Co-occurrence information between words is the basis of training word embeddings, besides, Chinese characters are composed of subcharacters, words made up by same characters or subcharacters usually have similar semantics, but these internal substructure information are usually neglected in popular models. In this work, We propose a novel method for learning Chinese word embeddings, which takes full use of external co-occurrence context information and internal substructure information. We represent each word as a bag of subcharacter n-grams, and our model learns the vector representation corresponding to the word and its subcharacter n-grams. The final word embeddings is represented as the sum of these two kinds of vector representation, which makes the learned word embeddings can take into account both internal structure information and external co-occurrence information possible. Experiments show that our method outperforms state-of-the-art performance on benchmarks.

Distant supervision is a widely applied approach in field of relation extraction, which could automatically generate large amounts of labeled training corpus with minimal manual effort. However, the labeled training corpus may have many false positiv

Distant supervision, a widely applied approach in the field of relation extraction can automatically generate large amounts of labeled training corpus with minimal manual effort. However, the labeled training corpus may have many false-positive data, which would hurt the performance of relation extraction. Moreover, in traditional feature-based distant supervised approaches, extraction models adopt human design features with natural language processing. It may also cause poor performance. To address these two shortcomings, we propose a customized attention-based long short-term memory network. Our approach adopts word-level attention to achieve better data representation for relation extraction without manually designed features to perform distant supervision instead of fully supervised relation extraction, and it utilizes instance-level attention to tackle the problem of false-positive data. Experimental results demonstrate that our proposed approach is effective and achieves better performance than traditional methods.