##Cross-lingual Dependency Parsing Based on Distributed Representations Jiang Guo, [email protected]
This parser can be used both for monolingual and cross-lingual dependency parsing. The monolingual component has been optimized and integrated into LTP.
###Compile
Require: cmake (available in most Linux systems)
- ./configure
- make
###Resources
Alignment dictionaries:
resources/align/PROJ
(target->source)resources/align/CCA
(source->target).
Note that the EN word embeddings in PROJ are used only for initialization, they get updated/finetuned while training.
###Running the executable
For robust projection: ./train-PROJ.sh
- set
fix_word_embeddings=false
(conf/nndep.cfg
)
For CCA: ./train-CCA.sh
Learning parameters are defined in conf/nndep.cfg
.
###Reference
@InProceedings{Guo:2015:clnndep,
author = {Guo, Jiang and Che, Wanxiang and Yarowsky, David and Wang, Haifeng and Liu, Ting},
title = {Cross-lingual Dependency Parsing Based on Distributed Representations},
booktitle = {Proceedings of ACL},
year = {2015},
}