geval/CHANGELOG.md
2021-08-03 14:36:04 +02:00

194 lines
3.4 KiB
Markdown

## 1.40.4.1 (2021-08-03)
Fixes:
* Improve diagnostics for broken metric definitions
## 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
## 1.40.3.0
Fix:
* Properly validate the train set
## 1.40.2.0
Enhancements:
* Matching specifications (e.g. fuzzy matching) can be used for Accuracy
## 1.40.1.0
Improvements:
* Handle DOS/Windows end-of-lines
## 1.40.0.0
New features:
* Add BIO-Weighted-F1 metric
* Handle filter & match combination of flags properly
## 1.39.0.0
New features:
* Add Haversine metric for distance on a sphere
## 1.38.0.0
New features:
* Add CER (Character-Error Rate) metric
* Spaces can be escaped with backslashes in configuration files
## 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
## 1.36.1.0
* Add "c" and "t" flags
## 1.36.0.0
* Add fuzzy matching for MultiLabel-F1
## 1.35.0.0
* Add simple "mix" ensembling
## 1.34.0.0
* Add filtering (f<...> op for metrics)
## 1.33.0.0
* Handle headers in TSV files
## 1.32.2.0
* Fix bug in cross-tabs
## 1.32.0.0
* Add option to mark worst features
## 1.31.0.0
* Fix validation of challenges with Bootstrap resampling
## 1.30.0.0
* Automatically set precision when in Bootstrap mode
## 1.29.0.0
* Bootstrap resampling for most metrics
## 1.28.0.0
* Add `s` flag for substitution
## 1.27.0.0
* Results are formatted in cross-tables (if possible)
## 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)
## 1.25.0.0
* Add --oracle-item-based
## 1.24.0.0
* Introduce metric priorities
* Use "Cartesian" strings in metrics
## 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
## 1.22.1.0
* Add "Mean/" meta-metric (for the time being working only with MultiLabel-F-measure)
* Add :S flag
## 1.22.0.0
* Add SegmentAccuracy
## 1.21.0.0
* Add Probabilistic-MultiLabel-F-measure
## 1.20.1.0
* Fix Soft2D-F1 metric
* Check for invalid rectangles in Soft2D-F1 metric
## 1.20.0.0
* Add --list-metrics options
* Add Soft2D-F1 metric.
## 1.19.0.0
* Fully static build
* Add preprocessing options for metrics
## 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
## 1.18.0.0
* Add --validate option
## 1.17.0.0
* Add Probabilistic-Soft-F-score
## 1.16.0.0
* Handle JSONL files (only for MultiLabel-F-score)
* Fix SMAPE metric
## 1.0.0.1
* Added `--version`, `-v` options handling