-1

I'm studying NLP and as example I'm trying to identify what feelings are in customer feedback in the online course platform.

I was able to identify the feelings of the students with only simple sentence, such as "The course is very nice, I learned a lot from it", "The teaching platform is complete and I really enjoy using it", "I could have more courses related to marine biology", and so on.

My doubt is how to correctly identify the various sentiments in one sentence or in several sentences. For example:

  1. A sentiment per sentence:

    • "The course is very good! it could be cool to create a section of questions on the site."

  2. More than one sentiment per sentence:

    • "The course is very good, but the site is not."

  3. Involving both:

    • "The course is very good, but the teaching platform is very slow. There could be more tasks and examples in the courses, interaction by video or microphone on the forum, for example."

I thought of splitting text in sentences, but it is not so good for the example 2.

Kadu
  • 343
  • 7
  • 17
  • Welcome to StackOverflow. Please read and follow the posting guidelines in the help documentation, as suggested when you created this account. [On topic](http://stackoverflow.com/help/on-topic), [how to ask](http://stackoverflow.com/help/how-to-ask), and [... the perfect question](https://codeblog.jonskeet.uk/2010/08/29/writing-the-perfect-question/) apply here. StackOverflow is not a design, coding, research, or tutorial resource. However, if you follow whatever resources you find on line, make an honest coding attempt, and run into a problem, you'd have a good example to post. – Prune Oct 29 '18 at 21:39
  • 1
    For starting, you can separate the data according to the the labels (plataform, course, forum etc) and train different models on all of them. – Vivek Kumar Oct 30 '18 at 11:41

2 Answers2

1

You can think that comas, other punctuation marcs and some conjunctions and prepositions actually split sentences. This actually goes beyond code into the field of linguistics as they sometimes, but not always, separate sentences.

In the 2nd case you actually have two sentences: "The course is very good" -, but- "The site is not [very good]".

I believe there are NPL packages that can split sentences (Probably by knowing that most sentences follow the subject/predicate/object structure, so if you wind more than one verb then propbably you'll find the same ammount of sentences) and you could use those to parse your text first. Take a look for libraries doing that for your language of choice.

epsilonmajorquezero
  • 559
  • 4
  • 19
0

There is a lib specific for multi-label classification:

scikit-multilearn

when you train your model you have to split classes into binary columns.

Kadu
  • 343
  • 7
  • 17