How to make sense of notes made by executives - Text Mining or Sentimental Analysis



Hi Vidhyarthees

I am working on a sentimental analysis of the data given by a company whose executives use a CRM tool to make notes to write about the visits to various clients.

I used an algorithm where I summed the count of each of the negative (-1) and positive words (1) based on English dictionary. Based on the resultant figure, I termed a particular note as negative (sum < 0) or positive (sum > 0).

What other approaches I could use?



Hi @er.lokeshsharma08

You are using 1-gram bag of words. This kind of models generally fail to catch sarcasm and instances with navigation. Like having “not good” in your text will be classified as positive but that’s negative. To avoid it, you can use multi-gram where “not” goes along with “good” and its classified as negative.

Also, to extend this, you can do opinion mining and create something like word clouds with the most frequent words. This will provide you the context of most frequently talked about things.



Thanks, Saurav. Yes, I did try generating a word cloud but it didnt make much sense. The business wants to know the sentiment behind each comment. Will try multi-gram approach.


@er.lokeshsharma08 Just a small correction to saurav’s answer, its “n-gram” approach and not multi-gram.

You can find a small implementation of n-gram in this article


Thanks @jalFaizy