Table of Contents
- Community Bonding Week 1: April 25-May 1
- Community Bonding Week 2: May 2-May 8
- Community Bonding Week 3: May 9-May 15
- Community Bonding Week 4: May 16-May 22
- Coding Week 1: May 23-May 29
- Coding Week 2: May 30-June 5
- Coding Week 3: June 6-June 12
- Coding Week 4: June 13-June 19
- Coding Week 5: June 20-June 26
- Coding Week 6: June 27-July 3 (Midterm deadline June 27)
- Coding Week 7: July 4-July 10
- Coding Week 8: July 11-July 17
- Coding Week 9: July 18-July 24
- Coding Week 10: July 25-July 31
- Coding Week 11: August 1-August 7
- Coding Week 12: August 8-August 14
- Submit code and evaluations: August 15-August 23
Community Bonding Week 1: April 25-May 1
- Introduction in the community.
- In the process of clearing doubts related to the project by discussion on mailing list.
Community Bonding Week 2: May 2-May 8
- Discussed the testing process of existing weighting functions on mailing list.
- Xapian-evaluation module: Started discussion regarding it with Gaurav.
Community Bonding Week 3: May 9-May 15
- Added weighting functions details in Project Plan
- Trying to set up xapian-evaluation module on my system.
Community Bonding Week 4: May 16-May 22
Semester exams started this week.
Coding Week 1: May 23-May 29
Exams continue. Will be getting back to project work this weekend.
Coding Week 2: May 30-June 5
- Modified the original bm25weight.cc that contains implementation of BM25 weighting function to add support for BM25+ weighting function.
- Also, modified existing test cases of BM25 to extend test coverage for BM25+ function.
- Added API support in weight.h for the same.
- Opened a pull request on Github repo for code review and feedback.
Coding Week 3: June 6-June 12
- Have changed the approach a bit for the implementation of BM25+ as the previous approach didn't turn out very well.
- Implemented a new subclass Xapian::BM25PlusWeight.
- Have written new test cases -- all passing.
- Pushed changes to update the previous PR.
Coding Week 4: June 13-June 19
- Mostly, used this week to make further improvements in the PR.
- Improved test coverage of BM25PlusWeight using the fact that k2 should have no effect on weight because k2 is associated with extra weight component and BM25+ has such no extra weight.
- Fixed some pesky issues related to tab-width in the code.
- Also, wrote an initial draft for the documentation of BM25+ weighting function and opened a PR here.
Coding Week 5: June 20-June 26
- Made remaining changes in the PR to fix some formatting issues.
- Successfully installed xapian-evaluation module on my system after fixing some issues faced during compilation. Will be performing the evaluation once we have the FIRE data hopefully next week or so.
- Started looking at the implementation of next weighting function i.e. PL2+.
Coding Week 6: June 27-July 3 (Midterm deadline June 27)
- Using the same approach used for implementing BM25+ previously, implemented a new subclass Xapian::PL2PlusWeight for PL2+ weighting function.
- Have written new feature tests -- all passing.
- Pushed changes to update the PR here.
- [Edit] As discussed in mid-term meeting, documentation is due for later weeks or after we have evaluation results of new weighting schemes.
Coding Week 7: July 4-July 10
- Xapian-evaluation module was successfully set up after quite a bit of effort. Thanks to Gaurav for all the help.
- Have run evaluation of existing schemes to get familiar with its working.
- Will be evaluating BM25+ and PL2+ weighting schemes in the coming week alongside the implementation of Dir+.
Coding Week 8: July 11-July 17
- Modified Xapian::LMWeight subclass to implement support for new smoothing method; Dir+.
- New tests added to validate changes.
- Started looking at the implementation Piv+ normalization function.
Coding Week 9: July 18-July 24
- Added more weighting schemes support in xapian-evaluation module. Pull request opened here.
- Ran evaluation tests of existing and new weighting schemes alike.
- Results files put together on Github Gist here for easy access.
- Next up, implementing the Piv+ normalization function in the next week
Coding Week 10: July 25-July 31
- Started implementing support for pivoted normalization in Tf-Idf weighting scheme.
Coding Week 11: August 1-August 7
- Completed pivoted normalization support.
- Added new test cases for normalization strings "PPP", "Ptn", "nPn" and "ntP".
- Evaluated & compared above mentioned normalization strings along with the default tf-idf normalization i.e. "ntn". Results published here.
- Discovered that "Ptn" normalization does fairly better job than all other tf-idf normalizations evaluated so far.
Coding Week 12: August 8-August 14
- Started writing documentation for new weighting schemes as well as previously implemented-but-not-documented schemes.
- Made some significant tweaks in BM25+ implementation following the code review by Olly.
- Classes started for the new semester.
Submit code and evaluations: August 15-August 23
- Completed documentation. PR (to be merged later on)
- Working on final tweaks and code clean-up following the code review of PL2+, Dir+, Piv+ implementation.
Note: See TracWiki for help on using the wiki.