Smarter formation of topic patterns #41
Description
At the current time to topics are generated along with patterns created as topi names via the TextMiner class
https://github.com/aigents/aigents-java/blob/master/src/main/java/net/webstructor/data/TextMiner.java
and its underlying clustering implementation
https://github.com/aigents/aigents-java/blob/master/src/main/java/net/webstructor/data/Miner.java
The patterns created this way are just disjunctive sets of words, missing the few things
- conjunctions
- ordered conjunctions ("frames")
- regular expressions (to handle suffixes)
It should be improved with more complex pattern formation involving symbolic pattern regression producing hierarchical patterns like discussed here:
https://www.youtube.com/watch?v=FzKMtNILmDk