Skip to main content
 

Snigdha Chaturvedi discussed “Data Storytelling through Natural Language Generation,” and presented projects that implement natural language processing (NLP) to convert computer data to human text. Using NLP, human text can be generated through text, graph, or a table. One of the projects Chaturvedi works on is the aspect aware summarization of a collection of product reviews, where an NLP system will read hundreds of reviews and summarize them into a short paragraph, based on specific attributes the user is searching for. An additional project involves story generation, where the NLP system will be given a title and emotional arc to automatically generate a short story. Chaturvedi explained that her team also conducts research on the issues of fairness and implicit bias in text generation by studying auto-generated short stories and comparing the motivations and experiences of the male and female protagonists. Her team found that stories with female protagonists involved themes such as body image, religion, and  family, while male protagonists’ stories involved power and death. Chaturvedi ended her talk by discussing efforts to debias text representation before text is used by the NLP system. Click here to view the talk on YouTube.

 

Snigdha Chaturvedi | Department Bio

Assistant Professor, Computer Science

Featured on: June 23, 2022 (Event Page)

Session Title: Data Science: More than Numbers (Event Recap

Tools, Information, and Resources: