# A Character-Level Neural Approach to Sinhala Sandhi Splitting

> Source: <https://aclanthology.org/2026.findings-aacl.14/>
> Published: 2026-10-31 00:00:00+00:00

**Key takeaway:** A bidirectional LSTM encoder with a unidirectional LSTM decoder reaches 68.40% exact-match and 82.08% character-level accuracy on hard Sinhala Sandhi cases and 94.00% exact match on affixational cases, with bidirectional encoding the largest single contributor and native Sinhala script outperforming

##### Abstract

Sinhala Sandhi splitting recovers the constituent words or morphemes hidden inside a phonologically merged surface form. The task is important for Sinhala NLP because Sandhi obscures lexical boundaries, but no prior published work has established a neural benchmark for Sinhala Sandhi splitting. We present a character-level sequence-to-sequence study based on SandhiLex, using native Sinhala Unicode input and evaluating recurrent encoder-decoder models for affixational and more complex lexicalized, derivational, and etymological Sandhi. A bidirectional LSTM encoder with a unidirectional LSTM decoder yields the best results on the hard subset, achieving 68.40% exact-match accuracy and 82.08% character-level accuracy. On the full affixational subset, the same model achieves 94.00% exact-match accuracy. Ablations show that bidirectional encoding is the largest contributor to performance, while native Sinhala script improves exact match accuracy over romanized input. Qualitative analysis indicates that many errors are near misses involving boundary adjacent characters or plausible but incorrect phonological substitutions. These results establish an empirical baseline for Sinhala Sandhi splitting and identify data scale, Sandhi type conditioning, and attention-based decoding as the main directions for future work.
- Anthology ID:
- 2026.findings-aacl.14
- Volume:
- [Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026](https://aclanthology.org/volumes/2026.findings-aacl/)
- Month:
- November
- Year:
- 2026
- Address:
- Hengqin, China
- Editors:
- [Yue Zhang](https://aclanthology.org/people/yue-zhang/unverified/) ,[Ruifeng Xu](https://aclanthology.org/people/ruifeng-xu/) ,[Ayu Purwarianti](https://aclanthology.org/people/ayu-purwarianti/) ,[Tong Rong](https://aclanthology.org/people/tong-rong/unverified/) ,[Jun Suzuki](https://aclanthology.org/people/jun-suzuki/)
- Venue:
- [Findings](https://aclanthology.org/venues/findings/)
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 257–266
- Language:
- URL:
- [https://aclanthology.org/2026.findings-aacl.14/](https://aclanthology.org/2026.findings-aacl.14/)
- DOI:
- Cite (ACL):
- Yasas Ekanayaka and Deshan Sumanathilaka. 2026. [A Character-Level Neural Approach to Sinhala Sandhi Splitting](https://aclanthology.org/2026.findings-aacl.14/) . In*Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026* , pages 257–266, Hengqin, China. Association for Computational Linguistics.
- Cite (Informal):
- [A Character-Level Neural Approach to Sinhala Sandhi Splitting](https://aclanthology.org/2026.findings-aacl.14/) (Ekanayaka & Sumanathilaka, Findings 2026)
- PDF:
- [https://aclanthology.org/2026.findings-aacl.14.pdf](https://aclanthology.org/2026.findings-aacl.14.pdf)
