Ensure that multi-line TREC topic titles are fully parsed. Update the… #1482
Conversation
… tests to test this, and update the regressions with numbers that take this into account
Codecov Report
@@ Coverage Diff @@
## master #1482 +/- ##
=========================================
Coverage 46.32% 46.32%
- Complexity 830 831 +1
=========================================
Files 197 197
Lines 10248 10242 -6
Branches 1349 1348 -1
=========================================
- Hits 4747 4745 -2
+ Misses 5109 5107 -2
+ Partials 392 390 -2
Continue to review full report at Codecov.
|
|
Great! Can you actually change the regression numbers here? https://github.com/castorini/anserini/blob/master/src/main/resources/regression/disk12.yaml The documentation is auto-gen'ed based on yaml above. |
…arkdown files, and the world goes round and round!
|
The YAML file's connected to the MD file, the MD file's connected to the... Updated the YAML and it re-generated the MD file, so I committed both. |
|
@eelstretching great, thanks! I'll queue up a run of all regressions on my end to make sure it doesn't break anything else, and then I'll merge. |
Documentation of regression change - fixes bug in reading of multi-line topics, issue #1482
Incorporated the change from @axlliu for ensuring that multi-line TREC topic titles are processed correctly. Added a test for title parsing to the set of tests.
Note that one of the existing tests (
testNewswireTopics) happened to hit on a multi-line title (topic 200), and that test needed to be fixed to include the second line of the title.As these changes change the queries, the numbers in the regressions for disk 1 and 2 need to change. I re-ran the index and the evaluations using the new code, which resulted in quite a few changes for the regression numbers (some better, some worse.)