In contrast to the assessment criteria, which specify what to assess (e.g. whether a text is fluent, or whether a translation is good), metrics specify how to assess it. For example, the fluency assessment criteria can be assessed using the BLEU metric.
For most users, the default metrics for each assessment criteria will be sufficient. However, for researchers or other power users you may want to try different metrics as well. This page details the metrics supported by Critique.
Importantly, each page details the JSON Schema format of the:
- Configuration of each metric specifying various parameters that can be set.
- Data that is passed in to each metric.
- Result that is returned by each metric.
Currently supported metrics include:
If you have any questions or are interested in having other metrics supported by Critique, get in contact with us via our support chat or discord channel via the buttons on the left!