--- title: "Morphological Parsing for Media Lengua — Stewart 2026" description: "The paper looks at a recurring problem in NLP for endangered languages: published models often never reach the communities they are meant to help…" authors: ["Jesse Stewart", "Olga Kriukova"] year: 2026 venue: "Proceedings of the Ninth Workshop on the Use of Computational Methods in the Study of Endangered Languages (ComputEL-9)" source: "https://aclanthology.org/2026.computel-1.1/" doi: "10.18653/v1/2026.computel-1.1" licence: "CC BY 4.0 (ACL Anthology)" licence_url: "https://creativecommons.org/licenses/by/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/stewart-2026-morphological-parsing-media-lengua-when/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/stewart-2026-morphological-parsing-media-lengua-when/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/7/) / Morphological Parsing for Media Lengua: When Accessibility Matters More Than State-of-the-Art # Morphological Parsing for Media Lengua: When Accessibility Matters More Than State-of-the-Art **Authors:** Jesse Stewart, Olga Kriukova · **Year:** 2026 · **Venue:** Proceedings of the Ninth Workshop on the Use of Computational Methods in the Study of Endangered Languages (ComputEL-9) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2026.computel-1.1/) · [PDF](https://aclanthology.org/2026.computel-1.1.pdf) · [DOI 10.18653/v1/2026.computel-1.1](https://doi.org/10.18653/v1/2026.computel-1.1) · [OpenAlex](https://openalex.org/W7166824061) **Citations:** 0 citations (openalex-snapshot · 2026-10-14) ## Summary The paper looks at a recurring problem in NLP for endangered languages: published models often never reach the communities they are meant to help, because running them requires servers and technical skills. Their case study is an endangered language of Ecuador, Media Lengua (ISO 639-3: mue), whose mixed character comes from pairing roots of Spanish origin with very regular agglutinative morphology inherited from Quichua. They built two morphological parsers. The first is a rule-based JavaScript tool (RB-parser) that runs entirely in a web browser. It relies on dictionary lookup, a fallback predictive segmentation step, IPA normalization of variable spellings, and interlinear glossing. The second is a CRF-parser intended mainly for linguists. The rule-based tool reached 98.6% accuracy, higher than the CRF model's 95.7% F1. It also needs no installation, and its errors can be corrected by editing transparent dictionary entries rather than retraining a model. The authors argue that when a language's morphology permits accurate rules, accessibility and maintainability should be weighed alongside benchmark scores. ## Key points - Lexical data come from the only published Media Lengua dictionary, which has 3210 lemmas and 1974 spelling variants. This is supplemented by 24 hours of glossed speech archived at AILLA. - The RB-parser draws on five JSON dictionaries totalling 9,965 entries: - verbal morphology (178) - non-verbal morphology (113) - verbs (1,848) - non-verbs (6,847) - 979 morpheme clusters taken from 12 hours of speech - Because the language has no standard orthography, input is normalized to IPA with 94 ordered regular expressions before parsing. - Most errors (1.4%) come from morphological homography, such as *-ta* marking both accusative and interrogative. The authors resolved 123 such cases with context-sensitive rules, lowering the error rate below 0.5%. - The CRF model was trained on only 399 word types and is available on GitHub (HelgaKr/ML_CRF). The RB-parser is hosted free online. No formal user study has been conducted yet. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/) ## Cite The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{stewart2026morphological, title = {{Morphological Parsing for Media Lengua: When Accessibility Matters More Than State-of-the-Art}}, author = {Stewart, Jesse and Kriukova, Olga}, booktitle = {Proceedings of the Ninth Workshop on the Use of Computational Methods in the Study of Endangered Languages (ComputEL-9)}, year = {2026}, doi = {10.18653/v1/2026.computel-1.1}, url = {https://doi.org/10.18653/v1/2026.computel-1.1} } ``` - [← previous MorphoFiltered-Gemini at MWE-2026 PARSEME 2.0 Subtask 1: Tackling LLM Overgeneration via Universal POS-based Constraints](https://md.research.elamur.ai/free/en/papers/moise-2026-morphofiltered-gemini-mwe-2026-parseme/) - [next → MORQA: Benchmarking Evaluation Metrics for Medical Open-Ended Question Answering](https://md.research.elamur.ai/free/en/papers/yim-2026-morqa-benchmarking-evaluation-metrics-medical/) - [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/7/) ## References 4 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Steven Bird. 2020. Decolonising speech and language technology. In Proceedings of the 28th international conference on computational linguistics, pages 3504– 3519. [find](https://scholar.google.com/scholar?q=Steven+Bird%2E+2020%2E+Decolonising+speech+and+language+technology%2E+In+Proceedings+of+the+28th+international+conference+on+computational+linguistics%2C+pages+3504%E2%80%93+3519%2E) [our card](https://md.research.elamur.ai/free/en/papers/bird-2020-decolonising-speech-language-technology/) 2. Michael Collins. 2002. Discriminative training methods for hidden Markov models: Theory and experiments with perceptron algorithms. In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP 2002), pages 1–8. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Michael+Collins%2E+2002%2E+Discriminative+training+methods+for+hidden+Markov+models%3A+Theory+and+experiments+with+perceptron+algorithms%2E+In+Proceedings+of+the+2002+Conference+on+Empirical+Methods+in+Natura) [our card](https://md.research.elamur.ai/free/en/papers/collins-2002-discriminative-training-methods-hidden-markov/) 3. Ewa Czaykowska-Higgins. 2009. Research models, community engagement, and linguistic fieldwork: Reflections on working within canadian indigenous communities. [find](https://scholar.google.com/scholar?q=Ewa+Czaykowska%2DHiggins%2E+2009%2E+Research+models%2C+community+engagement%2C+and+linguistic+fieldwork%3A+Reflections+on+working+within+canadian+indigenous+communities%2E) 4. Ben Foley, Josh Arnold, Rolando Coto-Solano, Gautier Durantin, Mark T. Ellison, Daan van Esch, Scott Heath, František Kratochvíl, Zara Maxwell-Smith, David Nash, Ola Olsson, Mark Richards, Nay San, Hywel Stoakes, Nick Thieberger, and Janet Wiles. 2018. Building speech recognition systems for languag… [find](https://scholar.google.com/scholar?q=Ben+Foley%2C+Josh+Arnold%2C+Rolando+Coto%2DSolano%2C+Gautier+Durantin%2C+Mark+T%2E+Ellison%2C+Daan+van+Esch%2C+Scott+Heath%2C+Franti%C5%A1ek+Kratochv%C3%ADl%2C+Zara+Maxwell%2DSmith%2C+David+Nash%2C+Ola+Olsson%2C+Mark+Richards%2C+Nay+San%2C+Hy)