In case of emergency, order pizza: an urgent case of action formation and recognition

The biggest challenge for voice technologies is action recognition. This is partly because current approaches prioritize abstract context over practical action, and tend to ignore the detailed, sequential structure of talk by emulating scripted, often stereotypical dialogue. This provocation paper analyzes an urgent case of how a caller and a 911 dispatcher work together to achieve action recognition. We outline their 'seen but unnoticed' interactional methods and suggest how computational systems can learn from conversation analysis and use micro-analytic detail to recognize social actions.