A novel unsupervised fine-tuning method for text summarization, and highlighting the limitations of ROUGE score
Bibliographic record
Abstract
The limited availability of datasets for text summarization tasks and their similar characteristics (e.g. news articles) make it crucial to focus on unsupervised learning techniques to enable summarization across different domains. Moreover, since summarization produces text output, effective methods developed for news articles can be applied to other domains lacking sufficient labelled data. This study introduces a novel target selection process to be used as an unsupervised learning method for fine-tuning text summarization models with unlabeled data. The process involves two-steps: first, generating an extractive summary (Ext-Reference) from the article, and second, using an abstractive model to create a pool of candidate summaries. The most suitable summary (to be used as the target) is then selected by calculating the cosine similarity between the Ext-Reference’s embedding and each candidate’s embedding. Furthermore, this project underscores the limitations of the ROUGE score, which assigns a relatively low score to this method. However, extended analysis with various metrics, including using GPT-4 as a judge, demonstrates the effectiveness of this technique for fine-tuning models without a specific target reference. It highlights the importance of using a combination of metrics, like those included in the SumEvaluator package released alongside this paper. SumEvaluator package on Github: https://github.com/AlaFalaki/SumEvaluator .
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 | 0.003 | 0.013 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.004 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.003 |
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".