Multiclass TextClassification - sentence structure

[tombburnell]

Is there somewhere that explains clearly the process used to determine create the vector for each sentence?

• Is there any preprocessing is applied? Eg are stop words removed?
• Are the sentences treated as just a bag of words?
• Is the structure of the sentence considered?
• is TFIDF applied to apply weighting to words prior to vectorisation?
• Are distances between "important" words considered?

Is any of this customisable? Is the expectation that we pre-process and manipulate our corpuses first?

I would be particularly interested in mechanisms that could be applied to determine the nature of text. For example 3 documents may be about the same subject matter - but one is a factual, one is fictional and another is opinion. I think this would require the structure of the sentence to be considered as well as the words themselves.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.