ELSSA: Explainable Large Language Model-Based Decision Support for Public Transit Incident Management
Bibliographic record
Abstract
Rapid incident response in public transit requires supervisors to make defensible decisions that align with both formal standard operating procedures (SOPs) and practical agency experience recorded in past reports. However, conventional retrieval-augmented generation (RAG) often fragments handbook context and underutilizes precedent knowledge embedded in unstructured control narratives. We present ELSSA, an Experience- and Logic-integrated System for Supervisor Assistance, which unifies rule-based and experience-based evidence for explainable incident-management support. ELSSA couples (i) a hierarchy-aware RAG pipeline over a structured ontology of the Toronto Transit Commission (TTC) incident handbook, and (ii) a knowledge-graph–based RAG module that converts TTC service occurrence reports into a regularized transit-incident knowledge graph via lightweight frequency-based auto-regularization; hybrid semantic and graph retrieval then surfaces relevant clauses and structurally similar precedents. Experiments using TTC supervisor training materials and Route 29 service occurrence reports show that ELSSA improves rule-based SOP question answering over strong RAG baselines and produces more robust experience-based control suggestions across original, radio-style, and paraphrased incident descriptions. A temporally separated 2023 held-out database further provides an initial test of generalization to newly received reports. A compliance analysis shows that service recovery and core documentation elements are highly visible in historical records which reflacts the reliability of the suggestions from ELSSA. Overall, ELSSA demonstrates how hierarchy-aware SOP retrieval and regularized incident knowledge graphs can support evidence-linked, auditable, and human-supervised decision assistance for public transit operations.
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 | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.018 | 0.004 |
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".