Comparative Analysis of Deep Natural Networks and Large Language Models for Aspect-Based Sentiment Analysis
Bibliographic record
Abstract
Sentiment analysis is essential for comprehending public opinion, particularly when considering e-commerce and the expansion of online businesses. Early approaches treated sentiment analysis as a document or sentence-level classification problem, lacking the ability to capture nuanced opinions about specific aspects. This limitation was addressed by the development of aspect-based sentiment analysis (ABSA), which links sentiment to specific aspects that are mentioned explicitly or implicitly in the review. ABSA is relatively a new field of sentiment analysis and the existing models for ABSA face three main challenges, including domain-specificity, reliance on labeled data, and a lack of exploration into the potential of newer large language models (LLMs) such as GPT, PaLM, and T5. Leveraging a diverse set of datasets, including DOTSA, MAMS, and SemEval16, we evaluate the performance of prominent models such as ATAE-LSTM, flan-t5-large-absa, Deberta, PaLM, and GPT-3.5-Turbo. Our findings reveal nuanced strengths and weaknesses of these models across different domains, with Deberta emerging as consistently high-performing and PaLM demonstrating remarkable competitiveness for aspect term sentiment analysis (ATSA) tasks. In addition, the PaLM demonstrates competitive performance for all the domains that were used in the experiments including the restaurant, hotel, books, clothing, and movie reviews. Notably, the analysis underscores the models’ domain sensitivity, shedding light on their varying efficacy for both ATSA and ACSA tasks. These insights contribute to a deeper understanding of model applicability and highlight potential areas for improvement in ABSA research and development.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.004 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| 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 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".