Topic Modeling: Sherlock Holmes

Looking at the topics I selected earlier this week I could notice that even if in some cases there are words that are repeated or some words that shouldn’t be on the topic, the topic tool is really effective and give us a good general idea.

For example if you look at the topic that I named Mr. Watson:

watson case find point friend doubt sherlock interest facts remarkable fact singular remarked account dear strange present points curious reason

You can link most of the words with the character. For example, he is Shelock Holmes’ friend, interested in the facts of the case, he points things that Sherlock says and sometimes has doubt about it.

Some other topics for example can show us some steps of solving a case:

case point facts points fact obvious interest explanation investigation mystery simple confess theory present admit solution formed true problem connection

First of all you have the case and then you have to point the facts and even take in account the obvious ones.In the end of it you need to have and explanation for the mistery that the case is. You can start by having a theory of what happened or by investigating the victim’s connections or maybe follow a clue of a confession to get to the true story. It doens’t matter your choice of technique but you need to solve the problem.

Obviously some topics can be a little messy and don’t mean anything at all but this tool is really helpfull. I tried to set as many iterations as I could and it was really fun seeing the result.