A simple web app that takes a text file upload, splits it by your own regexp, and uses mallet to perform topic analysis on it. At present, does so completely statelessly (nothing is saved anywhere).
For development you only need Leiningen installed (plus Java). Everything else will be downloaded automatically.
To start a web server for the app, execute lein run. You can connect your REPL on port 7000, and your browser on port 3000. To start front-end development, follow up with lein figwheel and watch your changes appear in real time without browser refresh.
Produce a .jar with lein uberjar and run with java -jar target/maladroit.jar, then route incoming traffic to the appropriate port. You may need to run lein clean prior to lein uberjar. For production you’ll want some kind of upstart, systemd, other boot process to run the java -jar command.
Even better, if you have a Wildfly server available, just do lein clean; lein uberwar, and then upload target/maladroit.war to the appropriate wildfly directory on your server. Wildfly will keep it running and tuned, and will automatically restart when a changed version is swapped in.
- Mind that spaces are meaningful in your split regexp
- You can add your own stop words from the web interface
- The “passes” parameter on the web interface has the most immediate implications for how long it takes to process your document. Play with different values to change the granularity of your topic modeling.
- [ ] Add visualization, a la d3js