Recall and precision

Recall and precision

I. query accuracy and precision are currently relatively reasonable indicators for measuring the search results.
Query completion rate = (retrieved information volume/total amount of information in the system) * 100%
Precision = (the amount of retrieved information/total amount of retrieved information) * 100%
The former is to measure the ability of the retrieval system and the searcher to detect relevant information, and the latter is to measure the ability of the retrieval system and the searcher to reject non-relevant information. Both indicate the retrieval efficiency.

Ii. Limitations in both the recall and precision rates
The limitations of the full query rate are as follows: it is the ratio of the retrieved information to all the relevant information stored in the retrieval system, however, the amount of information in the system is generally uncertain and can only be estimated. In addition, the query rate is more or less limited by "assumptions, this "hypothesis" means that the retrieved information has the same value to the user, but it is not actually the case. For the user, the degree of relevance of the information is more important than the number of information.
The limitations of precision are mainly manifested in the following: if the search results are question-based, rather than full-text, it is difficult for users to determine whether the retrieved information is closely related to the subject due to the simplicity of the question-based content, the full text of the question must be found to determine whether the information meets the needs of the Search topic. Meanwhile, the information mentioned in the precision ratio also has the limitation of "hypothesis.
Experiments show that there is an opposite dependency between the query-completion rate and the precision-if the output query-completion rate is increased, the precision rate is reduced, and vice versa.
For users, the main factors that affect the search results include the extensiveness of document indexing and the specialization of user search marks.