Can Quick Search / Full Text Search be configured?
I was able to determine the one result not appearing in the Unresolved search turned out to be the longest summary. Even though "Populate" was the first word, the length of the summary prevented it from showing in the results. As I removed words from the summary, the score went up and it started to appear. With only two words it was 100%, three or four words was 80%, five words was 70%, six words was 60%, seven words 50%, and eight or more words it dropped below 50% and didn't show in the results. I would think a search term appearing anywhere in the summary would net a score of at least 50%.
Can I configure what type of hit results in what percentage? Can I configure what percentage results in showing in the search results (such as 40% instead of 50%)? Can Quick Search be configured in any way? What I'd really like to do is make Quick Search ALWAYS show a match with a summary that contains the search word (score would be a minimum of 50% if the word is in the summary).
Accepted answer
RTC for mied searches (e.g. unresolved work items containing certain words) uses following algorithm:
1. Performs full text search on all work items from given PA. Full text search algorithm is not trivial and is described on https://jazz.net/library/article/824 This returns max 100 work items from given project area
2. Work items returned in 1 are filtered by additional criteria (e.g. display only unresolved work item).
There is currently RFE for that: https://jazz.net/jazz/web/projects/Rational%20Team%20Concert#action=com.ibm.team.workitem.viewWorkItem&id=238575. You can subscribe to it to show the interest to have improvement here.
Let us know if that helps.
Comments
Thanks. Indeed that is what is happening. I changed the limit from 100 to 200 and it had an affect on the results even when there are less than 100 to show. Suddenly after that change my search that returned 12 results is returning 24, with scores down to 25%. As you said, what the "unresolved" search is really doing is a search of ALL work items, limiting the results based on the limit setting, and then showing the unresolved results as a subset of the ALL results that were based on the limit. Awkward, but now that I know that I can work with it. I still would like a full word match in the summary to have a higher score, though. We a need to be able to configure Lucene. And now I have the image of Ricky Ricardo shouting "LU-CENE!" in my head. :-)