CREU Blog 2017 - 2018 Research of JN Matthews

Further Transcript Keyword Analysis

I’ve been continuing to verify participant data with team and help organize volunteer participants.

In addition I’m further analyzing the result of a study preformed last semester. It appears that in the prior study, many of the most common words in the transcripts are articles, conjunctive phrases, and pronouns (both personal and not). For example, the top 6 are “the, like, and, that, to, i”. This felt like a fairly typical breakdown of English word frequency, so in order to reduce that noise, the top 100 English words were filtered from the list before generating the results.

This produced results were the top words more apparently aligned with the study. (see. “um, pieces, uumm, sort, ok, think”). However, we still see a lot of hesitation markers, simple verbs, and adjectives that can’t give us cognitive content with out their context. Consequentially we decided to focus first on recognizing top nouns (“pieces, guess, baxter, tool, wheels”), specifically key ones of interest, as they provide a more concrete self-contained semantic content.

As such I’m working to modify the analysis code I had written to filter for nouns and have been researching libraries that may help with that. So far the NLTK module looks promising.

Concussed

Week Ending 29 October

I got a concussion last weekend. As a result I was unable to get much work done this week, since I had difficulty concentrating and wasn’t have to focus on screens.

Transcirpt Analysis

Over the past week I worked with volunteers to guide them through their participation in the study, as well working with the team to continue to facilitate the study.

We also now have transcripts from one a previously preformed study. I have been working on developing python code to analysis them, so that we can decide what type of content to look at in this and future studies. In the next week I plan to improve my code a bit and use it to collect data from all of the transcripts.

Data Verification And Application Search

Week Ending 15 October

I’ve continued to help facilitate scheduling volunteers. In addition as a team we have begun to go through the data we have collected and manually verify that it is sound before using it as training data.

I’m still looking for more applications, although I have also downloaded the tools found last time to start getting a feel for what their capabilities are. They have proven tricky to install, so hopefully I will have more information about that in the next couple days.

Note: My computer crashed for a few days last week. It has now been restored and is up and running again. And I hope to make up for that lost time this week.

Working With Participants

Week Ending 8 October

I’ve continued to help schedule volunteers, as well has to guide them through their participation in the study. We introduce them to basic premise of the study and make sure remind them of their right to remove consent if they wish. Participants are then shown the FACES software. I show them how the different feature work, and how to use the various selection menus, and then observe their sketch as the create it so that I can offer suggestions for how to make it more accurate.