OPUS-MT Dashboard

Language pair: [eng-fra] all languages
Models: [all models] [OPUS-MT] [external] [compare]
Selected: Tatoeba-MT-models/eng-fin/opusTCv20210807+nopar+ft95-jointvoc_transformer-tiny11-align_2022-01-27
Benchmark: all benchmarks [download]
Evaluation metric: bleu [spbleu][chrf][chrf++][comet]
Chart Type: [barchart] [heatmap]

orange = OPUS-MT, blue = Tatoeba-MT models, red = HPLT-MT models
green = student models, grey = external models, purple = user-contributed
render chart with [gd] [plotly]

Model Scores (selected model)

ID	Language	Benchmark	Output	bleu
0	eng-fin	flores101-devtest	show	21.9
1	eng-fin	flores200-devtest	show	21.9
2	eng-fin	newstest2015	show	21.1
3	eng-fin	newstest2016	show	23
4	eng-fin	newstest2017	show	25.4
5	eng-fin	newstest2018	show	16.1
6	eng-fin	newstest2019	show	22.6
7	eng-fin	newstestALL2016	show	26.1
8	eng-fin	newstestALL2017	show	28.8
9	eng-fin	newstestB2016	show	18.6
10	eng-fin	newstestB2017	show	21.3
11	eng-fin	ntrex128	show	16.8
12	eng-fin	tatoeba-test-v2020-07-28	show	36.7
13	eng-fin	tatoeba-test-v2021-03-30	show	36.6
14	eng-fin	tatoeba-test-v2021-08-07	show	36.6
			average	24.900