Rtasr

Latest version: v0.0.7

Safety actively analyzes 685525 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 1 of 2

0.0.7

This new release is correcting some nits on Wordcab side.

* Adjust the pricing for Wordcab paid services 101
* Fix a bug in the WordcabWord schema: score -> probability 104

0.0.6

🚀 ASR Providers

* Moved from Deepgram Nova to Nova 2 model 96
* Allowed user to run transcription on local audio files or folders 90
* Added the possibility to run transcription on self-hosted wordcab-transcribe version 94
* Improved the way we handle errors for ASR providers 83
* Fixed the retries strategy 81
* Decreased the waiting time between job status 93

🎯 Evaluation

* Added the `wrr` metric 87
* Updated the results with the new Deepgram Nova 2 model 98

📁 Datasets

* Fixed the `fleurs` dataset file path finding 80

💬 CLI commands

* Added the pricing feature to the `audio-length` command 88

0.0.5

🚀 ASR Providers

* Simplified the `launch`/`get_transcription` function for each ASR provider 55
* Implemented WER data preparation for 5 providers 66 67

🎯 Evaluation

* Implemented the WER evaluation process 70

📁 Datasets

* Added a new dataset `google/fleurs` for WER 77
* Fixes for AMI to avoid server overload 59
* Added the number of files per dataset splits 52
* Added compatible metrics for each dataset 60
* Removed one audio type for AMI to simplify transcription 59
* Added code to prepare AMI dataset for WER evaluation 57

💬 CLI commands

* Created a command to create plots from evaluation results 73
* Improved `list` command 60
* Added an `audio-length` command to get the duration of a dataset per split 74

⚙️ Tests
* Added a lot of unit tests 51

0.0.4

* Added `RevAI` and `Speechmatics` as asr providers for transcription 24 25
* Fixed the manifest file creation for AMI dataset 27
* Added AsrOutputs for 5 providers 30 32
* Added the `results_to_rttm` for 5 providers 30 34
* Implemented `use_cache` for transcription 35
* Fixed concurrency feature 32
* Fixed AMI dataset download by enabling concurrency to avoid saturating the server 32
* Define speaker mapping feature for all providers and datasets 37
* Add the `evaluation` command 38
* Fixed AssemblyAI transcription diarization params. 42
* Added DER evaluation 38 42 48
* Fixed `UU` speaker problem for Speechmatics 48
* Added `retries` for providers 48
* Added a script to generate plots for DER 48

0.0.3

* Define a base process for each ASR Providers 2
* Add Wordcab, AssemblyAI and Deepgram 12 13 20
* Add tests for the utils functions 15
* Write the full API transcription step for the benchmark cli command 2 17
* Fix a bug with the AMI dataset files 22

0.0.2

This release:
* Init the CLI f42c380b75021b2d637ddd54c0611ebc6f2532c5
* Added the `AMI` dataset e137f9d2bbb9b7f31dbb190c1bb3a12b649a9450
* Added the `VoxConverse` dataset edb9d9ded87973750376b79bd26a48bf96be5b22
* Docs are also deployed on main branch push 3c635f69c0bec8ecbd30a2164b70a17ba5afdf1e

Page 1 of 2

Links

Releases

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.