EnDex: Evaluation of Dialogue Engagingness at Scale
Xu, Guangxuan, Liu, Ruibo, Harel-Canada, Fabrice, Chandra, Nischal Reddy, Peng, Nanyun
–arXiv.org Artificial Intelligence
We propose EnDex, the first human-reaction based model to evaluate dialogue engagingness. EnDex is trained on 80k Reddit-based Engagement Dataset (RED) curated using a novel distant-supervision framework. Engagingness is a key measure that captures high-level quality of AI dialogue systems and closely reflects actual user experience. However, data shortage, plus the abstract and extensive definition of engagingness makes it challenging to develop an automatic metric. Our work departs from mainstream approaches that use synthetic negative examples to train binary classifiers, and instead, proposes a solution using distant-supervision from human-reaction feedback. To support the soundness of our EnDex metric, we offer a theoretical foundation for engagement, an extensive ablation study, and empirical evidence of high correlation on five engagingness related datasets. We will release code, off-the-shelf EnDex model, and a large-scale dataset upon paper publication to facilitate future research.
arXiv.org Artificial Intelligence
Oct-22-2022
- Country:
- Oceania > Australia
- North America
- United States
- Minnesota > Hennepin County
- Minneapolis (0.14)
- Louisiana > Orleans Parish
- New Orleans (0.04)
- California
- Los Angeles County > Los Angeles (0.14)
- San Diego County > San Diego (0.04)
- Minnesota > Hennepin County
- Canada > British Columbia
- United States
- Europe
- Sweden > Stockholm
- Stockholm (0.04)
- Spain > Catalonia
- Barcelona Province > Barcelona (0.04)
- Denmark > Capital Region
- Copenhagen (0.04)
- Sweden > Stockholm
- Asia > Taiwan
- Taiwan Province > Taipei (0.04)
- Africa > Ethiopia
- Addis Ababa > Addis Ababa (0.04)
- Genre:
- Research Report (0.40)
- Industry:
- Leisure & Entertainment > Games (0.68)
- Technology: