geval/CHANGELOG.md

213 lines
3.7 KiB
Markdown
Raw Normal View History

2019-08-10 16:31:54 +02:00
2021-10-28 08:52:44 +02:00
## 1.40.6.0 (2021-10-28)
Enhancements:
* Add p<P> flag for selecting the P% subset with the highest confidence scores
2021-10-09 18:32:31 +02:00
## 1.40.5.1 (2021-10-09)
Bug fixes:
* Improve line-by-line mode for BIO-F1
* Fix clarifications for submitting solutions
2021-08-20 19:40:16 +02:00
## 1.40.5.0 (2021-08-20)
Enhancements:
* Add PerplexityHashed metric
2021-08-03 14:36:04 +02:00
## 1.40.4.1 (2021-08-03)
Fixes:
* Improve diagnostics for broken metric definitions
2021-07-23 18:38:12 +02:00
## 1.40.4.0 (2021-07-23)
Fixes:
* Fix inconsistencies in handling probabilities in MultiLabel-F-score
* Plot calibration graphs for Probabilities-MultiLabel-F-score
* Improve diagnostics for unknown or broken metric definitions
2021-07-21 14:22:42 +02:00
## 1.40.3.0
Fix:
* Properly validate the train set
2021-07-19 17:09:29 +02:00
## 1.40.2.0
Enhancements:
* Matching specifications (e.g. fuzzy matching) can be used for Accuracy
2021-06-30 09:34:12 +02:00
## 1.40.1.0
Improvements:
* Handle DOS/Windows end-of-lines
2021-06-10 15:08:55 +02:00
## 1.40.0.0
New features:
* Add BIO-Weighted-F1 metric
* Handle filter & match combination of flags properly
2021-06-07 18:17:21 +02:00
## 1.39.0.0
New features:
* Add Haversine metric for distance on a sphere
2020-10-17 20:53:31 +02:00
## 1.38.0.0
New features:
* Add CER (Character-Error Rate) metric
* Spaces can be escaped with backslashes in configuration files
2020-08-08 21:52:03 +02:00
## 1.37.0.0
New features:
* Add --select-metric to select metric(s) by name (useful when you
have a complicated configuration with a large number of metrics,
and you want to see the result only for a specific metric, especially
in --line-by-line or --worst-features more)
* Add --show-preprocessed so that in --line-by-line or similar modes you
will be shown the results after the proprocessing
* --validate checks whether at least one metric is of priority 1
Bug fixes:
* Fix handling MultiLabel-F1 in --line-by-line mode when used with flags
2020-08-01 21:46:59 +02:00
## 1.36.1.0
* Add "c" and "t" flags
2020-07-02 18:38:20 +02:00
## 1.36.0.0
* Add fuzzy matching for MultiLabel-F1
2020-05-20 11:56:37 +02:00
## 1.35.0.0
* Add simple "mix" ensembling
2020-05-13 15:39:36 +02:00
## 1.34.0.0
* Add filtering (f<...> op for metrics)
2020-02-22 13:21:52 +01:00
## 1.33.0.0
* Handle headers in TSV files
2020-02-11 09:38:12 +01:00
## 1.32.2.0
* Fix bug in cross-tabs
2020-01-31 09:50:26 +01:00
## 1.32.0.0
* Add option to mark worst features
2020-01-28 22:41:48 +01:00
## 1.31.0.0
* Fix validation of challenges with Bootstrap resampling
2020-01-28 21:42:54 +01:00
## 1.30.0.0
* Automatically set precision when in Bootstrap mode
2020-01-27 22:53:15 +01:00
## 1.29.0.0
* Bootstrap resampling for most metrics
2020-01-11 17:03:57 +01:00
## 1.28.0.0
* Add `s` flag for substitution
2020-01-04 22:01:56 +01:00
## 1.27.0.0
* Results are formatted in cross-tables (if possible)
2019-12-21 16:05:51 +01:00
## 1.26.0.0
* Change the meaning of WER (WER is calculated for the whole set now
- similar to the way BLEU is calculated)
* Use `Mean/WER` if you want the old meaning (average of per-item results)
2019-12-16 12:49:18 +01:00
## 1.25.0.0
* Add --oracle-item-based
2019-12-14 21:12:46 +01:00
## 1.24.0.0
* Introduce metric priorities
* Use "Cartesian" strings in metrics
2019-12-13 20:38:08 +01:00
## 1.23.0.0
* New style of train data is preferred
- `in.tsv` and `expected.tsv` instead of `train.tsv`
- though this is not required as sometimes training data look different than test data
- `--validate` option was changed accordingly
2019-11-25 21:35:22 +01:00
## 1.22.1.0
* Add "Mean/" meta-metric (for the time being working only with MultiLabel-F-measure)
* Add :S flag
2019-11-18 21:53:40 +01:00
## 1.22.0.0
* Add SegmentAccuracy
2019-09-07 15:55:36 +02:00
## 1.21.0.0
* Add Probabilistic-MultiLabel-F-measure
2019-09-03 17:21:27 +02:00
## 1.20.1.0
* Fix Soft2D-F1 metric
* Check for invalid rectangles in Soft2D-F1 metric
2019-08-22 17:09:58 +02:00
## 1.20.0.0
* Add --list-metrics options
* Add Soft2D-F1 metric.
2019-08-20 07:55:22 +02:00
## 1.19.0.0
* Fully static build
* Add preprocessing options for metrics
2019-08-10 16:31:54 +02:00
## 1.18.2.0
* During validation, check the number of columns
* During validation, check the number of lines
* Validate train files
## 1.18.1.0
* During validation, check whether the maximum values is obtained with the expected data
2019-08-10 12:30:17 +02:00
## 1.18.0.0
* Add --validate option
2019-03-12 22:41:57 +01:00
## 1.17.0.0
* Add Probabilistic-Soft-F-score
2019-02-14 20:41:18 +01:00
## 1.16.0.0
* Handle JSONL files (only for MultiLabel-F-score)
* Fix SMAPE metric
## 1.0.0.1
2019-02-14 20:41:18 +01:00
* Added `--version`, `-v` options handling