Negation detection in sentiment analysis

Question

This basic function to detect negation while performing sentiment analysis. So that not good kind of words can be considered a negative.

This code is working fine. Can someone help to improve it or find if any bug present?

def negate_sequence(self,text):
    """
    Detects negations and transforms negated words into "not_" form.
    """
    negation = False
    delims = "?.,!:;"
    result = []
#Here rather then applying split, we can directly feed our extracted symptoms list
    words = text.split()
    prev = None
    pprev = None
    for word in words:
        # stripped = word.strip(delchars)
        stripped = word.strip(delims).lower()
        negated = "not_" + stripped if negation else stripped
        result.append(negated)
        if prev:
            bigram = prev + " " + negated
            result.append(bigram)
            if pprev:
                trigram = pprev + " " + bigram
                result.append(trigram)
            pprev = prev
        prev = negated

        if any(neg in word for neg in ["not", "n't", "no"]):
            negation = not negation

        if any(c in word for c in delims):
            negation = False

    return result

Answer 1

Here are a few suggestions on how to improve the code:

State the intent

What is the purpose of the function? What is the input, what is the output? In human words, what does the algorithm do? A docstring answering those basic questions would be helpful. A series of concise unit tests would be great.

I ran the function to have a look at the output:

>>> text = "The weather is not good."
>>> result = negate_sequence(self=None, text=text)
>>> print(result)

['the', 'weather', 'the weather', 'is', 'weather is', 'the weather is', 'not', 'is not',
'weather is not', 'not_good', 'not not_good', 'is not not_good']

This doesn't ring a bell with me, so I stopped trying to understand the purpose.

Avoid stateful loops

Iteration i is coupled to iteration i-1 by the negation variable, this makes the logic hard to understand and error prone. If you work on bigrams/trigrams, I'd create a list of bigrams/trigrams and iterate over the tuples. This decouples the iterations.

Breakup long functions

This has almost endless benefits, as a starting point see this article. Some possibilities:

• Have the text broken up into all lowercase and without punctuation by extract_words(text)
• Have the list of trigrams created by make_trigrams(words)
• Inspect the trigrams by process(trigrams)
• If needed, have some kind of aggregate(results)

Once this is done, I guess we are much better to prepared to identify bugs and to further improve functionality.

Answer 2

I can give a suggestion.

Can you just make it more useful by replacing it with real autonyms? That would be way more useful than making it negative without parsing up front.