Semantic representations of human actions across vision and language
Bibliographic record
Abstract
Humans can visually recognize many actions performed by others, as well as communicate about them. How are action concepts organized in the mind? Recent work has uncovered shared neural representations of actions across vision and language, yet the semantic structure of these representations is not well understood. To address this, we curated a multimodal, naturalistic action set containing 95 videos of everyday actions from the Moments in Time dataset (Monfort et al., 2019) and 95 naturalistic sentences describing the same actions. We labeled each action with four semantic features: a specific action verb (e.g., chopping); an everyday activity (representing a set of actions; e.g., preparing food); the target of the action (e.g., an object); and a broad action class (e.g., manipulation; Orban et al., 2021). We also annotated the actions with other relevant social and action-related features. We used these features to predict behavioral similarity measured in two multiple arrangement experiments. Participants arranged the videos (N = 39) or sentences (N = 32) according to the actions’ similarity in meaning. In both experiments, the action target explained more unique variance in behavior than any other feature. In a cross-modal analysis, we mapped the semantic, action, and social features to the video similarity judgments to predict the sentence similarity judgments, and vice versa. Our feature set explained approximately 80% of the shared variance across modalities. Of all features, action target and action class were the best cross-modal predictors. Together, our results demonstrate the shared semantic organization of human actions across vision and language. This organization reflects broad semantic features, including action target and action class. Our results challenge commonly used definitions of action categories, and open exciting avenues for understanding how action concepts are represented in the mind and brain.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.006 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".