Evaluation of Deep Learning Context-Sensitive Visualization Models
Bibliographic record
Abstract
The introduction of Transformer neural networks has changed the landscape of Natural Language Processing (NLP) during the recent years. These models are very complex, and therefore hard to debug and explain. In this context, visual explanation became an attractive approach. The visualization of the path that leads to certain outputs of a model is at the core of visual explanation, as this illuminates the features or parts of the model that may need to be changed to achieve the desired results. In particular, one goal of a NLP visual explanation is to highlight the most significant parts of the text that have the greatest impact on the model output. Several visual explanation methods for NLP models were recently proposed. A major challenge is how to compare the performances of such methods since we cannot simply use the usual classification accuracy measures to evaluate the quality of visualizations. We need good metrics and rigorous criteria to measure how useful the extracted knowledge is for explaining the models. In addition, we want to visualize the differences between the knowledge extracted by different models, in order to be able to rank them. In this paper, we investigate how to evaluate explanations/visualizations resulted from machine learning models for text classification. The goal is not to improve the accuracy of a particular NLP classifier, but to assess the quality of the visualizations that explain its decisions. We describe several methods for evaluating the quality of NLP visualizations, including both automated techniques based on quantifiable measures and subjective techniques based on human judgements.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".