Evaluation of Local Explainability Methods in Turkish Text Classification Tasks
Bibliographic record
Abstract
Complex transformer models have become popular in practice, however, they function as “black boxes”. Therefore, there is a growing need for the quantitative evaluation of existing explainability techniques. This evaluation becomes challenging when there are no ground-truth explanations available in the text data. We address this by exploring evaluation approaches for local explainability techniques in Turkish text classification tasks. We use BERT-based models, specifically BERTurk and TurkishBERTweet, and apply SHAP, LIME, and Integrated Gradients (IG). We evaluate the explainability techniques based on their ability to preserve the original text’s prediction probability when their most important tokens are used during inference. We employ evaluation approaches like Mean of Probabilities and Incremental Deletion to compare the explainability techniques with a baseline approach, aiming to measure the faithfulness of the explanations. Our results demonstrate that the gradient-based technique, IG, effectively identifies salient tokens that correlate with the class of the original text. We conclude that the IG method is effective in computing the saliency scores of the explanations when using an encoder-based model for Turkish text classification tasks. In understanding Turkish morphological complexity, it captures and highlights the nuanced contributions of context-dependent words and phrases.
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 | 0.012 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.000 | 0.001 |
| 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".