Precision and Recall
1 like1,994 views
The document discusses precision and recall, which are measures used to evaluate search engines. Precision measures the percentage of retrieved documents that are relevant, while recall measures the percentage of relevant documents that are retrieved. The document provides examples of precision and recall calculations and discusses how to improve precision by finding less irrelevant results and improve recall by finding more relevant results.
Precision and Recall
- 1. Recall Precision M. Sc. Johannes Schildgen 2016-06-09 schildgen@cs.uni-kl.de and
- 4. Hariboogle Search Color-Rado 42 Results in 0,002 Seconds ? ?
- 9. Relevant Found 5 141 Precision: |𝑹∩𝑭| |𝑭| 4246 = 𝟒𝟏 𝟒𝟐 = 0,98 = 98% „How many % of the found items are relevant?“
- 12. Hariboogle Suche
- 13. Precision: |𝑹∩𝑭| |𝑭| = 1 Perfect! No rubbish Only a few results 1 result
- 14. Improve Precision • Find less rubbish (=> Relevance Feedback) Precision: |𝑹∩𝑭| |𝑭| Hariboogle Search Color-Rado 42 Results in 0,002 Seconds
- 15. Improve Precision • Find less rubbish (=> Relevance Feedback) Precision: |𝑹∩𝑭| |𝑭| Hariboogle Search Color-Rado 42 Results in 0,002 Seconds
- 16. Relevant Found 5 141 Recall: |𝑹∩𝑭| |𝑹| 4246 = 𝟒𝟏 𝟒𝟔 = 0,89 = 89% „How much % of the relevant items were found?“
- 20. Recall: |𝑹∩𝑭| |𝑹| = 1 Perfect! All relevant items were found A lot of rubbish 803214523509235*1012 Billion results
- 21. Improve Recall • Find more relevant items (easier said than done) Recall: |𝑹∩𝑭| |𝑹|