrict the meaning of "related" (to "contains", for example).

Answer 2

You can use the nltk wordnet to calculate path similarity score between the word and words in your other text and estimate a heuristics based on that score:

from nltk.corpus import wordnet as wn hit = wn.synset('hit.v.01') slap = wn.synset('slap.v.01') wn.path_similarity(hit, slap)

You can find more nltk word-net usage examples here: http://www.nltk.org/howto/wordnet.html