I followed your tutorial series on nltk and loved the explanation. After going through it, I stumbled upon a scenario where-in unlike sentiment analysis where in you know its going to be + or -, what if you screened through a number of text and had to give a Title to it (something like google news) The complexity of this problem is highlighted by the fact that you might train your model for say 200 different categories of titles but what if you get a new one which should not be bucketed into an existing one but should rather be a completely new addition to the skel.
Help me out if am overthinking a rather simple solution.
Cheers! Ravish
You must be logged in to post. Please login or register an account.
Generally, when doing text analysis, I come up with the categories myself manually, and then search based on those categories. Having something that dynamically searched all content, and just automatically created categories for you would definitely be a research problem.
-Harrison 8 years ago
You must be logged in to post. Please login or register an account.