Measuring lexical distance between parallel corpora: the case of AI-generated news translation
Bibliographic record
Abstract
Since the University of Warwick’s news translation project in the mid-2000s, it has been a truism that journalists rarely translate whole articles but instead compose stories using texts in other languages as one source among others. However, the development of AI-based machine translation has brought about a shift in journalistic practices. Increasingly, multilingual news agencies are using these tools to produce similar stories in multiple languages. One consequence has been that researchers can now compile parallel corpora of translated stories. This article proposes a method to characterize such corpora by measuring the distance between source and target texts, a method it applies to stories published in English and French on the website SwissInfo.ch. It describes the mechanics of corpus-building, article vectorization, and the creation of a lexical substitution list that makes measurement possible. It then proposes three measures – Euclidean, Jaccard, and cosine – which have complementary strengths and weaknesses. The value of these measurement tools is heuristic: they make it possible to identify patterns that can be investigated using other methods more familiar to news translation researchers, such as interviews or direct observation.
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 | 0.011 | 0.088 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.006 | 0.012 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.004 | 0.007 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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".