Recurrent Point Review Models

Author: K. Cvejoski, C. Ojeda, B. Georgiev, C. Bauckhage, R. J. Sanchez
Journal: 2020 International Joint Conference on Neural Networks (IJCNN)
Year: 2020

Citation information

K. Cvejoski, C. Ojeda, B. Georgiev, C. Bauckhage, R. J. Sanchez,
2020 International Joint Conference on Neural Networks (IJCNN),
2020,
1-8,
IEEE,
Glasgow, UK,
https://doi.org/10.1109/IJCNN48605.2020.9206768

Deep neural network models represent the state-of-the-art methodologies for natural language processing. Herewe build on top of these methodologies to incorporate temporalinformation and model how review data changes with time.Specifically, we use the dynamic representations of recurrentpoint process models, which encode the history of how business orservice reviews are received in time, to generate instantaneouslanguage models with improved prediction capabilities. Simul-taneously, our methodologies enhance the predictive power ofour point process models by incorporating summarized reviewcontent representations. We provide recurrent network andtemporal convolution solutions for modeling the review content.We deploy our methodologies in the context of recommendersystems, effectively characterizing the change in preference andtaste of users as time evolves.