By: Mike Mirando & Nadia Sharif

The utilization of a variance in number of topics in mallet methodically effects the results and outcomes out the process. As we observed the data from previous findings, we concluded that if one has a fewer amount of topics, evidence is more clear and broad in certain aspects. The outputs are meticulously  strengthened and bolded as the number of iterations are increased, existing as a positive correlation.  A recommended setting for Mallet would be: 50 topics, 1500 iterations, and 20 topic words.

Money & Worth: business client money england hundred king pounds thousand large set gold photograph paid pay ten give draw fifty ready worth

a) The specific story that utilized this topic the most was, The Blue Carbuncle with 21% of the words assigned to this topic (34 words)

b) The specific story that utilized this topic the least was, The Boscombe Valley Mystery (2 words)

c) 1) Due to the fact that these explicit words are associated with money, can you imply that the entire story circulated around economics?

2) Can Mallet help us determine if all of the Sherlock Holmes are circulated around money?

Family: woman lady wife husband life love left boy child nature loved beautiful maid ferguson happy madam women mistress devoted wonderful

a) The story that this selection of words is mostly found is is, The Sussex Vampire (88 words)

b) The story that this selection of words is mostly found in is, The Red Circle (2 words)

c) 1) Can we determine what characters have been associated with these words from MALLET?

2) Can we comprehend the background of information associated with this selection of words?

Unfortunate Murder: death man poor terrible dead heard told creature happened died night life human met devil great wild fate dreadful killed

a) The most prevalent story that these words are associated with is, The Second Stain (64 words)

b) The least prevalent story that these words are associated with is, The Noble Bachelor (2 words)

c) 1) Can we find the specific event that occurred from MALLET?

2) At which time and point in the story did these words come up most and why?

As we overlooked the results a numerous amount of times, we found these topics to be the most interesting and most evident and specific when it came to comprehending the background information. Methodically understanding the information the topic sets out for you was a highly significant aspect of Mallet and can be proved to be a primary aspect of reading.

This blog was written by Kevin Finer and Megan Doty

While using MALLET it is important to use the right variables in your search. We found that 75 was a good number of topics for MALLET to create. If the number of topics being created is too high than the program is forced to cluster words into topics that won’t provide a realistic sense of the text. In terms of iterations run, the output will only strengthen by increasing that variable but those iterations come at the cost of processing time. We felt that 1,500 iterations was good enough to create reliable topics while not overloading the program.

Now we will examine three similar topics that both of our searches created.

Writing / Notes and Writing
For the topic of Writing from Megan’s search, the words MALLET returned included “paper note read book table wrote writing written handed sheet picked letter write page address pen piece pencil post learn,” with the top ranked document being “The Adventure of the Reigate Squire”. For the segment of the story where the topic was most prevalent, the count came in as making up 41% of the file, with words from the topic appearing 19 times. The story in which it appeared least is “The Adventure of Wisteria Lodge”, with only 1 usage appearing up from multiple parts of the story.

Meanwhile, a similar topic in Kevin’s search provided different story frequency. “The Five Orange Pips” brought a total of 17 text files linked to the topic of writing, ranging from 4 to 24 words. In that largest count the topic provided 27% of the file. Writing occurred the least in “The Yellow Face”, although the use of it is still larger than in Megan’s data set. This is because in Kevin’s search where this topic occurred there were only 30 topics made, thus increasing the frequency of all of them.

Question 1: Where this topic appeared most, what was the writing in pertinence to?

Question 2: In these instances is Holmes typically writing or reading?

Emotional Shock / Reactions upon realization

In a search with 1500 iterations and 75 topics output, the topic pertaining to reactions and emotional shock was most prevalent in the story “The Adventure of the Creeping Man,” with 26 appearances of words from within the topic, followed closely by “The Adventure of the Devil’s Foot,” with 21 words appearing from the segment of text.

In Kevin’s version of this topic “The Speckled Band” used the topic the most, but “The Adventures of the Creeping Man” was probably the second most prevalent story. In this case there was some similarity in the two topics. This particular version of the topic appeared the least in “The Norwood Builder”.

Question 1: What type of event is normally occurring when this topic appears?

Question 2: Could these topics perhaps have overlap with the topics of Murder or Crime?

Time / During the Night

Under the results for this topic, different parts of the story “The Adventure of the Bruce-Partington Plans” came up for the first three results of top ranked documents. Here, with Megan’s search, the results showed more specifically words that were relevant to times during the night. Looking at the list of stories with top usages, the first came in with 17 usages for the section, followed by 14, and then 12. The same story also ranks in for 7 – 12 under the results, which can infer that — without having read the story — much of the action unfolds at night.

In a different search with a similar topic, “The Red-Headed League” was easily the most common with the top two files belonging to that story. In total 27 files belong to that story. “The Mazarin Stone” had few instances of the topic at all, comprising a total of eight words through three files of the story.

