Topic Modeling Tool
A GUI for MALLET's implementation of LDA.
- Metadata integration
- Automatic file segmentation
- Custom CSV delimiters
- Alpha/Beta optimization
- Custom regex tokenization
- Multicore processor support
To start using some of these new features right away, consult the quickstart guide. For tinkerers, there's a guide to the tool's optional settings. You may also find useful information in the discussion threads under documentation issues.
The Topic Modeling Tool now has native Windows and Mac apps, and because of unicode issues, these are currently the best options for installation. Just follow the instructions for your operating system. Do not try to install by clicking on [Clone or download] > [Download ZIP]. It won't work.
- Open it by double-clicking.
- Drag the app into your
Applicationsfolder -- or into any folder at all.
- Run the app by double-clicking.
For Windows PCs:
- NOTE: The native PC build is out-of-date. Help wanted.
- Extract the files into any folder and open it.
- Double-click on the file called
TopicModelingTool.exeto run it.
If you want to run the
you'll need to have a fairly recent version of Java; the version that
came with your computer may not work, especially if your computer is
a Mac. Whatever your operating system, you can install an updated
version of Java by following the instructions for your operating
Windows Unicode Support:
Windows and Java don't play very well together when it comes to unicode
text. If you are using the
.jar build, and non-ascii characters are
getting garbled on a Windows machine, there's a quick fix involving
that may make things work.
Again, the best answer may just be to use the native app. It should now work correctly at every stage with UTF-8-encoded text. (If it doesn't, let us know and we will moan and gnash our teeth some more.)
Reporting and Replicating Bugs and Other Issues:
If you hadn't already guessed, most testing for this tool happens on a Mac. There are bound to be errors happening on other platforms that have slipped through the cracks. We need you to report them so we can keep improving the tool! But we cannot fix a problem that we don't fully understand, so...
When posting a bug report, please include vast amounts of detail.
Copy and paste everything from the tool's console output if you can, tell us
your operating system and version, and let us know the other tools you're
using to create and view input and output. It also helps if you verify that the
bug still exists in the most recent build of the tool (i.e. the one contained in
.zip files in the root directory).
We know that there are substantial problems with Windows support for unicode text; if you see problems, please post detailed information under the main issue so that we can start isolating and fixing these bugs.
We love getting new issues because it means the tool is improving! But again, when posting a bug report, please include vast amounts of detail.
Building the Development Version:
If you feel adventurous, you might want to modify the code and compile your
own version. To do so, you'll need to install Apache Maven
as well as the Java Development Kit. On Macs, Homebrew
is the best way to do so; simply install homebrew as described on the Homebrew
site, and then type
brew install maven at the command line. On Windows PCs --
you're on your own! But we did it and it wasn't terribly hard. You just need an
up-to-date JDK and maven package, with their
bin folders in your
With maven installed, simply use the terminal to navigate to the
$ cd topic-modeling-tool/TopicModelingTool
Then use maven's
$ mvn package
We now have experimental support for compiling the tool as a native app using the javafx plugin for maven. This will build a native package able to run on your operating system. This has been tested on both Macs and Windows PCs.
$ mvn jfx:native
Previous work on the GUI for MALLET has been supported by a National Leadership Grant (LG-06-08-0057-08) from the Institute of Museum and Library Services to Yale University, the University of Michigan, and the University of California, Irvine. The Institute of Museum and Library Services is the primary source of federal support for the nation's 123,000 libraries and 17,500 museums. The Institute's mission is to create strong libraries and museums that connect people to information and ideas.