Empirical evaluation of feature projection algorithms for multi-view text classification
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
This study aims to propose (i) a multi-view text classification method and (ii) a ranking method that allows for selecting the best information fusion layer among many variations. Multi-view document classification is worth a detailed study as it makes it possible to combine different feature sets into yet another view that further improves text classification. For this purpose, we propose a multi-view framework for text classification that is composed of two levels of information fusion. At the first level, classifiers are constructed using different data views, i.e. different vector space models by various machine learning algorithms . At the second level, the information fusion layer uses input information using a features projection method and a meta-classifier modelled by a selected machine learning algorithm. A final decision based on classification results produced by the models positioned at the first layer is reached. Moreover, we propose a ranking method to assess various configurations of the fusion layer. We use heuristics that utilise statistical properties of F-score values calculated for classification results produced at the fusion layer. The information fusion layer of the classification framework and ranking method has been empirically evaluated. For this purpose, we introduce a use case checking whether companies’ domains identify their innovativeness . The results empirically demonstrate that the information fusion layer enhances classification quality. The Friedman’s aligned rank and Wilcoxon signed-rank statistical tests and the effect size support this hypothesis. In addition, the Spearman statistical test carried out for the obtained results demonstrated that the assessment made by the proposed ranking method converges to a well-established method named Hellinger - The Technique for Order Preference by Similarity to Ideal Solution (H-TOPSIS). Thus, the proposed approach may be used for the assessment of classifier performance.
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.
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 | 0.001 | 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 | 0.000 |
| Open science | 0.001 | 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 it