After discussing the Mallet project together, we came to the conclusion that less topics makes its easier to identify the subject of the topic. And the more iterations, the better relation of words to the topic. However, too little topics can make the topic modeling too broad and too many iterations will take Mallet longer to come up with the topic models. We recommend 50 topics and 1500 iterations will give the best and fastest results for topic models.
House Decor (50 topics, 3000 iterations)- house front large round side led window left garden windows small low standing lawn close houses lane high centre building huge gate iron wooden grounds narrow park trees chamber elderly lined walls doors upper ancient rain oak fashioned leads avenue appeared winding entrance century barred modern evidently brick surrounded plainly
This data was taken most from The Speckled Band and least from Resident Patient.
Question 1: When are home landscapes an important part to Sherlock Stories?
Question 2: Can descriptions of the home factor into the plot of the story?
Feminine (100 topics, 200 iterations) lady young woman wife maid child girl left beautiful ill give ferguson making mistress notice devoted possibly nurse poor frances
This data was taken most from A Scandal in Bohemia and least in Six Napoleons.
Question 1: Do women play a prominent role in Sherlock Stories?
Question 2: What can be inferred about the time period from the words used to describe women?
Murder (30 topics, 2000 iterations) found dead body death crime left murder police close attention
finally struck blood part blow remained knife examination stick tragedy
These words were found most in The Final Problem and The Adventure of the Blue Carbuncle
Question 1: Are all of Sherlock mysteries based off of murder?
Question 2: How graphic is the murder described in the stories?