XLM-R Finnish QA Extractor
This XLM-RoBERTa Base token classifier targets Finnish question-answer extraction. Its outputs distinguish QUESTION, ANSWER and O, rather than generating a response.
The trained task
The card describes manual Finnish annotations, ChatGPT annotations and semi-synthetic LFQA-based data. It uses a 512-token training limit and recommends raw token predictions followed by its linked post-processing script. Word-level aggregation alone is not the full published extraction procedure.XLM-R Finnish QA Extractor: model card and trained artifactsXLM-R Finnish QA Extractor: output labels and configuration
Practical assessment
Practical guidance. Keep the token offsets and the post-processing rules with each extracted pair. The noncommercial license differs from the selector's terms, so combining the two stages does not produce a single shared license.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | xlmr-qa-extraction-fi |
| Output labels | QUESTION, ANSWER and O.XLM-R Finnish QA Extractor: output labels and configuration |
| Training learning rate | 1e-5.XLM-R Finnish QA Extractor: model card and trained artifacts |
| License | CC BY-NC-SA 4.0.XLM-R Finnish QA Extractor: model card and trained artifacts |
Related models
- XLM-R English QA Extractor
TurkuNLP · QA web mining
The English QA Extractor is a separate token classifier for locating question and answer spans.
Launch date not confirmedDownloadable - XLM-R QA Register Selector
TurkuNLP · QA web mining
The QA Register Selector identifies documents likely to contain question-and-answer material.
Launch date not confirmedDownloadable