Detail of the student project

List
Topic:Automatic email reply generation
Supervisor:Jan Sedivy
Announce as:DP,BP,PTO
Description:In this project, we want to research methods for automatic generation of short responses to email or social networks messages. Specifically, on a cell phone, it can be a great advantage to select from a set of semantically diverse replies. We want first cover short messages of few words. The initial steps will include a review of LSTM and GRU neural networks architectures and a meaningful training set construction.
Instruction:Select a source of messages social media or email and collect a training set.
Review and select recurrent Neural Network for reply generation.
Train an NN model.
Provide a language model for simple and semantically diverse replies.
Write a simple application for the purpose of testing.
Bibliography:https://arxiv.org/pdf/1606.04870v1.pdf
Date:15.02.2017
Max.number of students:0
Responsible person: Petr Pošík