A deep learning based end-to-end system (F-Gen) for automated email FAQ generation

With overwhelming volumes of official emails being exchanged in enterprises every day, emails have become vital information storehouses. Automatic generation of FAQs from email systems helps in identifying important information and could serve potential applications such as chatbots and intelligent email answering. While there exist studies in the literature focusing on automatic FAQ generation and automated email answering, there are few studies that apply recently developed deep learning techniques to fetch FAQs from emails. This paper proposes a novel framework named F-Gen, which is an expert system that generates potential FAQs from emails utilizing state-of-the-art methodologies. The key characteristics of this study are as follows: 1. Designing F-Gen with various subsystems that interoperate together for the FAQ generation 2. Identifying the parameters that determine a valid FAQ. The three subsystems of F-Gen are: (a) query classifier subsystem (QC subsystem) for email texts, (b) FAQ group generator subsystem (FGG subsystem) for generating FAQ groups from email queries. And (c) FAQ generator subsystem (FG subsystem) for conversion of email query clusters into FAQs. Experiments on the email dataset that practically reflect the above-mentioned problem resulted in FAQs with a ROUGE-1 F-Score of 74.10% when compared with the ground truth. •A novel system for automatic generation of FAQs from emails•Models the characteristics of FAQs•Applies deep learning techniques for information extraction•Experiments on real world email datasets•Provides a high level Pseudocode