Intelligent sentence completion based on global context dependent recurrent neural network language model
Tao Yang, Hongli Deng · 2019
This paper proposes a global context dependent recurrent neural network language model to obtain the global semantics of the target in the sentence completion task, in order to overcome the problem of the limited context in existing recurrent neural network language model. The model uses all the preceding and succeeding words of the target to construct the global context of the target, so as to capture the complete semantic information of the target without resorting to any external sentence information. This paper verifies the validity of the model on the data in Microsoft Sentence Completion Challenge. The experimental results show that the proposed model obtains higher completion accuracy than other language models. Moreover, it also obtains higher completion accuracy than the commonly used word similarity model, even better than RNNMEs and Skip-gram+RNNMEs combination models.