Optical Character Recognition2020en

mldoc-zero-shot-english-to-french

Dataset from Papers With Code

Metrics:accuracy, cer, wer, f1
Current State of the Art

XLMft UDA

Unknown

96.05

accuracy

accuracy Progress Over Time

Showing 3 breakthroughs from May 2018 to Sep 2019

72.478.885.391.798.2May 2018Dec 2018Sep 2019accuracyDate

Key Milestones

May 2018
BiLSTM (UN)

From paper: A Corpus for Multilingual Document Classification in Eight Languages

74.5
Dec 2018
Massively Multilingual Sentence Embeddings

From paper: Massively Multilingual Sentence Embeddings for Zero-Shot Cross-Lingual Transfer and Beyond

78.0
+4.6%
Sep 2019
XLMft UDACurrent SOTA

From paper: Bridging the domain gap in cross-lingual document classification

96.0
+23.2%
Total Improvement
28.9%
Time Span
1y 4m
Breakthroughs
3
Current SOTA
96.0

Top Models Performance Comparison

Top 6 models ranked by accuracy

accuracy1XLMft UDA96.0100.0%2MultiFiT, pseudo89.493.1%3Massively Multilingual Se...78.081.2%4BiLSTM (UN)74.577.6%5BiLSTM (Europarl)72.875.8%6MultiCCA + CNN72.475.4%0%25%50%75%100%% of best
Best Score
96.0
Top Model
XLMft UDA
Models Compared
6
Score Range
23.7

accuracyPrimary

Related Papers4

Other Optical Character Recognition Datasets