• add n-gram word model to dictionary, so that each entry would have not a single word, but multiple ones, following each other

• for starters add a vector to stretch or ignore the values of each text Vector
• later add a Matrix or other art of distortions

Compare given vector to each vector in vector cloud and calculate the closes vector.

(alternative) - create n-dimentional figure using points in the vector cloud, and calculate the closest distance to the edge of the vector clouds

Have the ability to compare the prior AI state to the state after adding new data points to cloud and be able to judge the result or improve it

Compare vector clouds and find the most distinctive dictionary entries and ignore less impactful ones

The Code Base is growing and we may want to use different front-ends and back-ends, thus it seems like a good point to make a clear separation in the structure and make the API-part a library.

Allow the User to save their models to a file and load them back up.
Suggestion to differentiate between two formats:

• Compact Format (only contains resource reference and type, i.e.: web:https://...) (disadvantage: high start-up times for models and potentially requires Internet access)
• Full Format - contains the fully parsed model (disadvantage: huge file-sizes for models)

I think giving the user the choice here would be the smartest and then save one as a model.wts (What's This Sources) and the other as model.wtm (What's This Model)

Implement a Word Dictionary that acts as a look-up table for Word Vectors.

• change unknown words to be depicted as <UNK>
• change typical parts to be depicted uniformly (<IMG>, <EMAIL>, <NUMBER>, <FORMULA>,...)

Text Data Type Fields:

• Vector
• coefficient (combine unknown ascii signs, amount of modified characters, images)

The internals of the event management need to be reworked.
At the moment it's a heavily observer-based system, in the future it will work through annotations, events and listeners.

Add Unit Tests for all methods within the .nlp & .math packages

Create methods to allow 'text-to-vector' conversion based on a dynamically growing dictionary.

(alternative) - compress sentences to isolate the sense of each one

Create a (Swing) UI for the Application.