报告题目：Bidirectional Attentional Encoder-Decoder and Bidirectional Beam Search
报告时间：2019年4月9日（星期二）上午9:00-11:00
报告地点：校本部计算机楼212
报 告 人：Kamal Al-Sabahi(扎伊德), 16级留学博士，威廉希尔williamhill中文，威廉希尔williamhill中文
Abstract:
Sequence generative models with RNN variants, such as LSTM, GRU, show promising results. However, they still have some issues that limit their performance, especially while dealing with long sequences. One of the issues is that all current models employ a unidirectional decoder, which reasons only about the past and still limited to retain future context while giving a prediction. This makes these models suffer on their own by generating unbalanced outputs. To this end, bidirectional encoder-decoder architecture is used to tackle the aforementioned issues. Moreover, a bidirectional beam search mechanism is used to make inference from the bidirectional model.
