| Title: |
SATLab at BAREC Shared Task 2025: Optimizing a Language-Independent System for Fine-Grained Readability Assessment |
| Authors: |
Bestgen, Yves; The Third Arabic Natural Language Processing Conference @ EMNLP 2025 |
| Contributors: |
UCL - SSH/IPSY - Psychological Sciences Research Institute |
| Publication Year: |
2025 |
| Collection: |
DIAL@USL-B (Université Saint-Louis, Bruxelles) |
| Description: |
This paper presents SATLab’s participation in the BAREC shared task on estimating the read- ability of sentences and documents. The pro- posed system is based on character n-grams fed into a support vector regression. A pro- cedure is then applied to try to optimize the Quadratic Weighted Kappa, the main chal- lenge measure, by tuning the decision thresh- olds used to transform continuous values into ordered categories. Performance is signifi- cantly lower than that of the best systems, but nevertheless superior to that of several deep learning approaches. |
| Document Type: |
conference object |
| Language: |
English |
| Relation: |
boreal:306535; https://hdl.handle.net/2078.1/306535 |
| Availability: |
https://hdl.handle.net/2078.1/306535 |
| Rights: |
info:eu-repo/semantics/openAccess |
| Accession Number: |
edsbas.FC8783C0 |
| Database: |
BASE |