Table 3.
Comparison on tokens (NLP libraries vs. Manual benchmark)
Feature request | ReadMe file | JavaDoc | |||||||
---|---|---|---|---|---|---|---|---|---|
Identical | All | ACC | Identical | All | ACC | Identical | All | ACC | |
NLTK | 2484 | 2545 | 98% | 3154 | 3238 | 98% | 4209 | 4256 | 99% |
spaCy | 2453 | 2580 | 96% | 3113 | 3335 | 95% | 4188 | 4348 | 97% |
Stanford CoreNLP | 2478 | 2529 | 98% | 3170 | 3243 | 98% | 4222 | 4245 | 99% |
OpenNLP | 2382 | 2535 | 94% | 3080 | 3218 | 96% | 4105 | 4221 | 97% |
Combined method | 2496 | 2543 | 99% | 3171 | 3215 | 99% | 4219 | 4243 | 99% |
Overall improvement | 1% | 1% | 0% |