speechmatics-diarization-metrics 0.0.0

Python module for evaluating speaker diarisation.

# diarisation-metrics

This repository is Speechmatics’ fork of https://github.com/pyannote/pyannote-metrics. We have changed the project to fit our python project template. ## Getting Started ### Wheel

This project is built as a Python wheel by the CI. To download and use an existing wheel, get the latest version of sm_diarisation_metrics from Artifactory. e.g.

` $ pip3 install --extra-index-url https://artifacts.speechmatics.io/speechmatics/api/pypi/pypi-master/simple sm_diarisation_metrics `

This package has a CLI supporting ctm, lab, or V2 JSON format that can be run using eg python3 -m sm_diarisation_metrics.cookbook reference.json transcript.json ### Docs

To build the docs run make docs. This will create a directory containing HTML docs under docs/build/html. Run make serve-docs to run a web server to serve these docs locally at http://localhost:8080. ### Command line interface

For convenience we have created aladdin.diarization.metrics.sm_diarisation_metrics.cookbook which features a simplified CLI for the most common tasks that we need to perform at Speechmatics for evaluating diarisation.

Currently this CLI just offers one interface which accepts a reference and hypothesis diarisation data file in either ctm, lab, or V2 JSON format and returns the diarisation error rate. To use this run for example python3 -m aladdin.diarization.metrics.sm_diarisation_metrics.cookbook ref.ctm hyp.lab. ### Tests

Run the unit tests with make test. ### License

This project is released under a proprietary [Speechmatics license](insert-link-here)