Bibliographic record
Abstract
Distributed representations, predominantly acquired via neural networks, have been applied to natural language processing tasks including speech recognition and machine translation with a success comparable to sophisticated state-of-the-art algorithms.The present thesis offers an investigation of the application of such representations to information extraction.Specifically, I explore the suitability of applying shallow distributed representations to the automatic terminology extraction task, as well as the bridging reference resolution task.I created a dataset as a gold standard for automatic term extraction in the mathematical education domain.I carefully assessed the performance of the existing terminology extraction methods on this dataset.Then, I introduce a novel method for automatic terminology extraction for one word terms, and I evaluate the performance of the novel algorithm in various terminological domains.The introduced algorithm leverages the distributed representation of words from the local and global perspectives to encode syntactic, semantic, association, and frequency information at the same time.Furthermore, this novel algorithm can be trained with a minimal number of data points.I show that the algorithm is robust to the change of domain, and that information can be transferred from one technical domain to another, leveraging what we call anchor words with consistent semantics shared between the domains.As for the bridging reference resolution task, a dataset is built on the letter portion of the Open American National Corpus and I compare the performance of a preliminary method against a majority class baseline.Eshaghi for their support and sacrifice that first and foremost made this endeavor possible.I could travel, as a result, to the other side of the world in pursuit of knowledge and to contribute to the scientific community.Amir and Khadijeh, without your great help and sacrifice this journey would have never started.Thank you!
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.005 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.003 |
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 source (direct Gemma or distilled Codex), 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".