Difference between revisions of "Fall 2022 Schedule"
CompSemUser (talk | contribs) |
CompSemUser (talk | contribs) |
||
| Line 27: | Line 27: | ||
Active learning (AL) uses a data selection algorithm to select useful training samples to minimize annotation cost. This is now an essential tool for building low-resource syntactic analyzers such as part-of-speech (POS) taggers. Existing AL heuristics are generally designed on the principle of selecting uncertain yet representative training instances, where annotating these instances may reduce a large number of errors. However, in an empirical study across six typologically diverse languages (German, Swedish, Galician, North Sami, Persian, and Ukrainian), we found the surprising result that even in an oracle scenario where we know the true uncertainty of predictions, these current heuristics are far from optimal. Based on this analysis, we pose the problem of AL as selecting instances which maximally reduce the confusion between particular pairs of output tags. Extensive experimentation on the aforementioned languages shows that our proposed AL strategy outperforms other AL strategies by a significant margin. We also present auxiliary results demonstrating the importance of proper calibration of models, which we ensure through cross-view training, and analysis demonstrating how our proposed strategy selects examples that more closely follow the oracle data distribution. | Active learning (AL) uses a data selection algorithm to select useful training samples to minimize annotation cost. This is now an essential tool for building low-resource syntactic analyzers such as part-of-speech (POS) taggers. Existing AL heuristics are generally designed on the principle of selecting uncertain yet representative training instances, where annotating these instances may reduce a large number of errors. However, in an empirical study across six typologically diverse languages (German, Swedish, Galician, North Sami, Persian, and Ukrainian), we found the surprising result that even in an oracle scenario where we know the true uncertainty of predictions, these current heuristics are far from optimal. Based on this analysis, we pose the problem of AL as selecting instances which maximally reduce the confusion between particular pairs of output tags. Extensive experimentation on the aforementioned languages shows that our proposed AL strategy outperforms other AL strategies by a significant margin. We also present auxiliary results demonstrating the importance of proper calibration of models, which we ensure through cross-view training, and analysis demonstrating how our proposed strategy selects examples that more closely follow the oracle data distribution. | ||
|- style="border-top: 2px solid DarkGray;" | |- style="border-top: 2px solid DarkGray;" | ||
| − | | 3.17.21 || | + | | 3.17.21 || Conflict with DARPA KAIROS PI Meeting |
|- style="border-top: 2px solid DarkGray;" | |- style="border-top: 2px solid DarkGray;" | ||
| 3.24.21 || Skatje Meyers proposal | | 3.24.21 || Skatje Meyers proposal | ||
Revision as of 12:59, 8 February 2021
| Date | Title |
|---|---|
| 3.25.20 | Happy New Year! |
| 1.27.21 | Planning, Zihan Wang, Extending Multilingual BERT to Low-Resource Languages |
| 2.3.21 | cancelled because of DARPA AIDA PI Meeting conflict |
| 2.10.21 | Martha Palmer SCIL UMR practice talk |
| 2.17.21 | ACL paper discussion, led by Jon Cai & Sarah Moeller |
| 2.24.21 | Sarah Moeller practice talk |
| 3.3.21 | Clayton Lewis: Garfinkel and NLP - a discussion of challenges for Natural Language Understanding |
| 3.10.21 | guest of Alexis Palmer, Antonis Anastasapolous, [1]
Reducing Confusion in Active Learning [2] Active learning (AL) uses a data selection algorithm to select useful training samples to minimize annotation cost. This is now an essential tool for building low-resource syntactic analyzers such as part-of-speech (POS) taggers. Existing AL heuristics are generally designed on the principle of selecting uncertain yet representative training instances, where annotating these instances may reduce a large number of errors. However, in an empirical study across six typologically diverse languages (German, Swedish, Galician, North Sami, Persian, and Ukrainian), we found the surprising result that even in an oracle scenario where we know the true uncertainty of predictions, these current heuristics are far from optimal. Based on this analysis, we pose the problem of AL as selecting instances which maximally reduce the confusion between particular pairs of output tags. Extensive experimentation on the aforementioned languages shows that our proposed AL strategy outperforms other AL strategies by a significant margin. We also present auxiliary results demonstrating the importance of proper calibration of models, which we ensure through cross-view training, and analysis demonstrating how our proposed strategy selects examples that more closely follow the oracle data distribution. |
| 3.17.21 | Conflict with DARPA KAIROS PI Meeting |
| 3.24.21 | Skatje Meyers proposal |
| 3.31.21 | Capstone Projects |
| 4.7.21 | |
| 4.14.21 | |
| 4.21.21 | Rehan Ahmed proposal |
| 4.28.21 | Abhidip Bhattacharyya proposal |
| 5.05.21 |
Past Schedules
- Fall 2020 Schedule
- Spring 2020 Schedule
- Fall 2019 Schedule
- Spring 2019 Schedule
- Fall 2018 Schedule
- Summer 2018 Schedule
- Spring 2018 Schedule
- Fall 2017 Schedule
- Summer 2017 Schedule
- Spring 2017 Schedule
- Fall 2016 Schedule
- Spring 2016 Schedule
- Fall 2015 Schedule
- Spring 2015 Schedule
- Fall 2014 Schedule
- Spring 2014 Schedule
- Fall 2013 Schedule
- Summer 2013 Schedule
- Spring 2013 Schedule
- Fall 2012 Schedule
- Spring 2012 Schedule
- Fall 2011 Schedule
- Summer 2011 Schedule
- Spring 2011 Schedule
- Fall 2010 Schedule
- Summer 2010 Schedule
- Spring 2010 Schedule
- Fall 2009 Schedule
- Summer 2009 Schedule
- Spring 2009 Schedule
- Fall 2008 Schedule
- Summer 2008 Schedule
