Co-located with NAACL 2018
June 5th or 6th, 2018. New Orleans, Louisiana
Human storytelling has existed for as far back as we can trace, predating writing. Humans have used stories for entertainment, education, cultural preservation; to convey experiences, history, lessons, morals; and to share the human experience.
Part of grounding artificial intelligence work in human experience can involve the generation, understanding, and sharing of stories. This workshop highlights the diverse work being done in storytelling and AI across different fields.
The WorkshopThis one-day, multi-modal and interdisciplinary workshop will bring together researchers and practitioners in NLP, Computer Vision, and storytelling. The focus will be on human storytelling: What storytelling is, its structure and components, and how it’s expressed, connected to the state of the art in NLP and related ML/AI areas:
- What we can understand from stories (natural language understanding)
- What we can generate to create human-like stories (natural language generation)
- What we can recognize multimodally for story understanding and generation (e.g., with computer vision)
- Contributed talks and posters.
- A visual storytelling challenge.
- Invited talks given by researchers in NLP, Computer Vision, and Storytelling.
Call For PapersWe invite work involving human storytelling with respect to machine learning, natural language processing, computer vision, speech, and other ML/AI areas.
This spans a variety of research, including work on creating timelines, detecting content to be used in a story, generating long-form text, and related multimodal work.
Data input sources may include professional and social-media content.
We also encourage ideas about how to evaluate user experiences in terms of coherence, composition, story comprehensiveness, and other aspects related to the creation of stories.
Paper topics may include, but are not limited to:
- The role of storytelling in artificial intelligence
- Accessible and Assistive storytelling
- Affect and emotion in stories
- Augmenting human storytelling
- Character relationships
- Collaborative storytelling
- Event summary diversity
- Event-episodes detection and annotation
- Multimodal grounding for storytelling
- Multimodal event timelines
- Narrative structure
- Plot structure
- Story concept detection and annotation
- Story generation
- Story understanding
- Storytelling Applications/Demos
- Temporal/Event structure
- Temporal and semantic alignment
- Writing stories
- User studies
Visual Storytelling ChallengeThis challenge begins to scratch the surface on how well artificial intelligence can share in this cultural human experience.
Participants are encouraged to work on creating AI systems that can generate stories for themselves, sharing the human experience that they see -- and begin to understand.
Click here to see more about the dataset.
Participants may submit to two different tracks: The Internal track and the External track.
Submissions are evaluated on how well they can generate human-like stories given a sequence of images as input.
Dates2 Feb 2018: Data train set augmented with additional stories
16 May 2018: Submissions due on EvalAI. (You will need to create an account to view the challenge)
30 May 2018: Results announced
Submission TracksInternal Track
For apples-to-apples comparison, all participants should submit to the Internal track. In this track, the only allowable training data is:
Any of the VIST storytelling data (SIS, DII, and/or the non-annotated album images)
Data available here
Allowed pretraining Data from any version of the ImageNet ILSVRC Challenge (common in computer vision).
Data from any version of the Penn Treebank (common in natural language processing).
If you wish to use any other sources of data/labels or pre-training, please submit to the External track.
Participants can use any data or method they wish during training (including humans-in-the-loop), but all data should be publicly available or made publicly available. At test time, the systems must be stand-alone (no human intervention). Possible datasets include data from ICCV/CVPR workshops, such as LSMDC, and other vision-language datasets, such as COCO, and VQA.
EvaluationEvaluation will have two parts:
Automatic: On EvalAI, using the automatic metric of METEOR.
Human: Crowdsourced survey of the quality of the stories.
Please follow the instructions listed on the challenge webpage on EvalAI.