The Utility of Context When Extracting Entities From Legal Documents
Bibliographic record
Abstract
When reviewing documents for legal tasks such as Mergers and Acquisitions, granular information (such as start dates and exit clauses) need to be identified and extracted. Inspired by previous work in Named Entity Recognition (NER), we investigate how NER techniques can be leveraged to aid lawyers in this review process. Due to the extremely low prevalence of target information in legal documents, we find that the traditional approach of tagging all sentences in a document is inferior, in both effectiveness and data required to train and predict, to using a first-pass layer to identify sentences that are likely to contain the relevant information and then running the more traditional sentence-level sequence tagging. Moreover, we find that such entity-level models can be improved by training on a balanced sample of relevant and non-relevant sentences. We additionally describe the use of our system in production and how its usage by clients means that deep learning architectures tend to be cost inefficient, especially with respect to the necessary time to train models.
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 | 0.013 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.005 | 0.003 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.007 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 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".