The classifier currently strips out stop words to improve accuracy. The stop words are hard-coded into the library, which means they will not work with text in other languages.

Allow users to set new stop-words.

I couldn't find a simple way of doing this in the build. Write custom validation in Js to prevent the dist/index.js from growing above a certain size. If this custom validation is good, consider publishing in a separate package.

If your model contains user input it is possible that sensitive user information may make it into the model. If the model is exposed (for example in the browser) it may expose this information.

When creating a new WhichX objection, we should allow a configuration option to hash all words added to the model. This will also require that words are hashed during comparison so that they can effectively be compared.

• Specify new configuration option to allow model hashing.
• Check this configuration before adding to the model.
• Check this configuration before classifying against the model.

Two error criteria that now provide misleading error messages. These new issues are un-release in 3.0.0-SNAPSHOT.

• whichx.addLabels('total') should specify that labels must be unique. It complains about the type, which is incorrect.
• Further, 'total' is not just a duplicate, it is a reserved key-word. The same can be said of other properties of object.

The logic for adding a string label should be extracted to a private function so the error handling logic does not need to be duplicated.

One of the value adds of this classification library is that it is really really small. It's worth taking that one step further and making it as small as it can be.

• Add some kind of minify to the build.
• Ensure the minified version is published for new versions.
• Ensure that the project JSDoc is still displayed in the IDE of a project using whichx as a dependency.
• Add checks to ensure deliverable stays below a certain size to protect against bloat.

Title just about says it all. There are some details that need documenting, for example how to export/import a model or change the stop words for a different language.

• Create a documentation folder.
• Add a faq.md file to the folder.
• Link to the FAQ from the main README.md.

Hello

I was glad to see that I was able to have a good outcome from your library to solve a classification problem at work.
I actually noticed that your library worked better than https://github.com/ttezel/bayes (about 400 stars) with the same training data.
So I would recommend that you make available by a better name your Naive Bayes code, so that more people can find it (and benefit from it) beyond the pets domain.

Congrats!

Any way to do this?