Human Emotion and Sentiment in Natural Language Understanding and Generation using Large Language Models with Limited to No Labeled Data
Bibliographic record
Abstract
Natural Language Processing (NLP) aims to utilize computational resources to comprehend and generate human language. Emotion and sentiment are integral parts of human beings, and they are often reflected in human language. Consequently, these two closely related ideas are of paramount importance to NLP. In this thesis, we focus on several NLP tasks related to human emotion and sentiment. Particularly, we focus on the domains of Sentiment Analysis and Emotion-Cause Analysis (ECA). Like most other NLP tasks, machine learning technologies are frequently leveraged to perform various NLP tasks in these two domains. A common challenge in applying machine learning technology to context-dependent tasks like Sentiment Analysis is that they require a large amount of labeled data to develop a performant model. In this thesis, we develop several techniques leveraging Transformer-based large language models (LLMs) to perform various NLP tasks within these two domains in a limited to no labeled data setting. Specifically, we devise two technical architectures to perform multi-class Sentiment Analysis with limited labeled data. We introduce two new NLP tasks within the domain of ECA, which are also the first Natural Language Generation (NLG) tasks in this domain. We devise technical solutions to perform these NLG tasks, one with limited labeled data, and the other with no labeled data. We publish a new dataset for one of these novel NLG tasks. Lastly, we propose leveraging conversational LLMs for the automatic evaluation of open-ended NLG tasks, which also does not require any new training or labeled data.
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) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| 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".