A Corpus for Multilingual Document Classification in Eight Languages
Holger Schwenk, Xi-An Li
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/facebookresearch/MLDocOfficialIn papernone★ 0
- github.com/n-waves/multifitnone★ 284
Abstract
Cross-lingual document classification aims at training a document classifier on resources in one language and transferring it to a different language without any additional resources. Several approaches have been proposed in the literature and the current best practice is to evaluate them on a subset of the Reuters Corpus Volume 2. However, this subset covers only few languages (English, German, French and Spanish) and almost all published works focus on the the transfer between English and German. In addition, we have observed that the class prior distributions differ significantly between the languages. We argue that this complicates the evaluation of the multilinguality. In this paper, we propose a new subset of the Reuters corpus with balanced class priors for eight languages. By adding Italian, Russian, Japanese and Chinese, we cover languages which are very different with respect to syntax, morphology, etc. We provide strong baselines for all language transfer directions using multilingual word and sentence embeddings respectively. Our goal is to offer a freely available framework to evaluate cross-lingual document classification, and we hope to foster by these means, research in this important area.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| MLDoc Zero-Shot English-to-Chinese | MultiCCA + CNN | Accuracy | 74.73 | — | Unverified |
| MLDoc Zero-Shot English-to-Chinese | BiLSTM (UN) | Accuracy | 71.97 | — | Unverified |
| MLDoc Zero-Shot English-to-French | BiLSTM (UN) | Accuracy | 74.52 | — | Unverified |
| MLDoc Zero-Shot English-to-French | BiLSTM (Europarl) | Accuracy | 72.83 | — | Unverified |
| MLDoc Zero-Shot English-to-French | MultiCCA + CNN | Accuracy | 72.38 | — | Unverified |
| MLDoc Zero-Shot English-to-German | MultiCCA + CNN | Accuracy | 81.2 | — | Unverified |
| MLDoc Zero-Shot English-to-German | BiLSTM (Europarl) | Accuracy | 71.83 | — | Unverified |
| MLDoc Zero-Shot English-to-Italian | BiLSTM (Europarl) | Accuracy | 60.73 | — | Unverified |
| MLDoc Zero-Shot English-to-Italian | MultiCCA + CNN | Accuracy | 69.38 | — | Unverified |
| MLDoc Zero-Shot English-to-Japanese | MultiCCA + CNN | Accuracy | 67.63 | — | Unverified |
| MLDoc Zero-Shot English-to-Russian | BiLSTM (UN) | Accuracy | 61.42 | — | Unverified |
| MLDoc Zero-Shot English-to-Russian | MultiCCA + CNN | Accuracy | 60.8 | — | Unverified |
| MLDoc Zero-Shot English-to-Spanish | MultiCCA + CNN | Accuracy | 72.5 | — | Unverified |
| MLDoc Zero-Shot English-to-Spanish | BiLSTM (UN) | Accuracy | 69.5 | — | Unverified |
| MLDoc Zero-Shot English-to-Spanish | BiLSTM (Europarl) | Accuracy | 66.65 | — | Unverified |
| MLDoc Zero-Shot German-to-French | BiLSTM (Europarl) | Accuracy | 75.45 | — | Unverified |