Semi-supervised learning and opinion-oriented information extraction
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
Recently, information extraction (IE) has attracted much attention in the field of natural language processing (NLP). Technology-users are no longer satisfied with factual information extraction, so researchers have become attracted to the study of opinion-oriented IE. However, past investigations of opinion-oriented IE have used supervised learning algorithms that require large amounts of data and manually labeled training corpora - time and money restrictions that are widely recognized as a bottleneck in the use of machine learning algorithms for opinion-oriented IE. Attempting to break this bottleneck, researchers are turning to semi-supervised learning. This thesis thus focuses on three types of semi-supervised learning algorithms, namely, self-training, co-training, and graph-based methods. For self-training, we apply the value difference metric (VDM) as the selection metric and use naive Bayes and decision tree algorithms as underlying classifiers. For co-training, we propose an unsymmetrical co-training algorithm that combines an EM classifier and a self-training classifier together within an unsymmetrical structure without splitting the attribute set. For graph-based methods, we put forward a probability propagation algorithm based on the instance-attribute graph, for which there are two kinds of nodes, i.e., instance nodes and attribute nodes; and two types of messages, i.e., instance node messages and attribute node messages. The goal of the probability propagation algorithm is to propagate messages between nodes in order to balance the global and local situations and, ultimately, to smooth the graph. According to the experimental results, the new techniques and novel algorithms achieve better performances than do their corresponding opponents. Furthermore, several opinion-oriented IE tasks have been tackled by the semi-supervised learning algorithms in this thesis. We mainly focus on three tasks, that is, sentence subjectivity classification, contextual polarity recognition, and opinion entity identification. Self-training is used to solve the sentence subjectivity classification, co-training is used to deal with the contextual polarity recognition, and graph-based methods are used to tackle opinion entity recognition. The experiments have been designed to compare the performances of corresponding algorithms to their corresponding tasks. The results show that semi-supervised learning algorithms are suitable for the tasks of opinion-oriented IE.
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 | 0.000 | 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 | 0.002 |
| 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 it