RK-A

In MR https://gitlab.com/kordale/rk-ai-3/-/merge_requests/839 & https://gitlab.com/kordale/rk-ai-3/-/merge_requests/843, we worked on a problem of matching KPs with "-" (hyphen) with header variants that do not have hyphens. Just came across this issue in the "Numpy User Guide" book where a KP is being matched perfectly but the scores are below 0.5.

KP = "c-order"
Header Variant = "c order"

We need to investigate the reason for the low scores and fix it if possible.

Steps of investigation:
1) Identify where the score is reducing (BR3_IR3_tagger.py or tagging_utils.py)
2) Identify the reason for reduction (CW filter/POS tags/context matching)
3) Brainstorm the solution.

This ticket is primarily for steps (1) & (2).