Institute of Information Science, Academia Sinica

Storytelling is one of the oldest known human activities. People engage in storytelling to communicate, teach, entertain, establish identity, or simply relate to each other in meaningful ways. Storytelling is important, but writing a good story is a challenging and complicated task, and many creative writers struggle to come up with ideas throughout the process. Despite this common experience, research into technological writing support does not have much to say to help story writers. In this talk, we will introduce two lines of our work on storytelling support: (i) Heteroglossia and (ii) visual story post-editing.
  Heteroglossia is an add-on for Google Docs that allows writers to elicit story ideas from the online crowd using their text editors. Writers can share snippets of their working drafts and ask the crowd to provide follow-up story ideas based on it. Heteroglossia employs a strategy called "role play", where each worker is assigned a fictional character in a story and asked to brainstorm plot ideas from that character's perspective. Our deployment with two experienced story writers shows that Heteroglossia is easy to use and can generate interesting ideas. Heteroglossia allows us to gain insight into how future technologies can be developed to support ideation in creative writing.
  Another critical aspect of story writing is "editing". Professional writers emphasize the importance of editing. Given that professionals revise and rewrite their drafts intensively, machines that generate stories may also benefit from a good editor. Per the evaluation of the first Visual Storytelling Challenge, the ability of an algorithm to tell a sound story is still far from that of a human. Users will inevitably need to edit generated stories before putting them to real uses, such as sharing on social media. To this end, we introduce the first dataset for human edits of machine-generated visual stories and explore how these collected edits may be used for the task of visual story post-editing. We establish baselines for the task, showing how a relatively small set of human edits can be leveraged to boost the performance of large visual storytelling models. We also discuss the weak correlation between automatic evaluation scores and human ratings, motivating the need for new automatic metrics.

Ting-Hao (Kenneth) Huang is an Assistant Professor in the College of Information Sciences and Technology (IST) at the Pennsylvania State University. His research combines AI with crowdsourcing to create Crowd-AI systems that are usable, robust, and intelligent. Dr. Huang has published in both top HCI and AI conferences, including CHI, UIST, HCOMP, and AAAI. He has also published in top NLP conferences, including ACL, NAACL, COLING, EMNLP, IJCNLP, and LREC.