You can refer to the reference paper "Zhiyuan Liu, Xinxiong Chen, Yabin Zheng, Maosong Sun. Automatic Keyphrase Extraction by Bridging Vocabulary Gap. The 15th Conference on Computational Natural Language Learning (CoNLL 2011)." for more details about training a new model.
In summary, you need to install and use giza++ platform to train a word alignment model, here is the link for downloading: https://github.com/moses-smt/giza-pp
You can refer to the reference paper "Zhiyuan Liu, Xinxiong Chen, Yabin Zheng, Maosong Sun. Automatic Keyphrase Extraction by Bridging Vocabulary Gap. The 15th Conference on Computational Natural Language Learning (CoNLL 2011)." for more details about training a new model.
In summary, you need to install and use giza++ platform to train a word alignment model, here is the link for downloading: https://github.com/moses-smt/giza-pp