2023
- Altintas, E.
Development of Named Entity Recognition System Exploiting Language-Specific Features
Ph.D. Thesis. CoHE Thesis Center.- Summary: This study addresses the limitations of existing natural language processing models like BERT, which tokenize text independently of language-specific features. A language-specific approach for Turkish was developed, including a new tokenizer tailored to Turkish sound transformation features. Language models trained from scratch demonstrated significant improvements in training speed and named entity recognition performance compared to existing models like BERT.
2005
Altintas, E., Karsligil E., Coskun, V.
A New Semantic Similarity Measure Evaluated in Word Sense Disambiguation
The 15th Nordic Conference of Computational Linguistics (NODALIDA-2005), May 20-21 2005, Joensuu, Finland.- Electronic Edition: ACL Anthology
- Source Code: GitLab
- Summary: Presented a novel semantic similarity measure for word sense disambiguation (WSD), tested against other measures, and demonstrated improved accuracy in WSD tasks.
Altintas, E., Karsligil E., Coskun, V.
The Effect of Windowing in Word Sense Disambiguation
The 20th International Symposium on Computer and Information Sciences (ISCIS'05), Oct 26-28 2005, Istanbul, Turkey.- Electronic Edition: Springer DOI
- Summary: Investigated the impact of varying window sizes on the performance of WSD systems. Found that larger window sizes generally enhanced performance when combined with an effective weighting function.
Theses
Master’s Thesis
- Improving Word Sense Disambiguation Accuracy
- Summary: Focused on enhancing the accuracy of word sense disambiguation (WSD), building upon the foundational work presented in the two 2005 papers. This study advanced the understanding of techniques to improve WSD performance.
Ph.D. Thesis
- Development of Named Entity Recognition System Exploiting Language-Specific Features
(Details provided above under 2023 publication.)
Presentations
For more details, visit my Google Scholar profile or Academia.