NLIDB Progress

Going with multiclass multilabel classifier to counteract parsing issues and users unfamiliarity with underlying data structure. Also I can reduce the amount of shit I need to handle in the code. I think the overall time will be reduced significantly as incorrect I/O operations (due to imperfect parsing) takes a lot of time. Can probably also use a matrix comparison if I am going to only use a bag of words approach. Classification ml is only useful if I use crf sort of thing using ngrams or use data analysis based probability based data generation.

Using conceptnet for classification is also new element creation. Conceptnet can be quite useful for classification. Don't think many people are using it.

made good progress in attacher/matcher. will use pandas for storing and final scoring.

Using pandas for final scoring and storing is probably the best way to solve it. The final output is shaping up pretty nicely.