Twitter Translation using Translation-Based Cross-Lingual Retrieval

Microblogging services such as Twitter have become popular media for real-time usercreated news reporting. Such communication often happens in parallel in different languages, e.g., microblog posts related to the same events of the Arab spring were written in Arabic and in English. The goal of this paper is to exploit this parallelism in order to eliminate the main bottleneck in automatic Twitter translation, namely the lack of bilingual sentence pairs for training SMT systems. We show that translation-based cross-lingual information retrieval can retrieve microblog messages across languages that are similar enough to be used to train a standard phrasebased SMT pipeline. Our method outperforms other approaches to domain adaptation for SMT such as language model adaptation, meta-parameter tuning, or self-translation.