Cross-lingual Lexical Sememe Prediction

This is the open-source code of the EMNLP 2018 paper Cross-lingual Lexical Sememe Prediction [pdf].

Introduction

Sememes are defined as the minimum semantic units of human languages. As important knowledge sources, sememe-based linguistic knowledge bases have been widely used in many NLP tasks. However, most languages still do not have sememe-based linguistic knowledge bases. Thus we present a task of cross-lingual lexical sememe prediction (CLSP), aiming to automatically predict sememes for words in other languages. We propose a novel framework to model correlations between sememes and multi-lingual words in low-dimensional semantic space for sememe prediction. Experimental results on real-world datasets show that our proposed model achieves consistent and significant improvements as compared to baseline methods in cross-lingual sememe prediction.

Usage

bash run.sh

To change the training corpus, please just switch the -mono-train1 and -mono-train2 parameters in bash.sh. Notice that lang1 refers to the source language and lang2 refers to the target language.

Datasets

Process	Type	Source	Target
Training	Corpus	Sogou-T	Wikipedia
	Seed Lexicon	Google Translate API
	Sememe-based KB	HowNet_zh	-
Testing	Sememe Prediction	-	HowNet_en
	Bilingual Lexicon Induction	Chinese-English Translation Lexicon 3.0 Version
	Word Similarity Computation	Wordsim-240	WordSim-353
	Word Similarity Computation	WordSim-297	SimLex-999

Cite

If the codes or datasets help you, please cite the following paper:

@InProceedings{qi2018cross,
  Title      = {Cross-lingual lexical sememe prediction},
  Author     = {Qi, Fanchao and Lin, Yankai and Sun, Maosong and Zhu, Hao and Xie, Ruobing and Liu, Zhiyuan},
  Booktitle  = {Proceedings of EMNLP},
  Year       = {2018},
}