Turning interviews into process models — Technical https://thesis.uya.no/proposals/from-conversation-to-process/#technical-track BACHELOR Final result: A process-drafting prototype and documented error types. Task: Build a transcript-to-process draft using an existing model and a fixed prompt. Data: 12 line-numbered scripts across 3 processes (planned) Four role accounts each for leave/equipment/expenses; 2 same-context conflicts per process. Independently finalise step, sequence and conflict keys; separate real conflicts from valid role variants. Requires: Independent annotator, trainee analysts and inference compute; scripts must be authored. Method: Use the scripted interviews; check steps/edges against the independent key and record analyst corrections. Plan 4 participant sessions. STEPS Build and test 1. Read the starting sources and choose one established implementation method. 2. Write the requirements, data fields, system diagram and test cases. 3. Prepare 12 line-numbered scripts across 3 processes. Make the answer key and pass criteria before testing. 4. Build a working version of a process-drafting prototype and documented error types. 5. Run function, integration and failure-case tests. Record each result. 6. Run the practical evaluation and list the changes the system still needs. MASTER Final result: Validated JSON-to-BPMN pipeline, rendered diagrams, provenance links and error analysis. Research question: Investigate how source-linked process drafts handle conflicting stakeholder accounts. Separate syntax, step/edge correctness and analyst correction; compare identical drafts with/without source links and distinguish contradictions from legitimate process variants. Research result: Design principles for traceable process modelling under conflicting evidence. STEPS Design, build and test (DSR) 1. Read the newest papers and list the closest existing systems. 2. Write down the versions, fields, data split, case assignment, random seeds and correct answers for 12 line-numbered scripts across 3 processes. 3. Draw the user workflow, data model and system architecture. List the requirements and pass criteria. 4. Build a working version of validated json-to-bpmn pipeline, rendered diagrams, provenance links and error analysis. 5. Test every function, connection and failure case. Save the failed tests as well as the passed tests. 6. Compare the system with the named alternative. Then run the user task or decision task in the assignment. 7. Report the measured result, the failed cases and the design lessons another team can reuse. CURRENT PROJECT LITERATURE LLM-based Business Process Models Generation from Textual Descriptions (2025, peer-reviewed): https://doi.org/10.18653/v1/2025.findings-ijcnlp.31 Provides a current method and benchmark framing for text-to-process-model generation. Evaluating Large Language Models in Process Mining (2024, peer-reviewed): https://doi.org/10.1007/978-3-031-61007-3_2 Maps current LLM capabilities and limitations in process-analysis tasks. IS THEORY STARTING POINTS Berti et al. (2024) — Evaluating LLMs in Process Mining: https://doi.org/10.1007/978-3-031-61007-3_2 Separate model syntax, process meaning and usefulness to process analysts. Wang & Strong (1996) — Beyond Accuracy: https://www.tandfonline.com/doi/abs/10.1080/07421222.1996.11518099 Distinguish accuracy, completeness, representation and fitness for the task. Search Scopus or Web of Science and ACM Digital Library using the topic query, then follow citations to the thesis start date. Record searches and compare methods, data, findings and limitations in literature-matrix.csv. Use that review to confirm or revise the gap and choose a current comparator. The linked papers are starting points. PILOT QUESTION Do transcript references improve the correctness and reviewability of generated process models? EXAMPLE TOOLS Qwen2.5-7B-Instruct, Python, Pydantic, BPMN 2.0 XML, bpmn-js Equivalent tools are fine. FIRST THREE ACTIONS 1. Write the three role accounts for one approval process and manually draw a reference diagram. Confirm access to inference compute and record model, software and hardware versions. 2. Mark supporting transcript lines and deliberate contradictions before trying the model. 3. Run extraction with one fixed prompt and validate the JSON and source references manually. LITERATURE SEARCH business process modelling interview provenance ambiguity requirements DATA 12 line-numbered scripts across 3 processes (planned) Four role accounts each for leave/equipment/expenses; 2 same-context conflicts per process. Independently finalise step, sequence and conflict keys; separate real conflicts from valid role variants. Qwen2.5-7B-Instruct: https://huggingface.co/Qwen/Qwen2.5-7B-Instruct Transformers generation settings: https://huggingface.co/docs/transformers/main_classes/text_generation HOW TO TEST IT Use 12 line-numbered scripts and an independent dual-modeller gold BPMN. Report task/edge/conflict/citation precision-recall, graph edit distance, JSON/BPMN validity and token-flow soundness. Four trainee analysts inspect fixed drafts with links on/off; record justified corrections and time. FINAL RESULT Validated JSON-to-BPMN pipeline, rendered diagrams, provenance links and error analysis. ACCESS OR PEOPLE Independent annotator, 4 trainee analysts and inference compute; scripts must be authored. TEMPLATES https://thesis.uya.no/starters/from-conversation-to-process/technical/data-plan.csv https://thesis.uya.no/starters/from-conversation-to-process/technical/evaluation.csv https://thesis.uya.no/starters/from-conversation-to-process/technical/literature-matrix.csv Use participant codes instead of names or email addresses.