Change Details

Hoping to release the Discernatron (human judgment of search relevance) some time next week at https://relevance.wmflabs.org. Although i suppose i could change the URL to https://discernatron.wmflabs.org before release, might be good :). Anyways we should come up with some sort of plan to get some users onto the site and grading results for us. We plan to release a new set of 500 queries into it each month to be graded, and ideally we need each query graded by 10+ people. Possible avenues for getting users: * Village Pump Technical * Anywhere people have discussed how much search sucks * Moar?

Hoping to release the Discernatron (human judgment of search relevance) some time next week at https://relevance.wmflabs.org. Although i suppose i could change the URL to https://discernatron.wmflabs.org before release, might be good :). Anyways we should come up with some sort of plan to get some users onto the site and grading results for us. We plan to release a new set of 500 queries into it each month to be graded, and ideally we need each query graded by 10+ people. For each query users will be provided the set of titles found by cirrus, Bing, Google, and duckduckgo and asked to rank each title from 1 to 4, or leave the title unranked. We source up to 25 results per source, but due to duplicates there are probably 20 to 75 titles on each (not sure yet) Possible avenues for getting users: * Village Pump Technical * Anywhere people have discussed how much search sucks * wikitech-l and discovery mailing lists * Moar?

Hoping to release the Discernatron (human judgment of search relevance) some time next week at https://relevance.wmflabs.org. Although i suppose i could change the URL to https://discernatron.wmflabs.org before release, might be good :). Anyways we should come up with some sort of plan to get some users onto the site and grading results for us. We plan to release a new set of 500 queries into it each month to be graded, and ideally we need each query graded by 10+ people. For each query users will be provided the set of titles found by cirrus, Bing, Google, and duckduckgo and asked to rank each title from 1 to 4, or leave the title unranked. We source up to 25 results per source, but due to duplicates there are probably 20 to 75 titles on each (not sure yet) Possible avenues for getting users: * Village Pump Technical * Anywhere people have discussed how much search sucks * wikitech-l and discovery mailing lists * Moar?