Linguistic Landscape and National Infrastructure

Which feature best characterizes many Indian languages like Marathi and Sanskrit in NLP contexts?

• Isolating morphology with minimal affixes

• Agglutinative morphology with complex suffixation

• Analytic syntax with few inflections

• Polysynthetic morphology with noun incorporation

What does the “low-resource paradox” refer to for Indian languages?

• Many speakers yet scarce high-quality corpora

• Few native speakers but abundant corpora

• Rich annotations but limited tokenizers

• Abundant speech data but no text data

Why do standard tokenizers often fail for Indic scripts in agglutinative languages?

• They assume fixed-width characters only

• They ignore whitespace completely

• They rely on grammar rules not used

• They struggle with long affixed word forms

Which challenge requires robustness across script boundaries for Indian NLP models?

• Entity linking to global knowledge

• Prosody modeling in poetry

• Code-mixing and transliteration practices

• Dialect clustering across states

Which initiative operates as a unified ecosystem and promotes the ‘Bhasha Daan’ concept?

• ULCA under MeitY

• AI4Bharat at IISc

• IndicBERT consortium

• Mission Bhashini

ULCA (Universal Language Contributions API) is primarily described as which of the following?

• A crowdsourcing app for mobile keyboards

• A proprietary ASR engine for government use

• A closed benchmark leaderboard for models

• The largest open-source database for Indian language datasets

Which pair correctly matches organization and contribution in India’s NLP ecosystem?

• AI4Bharat — Bhasha Daan volunteer portal

• Mission Bhashini — Airavata translation model

• ULCA — IndicBART pretraining framework

• AI4Bharat — IndicTrans2 and Airavata models

What is a practical modeling response to morphological richness in Indian languages?

• Translate all text to English first

• Limit vocabulary to top fifty thousand words

• Adopt subword tokenization like BPE or Unigram

• Use character-level RNNs only

Which statement best describes Small Language Models (SLMs) in the 2026 trend toward “Right‑Sized AI”?

• They are compressed chatbots that ignore tokenization constraints

• They are trillion‑parameter general models for all tasks

• They are efficient domain models optimized for specific use cases

• They are rule‑based systems replacing neural networks entirely

In a Mixture of Experts (MoE) architecture used by models like Sutrim, what happens during a single query?

• Only a fraction of experts activate to process tokens

• All experts activate for every input equally

• Experts train but never participate in inference

• Experts activate randomly without learned routing