Attribute participant satisfaction to the shared representation

Determine whether participants’ satisfaction with SportSAGE highlights is attributable to the semantic action graph shared representation rather than to other aspects of the system, by comparing the graph-based approach against an unstructured baseline.

Background

SportSAGE is presented as a design probe for using a semantic action graph to support both agentic sports-highlight generation and viewer-facing interpretation. The evaluation used a single soccer match, one LLM, and no unstructured comparison system.

Because the study lacks an unstructured baseline, the authors explicitly state that they cannot yet determine whether participant satisfaction results from the shared semantic representation itself. A direct comparison with an unstructured baseline is identified as future work needed to resolve this attribution question.

References

As a design probe, SportSAGE was evaluated on a single match with one model and no unstructured baseline, so we cannot yet attribute participants' satisfaction to the shared representation itself. Future work should compare against an unstructured baseline, extend the schema with spatial and tactical attributes, and explore constructing the graph directly from video rather than official feeds.

— Semantic Action Graph: A Shared Representation for Agent Grounding and Human Interpretation of Sports Highlights  (2609.20768 - Lin et al., 17 Sep 2026) in Conclusion and Future Work, Section 6