List view
Instead of having only interactive visualizations, use the visualizations as a tool to correct or make suggestions to the topic model interactively. Such features would be the ability to combine two topics, eliminate topics, change the underlying probability distribution of words in a topic, remove documents from the analysis, etc. Ideally, this is accomplished in real time so that the user does not have to wait between corrections.
No due dateUsing the Pioneer Diaries dataset and/or other interesting data sets, perform A/B testing of the following: Visualization Preference - Do people like the force or chord diagram better when examining topics? Visualization Version Preference - Does a certain feature of the force diagram enhance it's utility to users? Exploration Route - What exploration order best helps a user become acquainted with a dataset? Is it helpful to look at topic, then relevant documents, or vice versa? How do word/metadata views aid that experience? Topic Models - Obtain an empirical measure of the correctness of a particular topic model. Does a certain dataset need 50 or 100 topics? How do other models compare to LDA? Which do users prefer? Etc
No due dateThe three main goals of this milestone are 1) To be able to import the pioneer diaries dataset containing over 20,000 documents (splitting on pages) 2) To be able to aggregate documents (their metadata and topical composition) so that the user can look at the dataset with different levels of granularity: Author, diary, chapter, page, paragraph. 3) Provide favorites support including user authentication and favorites support for all visualization views
No due date•8/8 issues closedWe want to get the demo server up and running with no new bugs with all the features we have been working on. We are using a PostgreSql database on an Apache Server.
No due date•29/29 issues closed