File size: 1,117 Bytes
d7a89cb 95792ab d7a89cb 95792ab d7a89cb |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 |
---
language:
- la
license: agpl-3.0
tags:
- robust-speech-event
- hf-asr-leaderboard
datasets:
- lsb/poetaexmachina-mp3-recitations
metrics:
- wer
model-index:
- name: wav2vec2-base-it-latin
results:
- task:
type: automatic-speech-recognition
name: Speech Recognition
dataset:
type: lsb/poetaexmachina-mp3-recitations
name: Poeta Ex Machina mp3 recitations
metrics:
- type: wer
value: 0.398
name: Test WER
---
---
# wav2vec2-base-it-latin
This model is a fine-tuned version of [wav2vec2-base-it-voxpopuli](https://huggingface.co/facebook/wav2vec2-base-it-voxpopuli)
The dataset used is the [poetaexmachina-mp3-recitations](https://github.com/lsb/poetaexmachina-mp3-recitations),
all of the 2-series texts (vergil) and every tenth 1-series text (words from Poeta Ex Machina's [database](https://github.com/lsb/poetaexmachina/blob/master/merged-scansions.db) of words with scansions).
It achieves the following [results](https://github.com/lsb/tironiculum/blame/trunk/wav2vec2%20base%20it%20latin.ipynb#L1234) on the evaluation set:
- Loss: 0.1943
- WER: 0.398
|