Entity-Dependent Long-Short Time Memory Network for Semantic Relation Extraction

Semantic relatedness between context information of two different entities, which are one of the most easily accessible features, have been shown to be very useful for detecting the semantic relations held in the text segments. However, these semantic relatedness information is often ignored by some researcher in the study of the problem of relation extraction. In this work, we propose a novel neural network for semantic relation extraction called entity-dependent long short-term memory network (ED-LSTM). This network can extract the relatedness of entity with the context, and select the relevant parts of contexts to infer the semantic relation towards the entity. We do some experiments on the SemEval-2010 Task 8 dataset. Extensive experiment and the results demonstrate that the proposed methods are effective for relation extraction, which can obtain state-of-the-art F1 result just with minimal feature engineering.