Social Modeler 0.3 Is Now Out

Version 0.3 is now officially out. Here is the Windows bundle with R. All other operatings systems can download the standard bundle.

The most significant addition pertains to metadata. You can now include additional data for each document in a model and Social Modeler will incorporate these data into the topic model. For example, each document can have an author associated with it (e.g., Facebook friend). With the new metadata features of 0.3 you can see which authors are posting about particular topics. Another example concerns news data. If you collect region information for news posts, you can use Social Modeler to determine where particular topic-related events are occuring. In the Social Modeler UI, you can see metadata in two forms. First, you can view top metadata values for a topic. Secondly, you can view a topic trend for each metadata value in the time series panel.

The other big addition is a Facebook scraper. It allows you to scrape either a known user's home feed or do a public search and scrape the results. The data are imported in Social Modeler's files format (described in the README), therefore a model can be built immediately after import. Furthermore, if you choose to import into an existing dataset, the scraper looks at existing FB objects and only scrapes and imports new ones.

Stay tuned for a YouTube tutorial demonstrating the new features.

Release 0.2 Is Now Out, Example Data Set Available

I have officially released version 0.2! I wanted to get this version out before I finished the TUTORIAL (which will be up soon). You can download the Windows bundle with R here. For other OSs, you can get the bundle without R here.

I have added validation for both preprocessing and model building. For both, the entire 'form' is validated when the 'Start' Button is pressed, and any errors are shown in the console. I have also added a 'Browse' dialog for all directory fields. Finally, crude interrupts have been added that attempt to stop the preprocessing or build threads when the 'Stop' button is pressed for either.

To lead into the tutorial, I recently added a download for an example data set, which you can get here. It is a web scrape of the AfghanistanNewsCenter.com archives from 2012 through February of 2013. There is README included that walks through how to load, preprocess, and build a model with the data.

Release 0.1 Now Out

I have officially released version 0.1! You can download the Windows bundle with R here. For other OSs, you can get the bundle without R here. The user interface is working fairly well, although there is still a bit of more work to do. Most noticeably, the stop buttons on the build panel do not work. I will soon implement a method for interrupting the preprocessing and build threads, which has been on the back burner until now. On the explore panel, the only table that sorts is the topics table. However, I realize there is likely a need to sort the features and, more likely, documents table by other things (notably, date for documents).

I am working on a TUTORIAL for Social Modeler. I feel that this will best showcase what it can do and provide users a way to quickly get used to the tool so they can make the most of it. Stay tuned, people!