SentenceSimilarityMatching

Speciality: 2D Feature Map of two sentences is constructed by calculating W2V-Cosine-Similarity of their N-grams
Program Design: http://blog.csdn.net/u012052268/article/details/73164454
Train Data are gotten from: https://github.com/IAdmireu/ChineseSTS. 5494 samples in total, 50% POS v.s. 50% NEG
Accuracy: 96%+(max_lenth=40, n-gram=3, round=1000, batch=400)