Question 1: Where in the story is the time usually mentioned?

Question 2: Is time more likely to be mentioned in stories that are a longer length?

As is evidenced by the differences in the most common stories these topics were found, even when the topic is similar the data for related topics does not always match. The change of even just a few words and settings in the search provides different data. What this means is that MALLET may not give you the full picture (like many of the tools we have used so far). However, it is still useful in drawing certain broad conclusions as well as interpreting data in new ways and considering new approaches to stories and information.

Drawing from the collection of Sherlock Holmes stories, MALLET was used to draw out keywords and sort them into separate topics seen throughout Conan Doyle’s work. From the topics that I’d gotten running MALLET, I’d been able to pick out a handful of them that felt more classifiable than others, many of which seemed to be filler words that, together in a topic, seem a bit nonsensical. After running MALLET a few times, using settings of 75 topics with 1500 iterations I was able to pull out transportation, murder, writing, trails, working in London, during the night, women, describing the dead, finances, Holmes seated, realization, and facial expressions.

Changing the settings, it seemed as though the amount of topics output did not affect the quality of each topic, rather the amount of topics that I could pull from effectively. Which is to say, I might be able to get 5 good topics with a setting of 40 topics output, and 10 good ones with 80 output. Further, the amount of iterations played an important role in coming up with usable results. It seemed as though the more iterations that the program ran through, the more cohesive each topic was. The words per topic seemed to be in the same vain as the number of topics output overall. Running the program, I stuck with 20. Like with topics, from the amount of words put out per each topic, some seemed nonsensical. Though, with less words I might have lost the words that helped connect the topic to a certain subject. With more words, the result for each topic may have become too convoluted to make sense of and categorize.

Particularly interesting about how this project with MALLET was run was the html pages output with the results. Being able to click on and see the words of a topic in context helped immensely with more vague topics which could have been applicable to a number of adjacent topics.

Putting the words from MALLET into Wordle to create a word cloud, I could immediately imagine more interesting ways that the data could be processed. It is intriguing to see commonalities between topics, though it would be increasingly more so to see how they connect overall. Particularly, to see how often they appeared throughout a timeline of Holmes stories — perhaps incorporating density with adjustments pertaining to the length of the work the topic appeared in.

This week’s assignment was actually really fun, I enjoyed experimenting different combinations of Topic Modeling and comparing the results, but the best part for sure was trying to unveil what possible labels I could use for each group of words. When I chose a combination of settings with more than 20 words printed it got tough to label them, I always had the feeling that the label I chose wasn’t good enough for most of the words. Also it is much better to chose less words in the same topic so that it looks more organized and better specified -you can just “jump in” to the part of the story or situation you want-.

When I first played with Malet I used the default settings but I changed the number of interations from 200 to 500, with10 words printed and 10 topics, it was very easy to label them once I had to work with only 10 words in the same topic, and as I said, when you have less words to group you can be even more specific.

For my second attempt I decided to do something a little different and contrasting to what I found to be easier, so I changed the settings to 30 topics, 2000 interations and 50 words printed. Worst decision ever! I had to look through at least 10 topics until I found one that I could relate all the words.

In my last time playing with the tool I decided to use all the settings that I liked better, so I chose 30 topics, 2000 interations and 10 words printed, the results were amazing and really easy to work with, also, the most interesting thing about this attempt is that all my labels were related in a certain way with crime and mystery.

I enjoyed using Mallet and was surprised how extremely fast it did the topic modeling. I’m glad it’s not a slow process; it actually takes longer to set it up and tell it exactly what to do. As for titling my topics, I found some easier to do than others. While some were rather clear in what they could be titled, some were more difficult and I ended up using a word that was already included in my list for some, such as Work. I also found myself determined to make my titles only one word, not realizing that two or even three words would do just fine as well. I found two lists that were very similar to each other (Features and Appearance) as I thought some of the words in the two lists could be interchangeable. I certainly expected the popular topics such as Crime and Murder to show up. A lot of my topics are also related to one another, such as Features and Appearance, Crime and Murder, and Communication and Literature. In my Communication topic, the word “tregennis” appeared and I had no idea what that was. It turns out to be a character’s name in the short story titled “The Adventure of the Devil’s Foot.”  This just goes to show that additional research is always necessary, no matter what academic tool is used. As I played with different iterations and topic numbers, I noticed that the higher the number, the more variety in words included in lists. However, too many topics/words may be hard too hard to analyze, therefore making the whole process of titling the topics a strenuous task. I like Mallet as a tool for distant reading, which is a concept that I think is definitely useful. It’s kind of like a more organized word cloud in a way, one that groups words together instead of just gathering them. As someone who isn’t very fond of reading, analyzing texts, especially this many at one time, is way more enjoyable for me.