Indicative Summarization of Fiction

点击翻译按钮获取中文摘要

Wattpad is a social platform that hosts collections of fiction uploaded by its users for others to read. It now has a collection of over 50 million stories, and has a membership of over 40 million monthly users. A major issue faced by the company in improving its platform is that it needs to provide a way for users to find stories of interest within the collection. This project aims to develop methods for generating summaries of the stories, in order to help users decide whether they are worth reading. This is a research problem in natural language processing (NLP), and will involve developing techniques to understand the structure of the story at some level (e.g., the setting, characters, genre, or plot), decide on the information relevant to a potential reader, and then express that in text. We plan to investigate these issues using probabilistic methods of language understanding and generation that learn from a corpus of stories and accompanying abstracts. The outcomes of this research could provide a competitive advantage to this Canadian company by helping it offer a more attractive platform and product.