Skip to content

Latest commit

 

History

History
117 lines (89 loc) · 2.9 KB

CHANGELOG.md

File metadata and controls

117 lines (89 loc) · 2.9 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

2022-01-17

Added

  • Support for Watson Assistant v2 API log extraction and analysis

2022-12-15

Added

  • Support for Watson Assistant v2 API accuracy (blind test and standard test only)

2022-03-03

Added

  • Optionally compare two separate blind test runs and compute comparison metrics with compare_results.py

2020-11-12

Changed

  • Pass Watson Assistant system_settings to test workspaces

2020-04-10

Changed

  • validateWS.py is now moved to log_analytics folder.
  • waObjects.py is extracted from that module to provide workspace parsing logic, usable in other modules including analytics

2020-01-15

Added

2019-12-02

Added

  • Configurable k-fold union output file name

2019-11-15

Added

  • Support for testing Natural Language Classifier models

2019-10-28

Removed

  • Support for username/password authentication. (Watson APIs require API key authentication)

2019-09-30

Added

  • Generates confusion matrix in k-fold and blind modes

2019-09-26

Added

  • Default values for most configuration values. Configuration effort is greatly reduced.

2019-09-23

Added

  • Utility for extracting conversation logs and utterances based on a given dialog node.

2019-09-03

Updated

2019-07-10

Added

  • Intent metrics include F1 score.

2019-06-21

Added

  • Support for IAM API key

2019-06-20

Added

2018-07-09

Added

  • Support optional partial credit intent tables

2018-07-03

Added

  • Support local workspace JSON as training input

2018-06-25

Added

  • Added intent metrics generation script

2018-06-14

Added

  • Made service base URL configurable

2018-05-21

Added

  • Refactored training input to only use workspace ID

2018-05-09

Added

  • Test cases for three modes and all sub-scripts are provided along with testing resources

2018-05-04

Added

  • Handling exception by cleaning out workspaces optionally after training phase.
  • Exposed maximum test rate and weight mode parameters.

2018-04-30

Fixed

  • Allowed intents with confidence less than 0.2 to be returned.

Added

  • Calculate precision using different weighting configuration.
  • Highlight confidence threshold on the precision curve figure.
  • Union all folds test output as kfold-test-out-union.csv.

2018-04-13

Added

  • Intent description generation script is provided in subfolder.

2018-04-13

Fixed

  • Reserved entities can be imported to workspaces.

2018-04-12

Added

  • Coroutine-based concurrent testing is provided in sub-script.