MediumNLPnlp-167

Aspect Sentiment Pairs

Problem

For each aspect word occurrence, sum the polarities of all lexicon-scored tokens within window positions on either side, inclusive of the aspect itself.

Return [aspect, score] pairs in order of occurrence. Aspects not present in the lexicon still produce a pair with score 0.0.

Hint

Clamp the window range to the sequence boundaries before summing.

Test cases

case 1

in: [["food","was","great"],["food"],{"great":1},2]

expected: [["food",1]]

case 2

in: [["bad","food","bad"],["food"],{"bad":-1},1]

expected: [["food",-2]]

case 3

in: [["food"],["food"],{},3]

expected: [["food",0]]

case 4

in: [["a","food","b","food"],["food"],{"a":1,"b":2},1]

expected: [["food",3],["food",2]]

case 5

in: [["nope"],["food"],{"nope":1},1]

expected: []

Starter codePython

def aspect_sentiment_pairs(tokens, aspects, lexicon, window):
    # Your code here
    pass

5 test cases · run your code against them in the editor