Comparative Analysis of Transformer-Based Language Models for Text Analysis in the Domain of Sustainable Development
Bibliographic record
Abstract
With advancements of Artificial Intelligence, Natural Language Processing (NLP) has gained a lot of attention because of its potential to facilitate complex human-machine interactions, enhance language-based applications, and automate processing of unstructured texts. The study investigates the transfer learning approach on Transformer-based Language models, abstractive text summarization approach, and their application to the domain of Sustainable Development with the goal to determine SDGs representation in scientific publications using the text summarization technique. To achieve this, the traditional transfer learning framework was expanded so that: (1) the relevance of textual documents to specified text can be evaluated, (2) neural language models, namely BART and T5, were selected, and (3) 8 text similarity measures were investigated to identify the most informative ones. Both the BART and T5 models were fine-tuned on an acquired domain-specific corpus of scientific publications extracted from Scopus Elsevier database. The relevance of recently published works to an SDG was determined by calculating semantic similarity scores between each model generated summary to the SDG’s description. The proposed framework made it possible to identify goals that dominated the developed corpus and those that require further attention of the research community.
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 | 0.004 | 0.013 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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".