如何在 Python 中将单词分类？

Question

I work on a project where I use Google vision to detect objects in images. The API returns a list of labels. So I have multiple words and I would like to put each word into a category. For example:

Google cloud vision returns:

['Head', 'Lamp', 'Eye', 'Green', 'Arm', 'Piano', 'Mobile phone', 'Blue', 'Toy']

And I would like to have something like:

{'Object' : ['Lamp', 'Piano', 'Mobile phone', 'Toy'],
'Color' : ['Green', 'Blue'],
'Body parts': ['Head', 'Eye', 'Arm']
}

I know that word2vec have something called similarity but it means that I have to train a model. Is there any pretained model I can use? Or maybe another solution to do this?

Answer 1

Check out WordNet , a free research lexicon which models word-relationships, and could thus help you group those labels in a variety of ways:

https://wordnet.princeton.edu/

You could try to use word-vectors, as well, to model degrees-of-similarity between words - and thus also potentially cluster related words. And, there are off-the-shelf sets of word-vectors in various languages you could try using rather than training your own.

However, the similarity reflected by such sets may or may not be what you want for your purposes. For example, antonyms like 'hot' and 'cold' are typically very 'similar' in most word-vector models, as they concern the same aspect of something and are used in similar contexts. And logical hierarchies – such as words being a more-specific example of others – won't necessarily be clear in word-vector spaces. (WordNet, as a manually-curated dataset, captures such hypernym/hyponym relationships explicitly.)

Answer 2

Try this hope it fits the criteria:

#This is the list of things you want sorted
lists = ['Head', 'Lamp', 'Eye', 'Green', 'Arm', 'Piano', 'Mobile phone', 'Blue', 'Toy']
#This is all the body parts this program searches for
body = ["Head","Eye","Arm"]
body2= []
#All the objects it searches for
object = ["Lamp","Piano","Mobile phone","Toy"]
object2 = []
#All the colors it search for
color = ["Blue","Green"]
color2 = []
#formats the dict according to your criteria 
dict = {"Object":None, "Color":None,"Body":None}
for i in lists:
    if i in object:
        if i in object2:
            continue
        object2.append(i)
        dict["Object"] = object2
    elif i in color:
        if i in color2:
            continue
        color2.append(i)
        dict["Color"] = color2
    elif i in body:
        if i in body2:
            continue
        body2.append(i)
        dict["Body"] = body2
    else:
          pass
print(dict)

Test it and see if it fits all your criteria