Towards the LLM-Based Generation of Formal Specifications from Natural-Language Contracts: Early Experiments with Symboleo
Bibliographic record
Abstract
Over the past decade, different domain-specific languages (DSLs) were proposed to formally specify requirements stated in legal contracts, mainly for analysis but also for code generation. Symboleo is a promising language in that area. However, writing formal specifications from natural-language contracts is a complex task, especially for legal experts who do not have formal language expertise. This paper reports on an exploratory experiment targeting the automated generation of Symboleo specifications from business contracts in English using Large Language Models (LLMs). Combinations (38) of prompt components are investigated (with/without the grammar, semantics explanations, 0 to 3 examples, and emotional prompts), mainly on GPT-4o but also to a lesser extent on 4 other LLMs. The generated specifications are manually assessed against 16 error types grouped into 3 severity levels. Early results on all LLMs show promising outcomes (even for a little-known DSL) that will likely accelerate the specification of legal contracts. However, several observed issues, especially around grammar/syntax adherence and environment variable identification (49%), suggest many areas where potential improvements should be investigated.
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 | 0.010 | 0.062 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.006 | 0.002 |
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".