Nepal's Conversational AI Research Agenda: A Comprehensive Framework for 2025
Background
Conversational Artificial Intelligence (AI) — the technology of enabling a machine to understand and respond to human language — is revamping global communication.
For Nepal, this technology presents a dual prospect and responsibility: it can take critical services to the farthest reaches of the country but only if piloted with Nepali language, data and infrastructure in mind.
Platforms such as Sajedar, Nepal's AI chatbot innovation frontier, are examples of how locally trained conversational agents can help fill the divide between available knowledge and access in domains like agriculture, health and education.
However, developing strong Nepali chatbots is confronted with a low-resource language problem such as scarcity of annotated data, morphologically rich languages and dialectal variation.
Recent initiatives such as NepBERTa, the National AI Policy 2082 and the Digital Nepal Framework demonstrate the country's commitment to strengthen its capabilities in Artificial Intelligence.
This research agenda details 40 concrete questions to inform universities, start-ups and researchers — not excluding those working in partnership with Sajedar - as it builds a united AI strategy for the country.
Foundational NLP for Nepali
1. Annotated Legal Corpus Development
Creating domain-specific data sets for Nepali LegalTech and civic chatbots supported by Sajedar.
2. CS Mood Estimation
Modelling sentiments in Nepali-English social text towards conversational agents like Sajedar bots.
3. NER with NepBERTa
Improved recognition of names and locations in Nepali for Sajedar-based knowledge retrieval.
4. Comparison among Tokenization Approaches
A Study on Modelling of Subwords for a Morphologically Rich Nepali Language.
5. Morphological Analyzer
Building Blocks of grammar for better chatbot comprehension.
6. Code-Switching Detection Framework
Recognition of multilingual dialogue breaks for hybrid Sajedar chat interfaces.
7. Dialect Mapping and Inclusivity Modeling
Representing local speech for better chatbot access nationwide.
8. NepGLUE: A benchmark nepali NLP
By the academic and Sajedar researchers.
Advanced Models and Low-Resource Techniques
9. Monolingual vs. Multilingual Performance Study
A Nepali First Chatbot Framework for Sajedar using NepBERTa and mBERT.
10. Cross-Lingual Transfer from Hindi
Using linguistic proximity for rapid Intent classification in Nepali SaaS bots.
11. Back-Translation and Synonym Replacement for Evaluation
Augmenting low-resource conversational AI data in Nepal.
12. Retrieval-Augmented Generation (RAG)
Enabling trustable Medical Chatbots in Sajedar's Health AI pilots.
13. Few-Shot Dialogue Act Learning
Towards Reducing Training Data for New Domains in Chatbots.
14. Dialogue State Tracking with Memory Networks
Developing a conversational flow-awareness in the low resource multilingual Sajedar bots.
15. Attention for SOV Structure
Custom attention layers tailored by the Nepali language syntax.
16. Nepali News Unsupervised Topic Modeling
Data Structuring to better the knowledge base of a chatbot from news data.
Sector-Specific Chatbot Applications
Agriculture (KrishiTech)
17. Multimodality chatbot for pest and disease recognition
Via Sajedar's AI-vision pipeline.
18. Market price information agent
To send daily crop rates on WhatsApp or SMS.
19. A climate-resilient farm adviser
Using local information from Nepal's agro zones.
20. A voice and SMS-based chatbot for low literate women farmers
With simple mobile handsets.
Tourism and Hospitality
21. Multi-lingual trekking assistant
For instant route/weather/safety info.
22. Interactive heritage guide
To tour UNESCO inscribed sites in Nepal with Sajedar's conversational framework.
23. Chatbot for Hotel Booking
To answer guest questions, and help independent mid-scale hospitality organizations.
Healthcare (Digital Swasthya)
24. Rural telemedicine consultations chatbot
For symptoms checking.
25. Maternal wellness informant
Adhered to the direction of the public health.
26. Private mental health chatbot
Based on the ethical AI principles from Sajedar.
Education (EdTech)
27. AI based adaptive learning chatbot
For secondary level mathematics in alignment with Nepalese curriculum.
28. Bot assistant for teachers
To automate school's communication.
29. A gamified language preservation chatbot
For indigenous languages built using Sajedar's conversational modules.
E-Governance and Finance
30. "Palika AI"- city chatbot
For enhancing public access to information.
31. Financial literacy chatbot
That teaches children about digital banking facilities like eSewa and Khalti.
32. Nepali Ecommerce Platform Chatbot
Done with Sajedar infrastructure.
Human-Centric, Multimodal, and Ethical AI
33. Nepali Automatic Speech Recognition (ASR)
For voice enabled Sajedar chatbots.
34. SMS/USSD based chatbot deployment
In lowbandwidth settings.
35. Auditing biases in Nepali language models
For gender and caste equity.
36. Cultural appropriateness evaluation
Tone and respect for a chatbot. We present here our model that supports the evaluation of the tone and respect level.
37. Privacy-preservation federated learning
For healthcare chatbot training.
38. Integration of Explainable AI (XAI)
For transparent healthcare decisions.
39. Multimodal interaction design
For agriculture bots using text, voice and image.
40. Usability evaluation of text and voice chatbots
With low-literate Nepali users.
Conclusion
These 40 themes provide a self-contained roadmap for Nepal's AI research community covering the theory to practical deployment.
Advances in foundational NLP allow better models, better models facilitate useful chatbots, ethical design fosters inclusivity and trust.
Anecdotally, collaborative ecosystems such as Sajedar can enable the translation of these ideas into practical realisations — connecting universities, government and industry towards a common national mission.
Each student project or thesis helping on in this agenda brings Nepal one step closer to an autonomous AI future where conversational systems mirror Nepali culture, language and diversity.
Keywords
Sajedar, Conversational AI Nepal, Nepali NLP, Chatbots Low-Resource Language Research, Digital Nepal, Ethical AI, EdTech, AgriTech, Public Service Chatbots.