CODEWITHSUNDEEP
pythonlistnlpspacycosine-similarity
venkatttaknev
venkatttaknev

Reputation: 667

Representing word similarity between words from two lists in a homogeneous way in python

I am trying an NLP technique to see similarity between words from two lists.

The code is as below

import en_core_web_sm
nlp = en_core_web_sm.load()

Listalpha = ['Apple', 'Grapes', 'Mango', 'Fig','Orange']

ListBeta = ['Carrot', 'Mango', 'Tomato', 'Potato', 'Lemon']

list_n =" ".join(ListBeta)     
doc = nlp(list_n)   


list_str = " ".join(Listalpha)    
doc2 = nlp(list_str)

newlist = []

for token1 in doc:
    for token2 in doc2:
        newlist.append((token1.text, token2.text,token1.similarity(token2)))


words_most_similar = sorted(newlist, key=lambda x: x[2], reverse=True)
print(words_most_similar)

I get the following output

[('Mango', 'Mango', 1.0), ('Potato', 'Mango', 0.71168435), ('Lemon', 'Orange', 0.70560765), ('Carrot', 'Mango', 0.670182), ('Tomato', 'Mango', 0.6513121), ('Potato', 'Fig', 0.6306212), ('Tomato', 'Fig', 0.61672616), ('Carrot', 'Apple', 0.6077532), ('Lemon', 'Mango', 0.5978425), ('Mango', 'Fig', 0.5930651), ('Mango', 'Orange', 0.5529714), ('Potato', 'Apple', 0.5516073), ('Potato', 'Orange', 0.5486618), ('Lemon', 'Fig', 0.50294644), ('Mango', 'Apple', 0.48833746), ('Tomato', 'Orange', 0.44175738), ('Mango', 'Grapes', 0.42697987), ('Lemon', 'Apple', 0.42477235), ('Carrot', 'Fig', 0.3984716), ('Carrot', 'Grapes', 0.3944748), ('Potato', 'Grapes', 0.3860814), ('Tomato', 'Apple', 0.38342345), ('Carrot', 'Orange', 0.38251868), ('Tomato', 'Grapes', 0.3763761), ('Lemon', 'Grapes', 0.28998604)]

How do I get an output in the format as below

[('Mango','Mango',1.0),('Mango', 'Fig', 0.5930651), ('Mango', 'Orange', 0.5529714),('Mango', 'Apple', 0.48833746),('Mango', 'Grapes', 0.42697987),('Carrot', 'Mango', 0.670182),('Carrot', 'Apple', 0.6077532)....]

Basically I want the mapping of the form (word in ListBeta, word in Listalpha, cosine score) and it should be uniform and not at random as I have currently. Also it needs to be in descending order of cosine value as depicted above.

Upvotes: 0

Views: 193

Answers (1)

Slam
Slam

Reputation: 8582

If it's indeed question of sorting results, you can use tuples as key result in sorted, i.e. your lambda could return tuple/list, and python will sort on it element-wise.

words_most_similar = sorted(newlist, key=lambda t: (t[0], -t[2]))

Upvotes: 2

Related Questions