文献紹介 Simple task-specific bilingual word embeddings Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL, pages 1386–1390, Denver, Colorado, May 31 – June 5, 2015. 長岡技術科学大学 自然言語処理研究室 勝田 哲弘
Approach 1. Ct, Csを連結してシャッフル Ct : target corpus, Cs : source corpus 2. 各単語 の時、確率1/2でランダムに置換 • R : 辞書 3. For example, the English sentence “build the house”: construire the house, build la maison, build the maison, etc. 7
Cross-language super sense tagging • BARISTA embeddings for English-Danishをテスト • English SemCor (1000文), Danish (320文) • baseline:most frequent sense (MFS), structured perceptron model trained only with ortographic and POS features • Metric: weighted average over F1-scores 13