Aspect term extraction via information-augmented neural network

Abstract Aspect term extraction (ATE) aims at identifying the aspect terms that are expressed in a sentence. Recently, Seq2Seq learning has been employed ATE and significantly improved performance. However, it suffers from some weaknesses, such as lacking ability to encode more informative information integrate of surrounding words encoder. The static word embeddings fall short modeling dynamic meaning words. To alleviate problems mentioned above, this paper proposes information-augmented neural network (IANN) which is novel framework. In IANN, specialized developed key module encoder, named multiple convolution with recurrence (MCRN), contextualized embedding layer designed capture sense. Besides, AO ({ A spect, O utside}) tags proposed less challenging tagging scheme. lot experiments have performed on three widely used datasets. These demonstrate IANN acquires state-of-the-art results validate powerful method for task.