Bibliographic record
Abstract
In practice, most approaches for text error detection and correction are based on a conventional domain-dependent background dictionary that represents a fixed and static collection of correct words of a given language and, as a result, satisfactory correction can only be achieved if the dictionary covers most tokens of the underlying correct text. Again, most approaches for text correction are for only one or at best a very few types of errors. The purpose of this thesis is to propose an unsupervised approach to detecting and correcting text errors, that can compete with supervised approaches and answer the following questions: Can an unsupervised approach efficiently detect and correct a text containing multiple errors of both syntactic and semantic nature? What is the magnitude of error coverage, in terms of the number of errors that can be corrected? We conclude that (1) it is possible that an unsupervised approach can efficiently detect and correct a text containing multiple errors of both syntactic and semantic nature. Error types include: real-word spelling errors, typographical errors, lexical choice errors, unwanted words, missing words, prepositional errors, article errors, punctuation errors, and many of the grammatical errors (e.g., errors in agreement and verb formation). (2) The magnitude of error coverage, in terms of the number of errors that can be corrected, is almost double of the number of correct words of the text. Although this is not the upper limit, this is what is practically feasible. We use engineering approaches to answer the first question and theoretical approaches to answer and support the second question. We show that finding inherent properties of a correct text using a corpus in the form of an n-gram data set is more appropriate and practical than using other approaches to detecting and correcting errors. Instead of using rule-based approaches and dictionaries, we argue that a corpus can effectively be used to infer the properties of these types of errors, and to detect and correct these errors. We test the robustness of the proposed approach separately for some individual error types, and then for all types of errors. The approach is language-independent, it can be applied to other languages, as long as n-grams are available. The results of this thesis thus suggest that unsupervised approaches, which are often dismissed in favor of supervised ones in the context of many Natural Language Processing (NLP) related tasks, may present an interesting array of NLP-related problem solving strengths.
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 | 0.004 | 0.022 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.005 | 0.003 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.003 | 0.004 |
| Open science | 0.004 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.002 |
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".