ACA: the Translation Bureau’s Assistant Client Advisor
Bibliographic record
Abstract
This document reports on project "Document Workflow (``ACA)", part of the BtB-NRC Collaboration Agreement 2021-22, titled "Artificial Intelligence for Translation Quality" (AI4TQ). This project is about building computer support tools for the Translation Bureau's client advisors, more specifically to identify the specialty domain of texts submitted for translation. In a previous project, we used Bureau data to create classifiers that can identify to which domain a given document belongs, with an accuracy close to 80%. We delivered a system to the Bureau, called the "Assistant Client advisor" (ACA), which provides document classification as a web service, accessible both through a web-based user interface (UI) and an Application Programming Interface (API). In this project, we have greatly expanded this system, by developing a set of functionalities that allow creating, updating, evaluating and deploying domain predictors. The API and UI of this new system will allow the Bureau to create and maintain domain predictors themselves. In addition, we have experimented with approaches to improve prediction accuracy, most notably through neural networks. The new API allows creating and using predictors based on the FastText neural network technology, in addition to the algorithms previously available, SVM and ProbCat. In a series of experiments on Confidence Estimation, we have analyzed the performance of the classifiers, and the relationship between classification accuracy and some numerical indicators produced by classifiers, with the goal of distinguishing between documents that can be handled automatically and documents that should be verified by a client advisor, with the goal of minimizing domain prediction errors and human workload. Finally, we have added functionalities to segment large documents into smaller pieces, based on the predicted domain of individual segments of text.
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 | 0.005 | 0.015 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.002 | 0.000 |
| Scholarly communication | 0.004 | 0.005 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.320 | 0.211 |
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".