# Digital Revival of Ingush: How PaydaDosh is Building a Dictionary, Corpus, and AI Models
Project PaydaDosh is an open linguistic platform that combines a dictionary, parallel corpus, and AI tools to support the Ingush language. The collected database includes over 66,000 dictionary entries and tens of thousands of aligned sentences, making it the largest resource for studying and preserving the language. Ingush, which belongs to the Nakh languages of the Caucasian family, is spoken by about 400–500 thousand people, but its digital presence was minimal until recently.
Platform Components: From Dictionary to AI Assistant
PaydaDosh provides developers and linguists with a comprehensive set of resources:
- Dictionary — 66,524 dictionary entries with translations, grammatical tags, word forms, and audio examples.
- Parallel Corpus — sentences from real texts aligned with Russian translations. Includes the historical novel Magas the Blessed (34,156 sentences), folklore (9,477 sentences), and translations of classics (Pushkin, Turgenev, and others).
- Thematic Collections — 2,156 proverbs divided into 16 categories (work, family, honor, etc.), and 171 parables with parallel translations.
- Phrasebook — ready-made phrases grouped by situations for practical use.
- Community — "Questions" section where native speakers answer queries on grammar and translation, with authoritative responses marked.
- AI Assistant — automatically generates preliminary answers to questions using dictionary and grammar data.
- Mobile Apps — two Android apps (one offline) and a Telegram bot with data synchronization.
All resources are open and freely available, which is crucial for supporting low-resource languages in the digital age.
Data Sources: From Classic Dictionaries to Modern Collections
The dictionary is based on the "Ingush Language Dictionary Compendium" project (ghalghay.github.io), but PaydaDosh has significantly expanded it by integrating data from diverse sources. Among them:
- M. G. Uzhaev _Ingush-Russian Dictionary_ (1927)
- Z. K. Malsagov _Ingush Terminological Collection_ (1933)
- Johanna Nichols. _Ingush-English and English-Ingush Dictionary_ (2004), translated into Russian by Ahmed Bekov
- A. S. Kurkiev _Ingush-Russian Dictionary_ (2005)
- A.-M. M. Dudarov _Ingush Agriculture_ (2015)
- N. D. Kodzoev _Dictionary of Computer Terms_ (2016)
- Children's magazine "SelaIad" (294 issues)
Each dictionary entry includes a link to the source as a short code, ensuring transparency and allowing researchers to verify originals. This approach is especially important for academic work, where data accuracy is paramount.
Technical Implementation: Morphology, Corpus, and AI
Ingush features an ergative structure, complex case system, and verb classes, requiring specialized processing tools. PaydaDosh addresses these through:
Fuzzy search with morphological analysis. The algorithm recognizes word forms and prioritizes exact matches. For a language where a single base word can generate dozens of forms, this boosts search relevance. For example, a query for "kho" (I) returns all pronoun forms but highlights the base form.
Linking the corpus and dictionary. Every sentence in the parallel corpus is tied to dictionary entries. Clicking a word in the text takes users to its entry, while the word's card shows usage examples from the corpus. Search works on both Ingush and Russian text, simplifying work with parallel data.
AI integration in the Q&A system. When a new query arrives, the system generates a preliminary response based on existing data, lowering the entry barrier for users. Native speaker responses are marked separately, and voting promotes the most accurate ones.
Cross-platform synchronization. Search history, favorites, and user contributions sync across the website, Telegram bot, and mobile apps, creating a unified ecosystem for language work in any format.
User contribution mechanism. Registered users can add new words, examples, or audio recordings. All submissions go through editor moderation before publication, ensuring data quality.
Value of the Parallel Corpus for Computational Linguistics
The parallel corpus is a key resource for developing NLP tools. For Ingush, like most low-resource Caucasian languages, such materials were virtually nonexistent. PaydaDosh fills this gap by providing:
- 218 aligned sentences from The Tale of Igor's Campaign for syntactic analysis.
- 2,918 sentences from Pushkin's The Captain's Daughter, enabling study of translation strategies.
- Folklore and modern literature texts reflecting everyday speech.
These datasets are invaluable for training language models, especially in machine translation tasks. Standard approaches developed for major languages often fail with ergative systems and non-standard scripts, so custom corpora are essential for adapting technologies.
Development Plans: From Mobile Apps to Educational Tools
The PaydaDosh team is focusing on three directions:
- Corpus expansion — adding original Ingush texts, including modern prose and journalism.
- Deepening AI functionality — developing models that account for Ingush grammar specifics to improve automatic response accuracy.
- Educational tools — porting mobile quizzes and flashcards to the web platform, creating interactive lessons.
Special attention goes to supporting researchers: an API for programmatic data access and corpus analysis tools are planned. This will turn the platform into not just a reference but a lab for computational linguistics.
Key Takeaways
- Data Scale: 66,524 dictionary entries and 43,633 sentences in the parallel corpus form the foundation for NLP development.
- Technical Adaptation: Fuzzy search with morphology tackles Ingush's high inflectional productivity.
- Openness: All resources are free to use, critical for preserving languages with limited digital ecosystems.
- Community: Combining AI with moderated native contributions creates a sustainable growth model.
— Editorial Team
No comments yet.