[4Pin1] インタラクティブ(2)

[4Pin1-23] Multi-source neural grammatical error correction

〇Cao Guolin¹, Hiroya Takamura^2,3, Manabu Okumura^2,4 (1. School of Engineering, Tokyo Institute of Technology, 2. Institute of Innovative Research, Tokyo Institute of Technology, 3. AIST, 4. RIKEN AIP)

Keywords:Natural language processing, Grammatical error correction, Machine translation

This is the first attempt to use a multi-source encoder-decoder model for the grammatical error correction task (GEC). In addition to the possibly erroneous sentence written in a second language, our model uses the sentence written in the mother tongue of the learner. With our model, we achieved up to 1.13 GLEU score increases than the single source baseline model.