In natural language processing, precision and recall are used to determine the accuracy of an entity based text analysis.
Say you write an article and then you run WordLift‘s text analysis to detect the entities which are relevant in that exact context. In this case:
- Precision tells you which percentage of entities are detected and classified correctly.
- Recall describes how many of the entities expected were actually detected.
The more a system is precise, the less recall you will have (meaning that you are going to miss some entities), and vice versa the highest the recall is, the less the system is precise. So, since WordLift works on the basis of a named entity recognition system, we need to continuously balance these two factors.