LiFE Case Studies

Language documentation / Active project

Project BoLI: empowering Indigenous languages of India

BoLI builds linguistic resources and technologies for Indigenous languages of India, bringing linguists, language experts, and NLP researchers into a shared preservation and revitalization effort.

Project: BoLI
Partner: UnReaL-TecE collaborative network
Location: India
Year: 2026

Challenge

Indigenous languages need durable linguistic resources that can support language preservation, language documentation, grammar writing, child language acquisition research, and future AI applications.

LiFE role

LiFE supports the project as the workspace for recording-oriented documentation workflows: audio management, transcription in IPA and native scripts, translation, morphological glossing, review, and reusable exports.

Impact

The project positions language resources as both archival and technological infrastructure, helping Indigenous languages become more visible in research, education, and language technology.

Outputs

Audio recordings representing Indigenous language speech Transcriptions in IPA and native scripts Interlinear morphological glosses Free English translations for wider accessibility

Signals

Indigenous languages Audio recordings IPA + native script Interlinear glossing

documentation indigenous languages audio ipa glossing translation

Original project page

Speech technology / Funded project

SpeeD-IL: bridging the speech technology gap in India

SpeeD-IL creates speech datasets and models for underrepresented Indian languages so speech technology can extend beyond commercially dominant languages.

Project: SpeeD-IL
Partner: Mission Bhashini, Government of India
Location: India
Year: 2026

Challenge

Many Indian languages have limited commercial speech support because there are not enough curated speech datasets, especially for non-scheduled Indo-Aryan and Dravidian languages and scheduled Tibeto-Burman and Austro-Asiatic languages.

LiFE role

LiFE supports collection, transcription, validation, data curation, model-ready export, and iterative AI workflow management for speech dataset creation.

Impact

The project is designed to make speech technology more inclusive by enabling accessible datasets and open models for underrepresented language communities.

Outputs

Transcribed speech datasets for Indian languages Phone sets for accurate phonetic representation Baseline ASR systems Language models and pre-trained models for further fine-tuning

Signals

100 hours per language target 10+ languages per family ASR baselines Open model licensing

speech datasets asr language models open data tibeto-burman indo-aryan

Original project page

Education research / Consultancy and corpus project

Multilingual education corpus for government schools

This project supports corpus development for multilingual education, teaching, learning, and assessment in India's government schools in the context of the National Education Policy.

Project: Multilingual Education
Partner: University of Cambridge
Location: India
Year: 2026

Challenge

Classroom language practices need careful corpus creation, transcription, translation, glossing, and technical support before they can be studied and shared through research repositories.

LiFE role

LiFE provides the in-house application layer for corpus management, AI-assisted transcription, translation, glossing, export to CHAT format, and ongoing support for storage, retrieval, and analysis.

Impact

The project helps researchers study translanguaging practices and supports evidence-driven multilingual education workflows for government-school contexts.

Outputs

Classroom audio recordings Transcribed, translated, and glossed corpus data CHAT-format export for CHILDES publication Application support for corpus creation and analysis

Signals

50+ hours classroom audio CHAT export CHILDES corpus target Translanguaging practices

education classroom recordings translanguaging childes translation glossing

Original project page

Sociolinguistic corpus / Research project

Swiss Tamil: mapping heritage language structure

The Swiss Tamil project creates a detailed spoken Tamil corpus across South India, Northern Sri Lanka, and Switzerland to study heritage language structure through sociolinguistic cues.

Project: Swiss Tamil Project
Partner: Universite de Lausanne, Switzerland
Location: South India, Northern Sri Lanka, and Switzerland
Year: 2026

Challenge

The project needs comparable, validated speech data from different Tamil-speaking communities, with transcription, translation, glossing, and sociolinguistic analysis of variation and code-switching.

LiFE role

LiFE's MATra Lab workflow supports speech data processing, AI-assisted transcription and translation, interlinear glossing, and human validation before data is used for analysis and model improvement.

Impact

The project connects heritage-language documentation with computational improvement, producing data that can support sociolinguistic research and stronger Tamil language technologies.

Outputs

Human-validated dataset of different Tamil varieties Transcription, translation, validation, and glossing outputs Data for fine-tuning Tamil ASR and translation models Evidence for language variation, code-switching, and language-contact analysis

Signals

200+ hours speech target Three communities Human validation ASR + translation fine-tuning

tamil heritage language speech corpus asr translation code-switching

Original project page

Computational morphology / Supported project

MorphGen: morphological generators for Indian languages

MorphGen generates inflectional wordforms and morphological features for more than 10 Indian languages, supporting synthetic dataset creation and system development.

Project: MorphGen
Partner: Oxford Languages
Location: India
Year: 2026

Challenge

Indian-language NLP systems need large, transparent, morphologically rich resources, but creating complete inflectional datasets manually is slow and difficult to scale.

LiFE role

LiFE supports the broader resource pipeline around lexical data, linguistic annotation, export, and downstream model or dataset preparation for rule-based and AI-assisted workflows.

Impact

The rule-based approach gives teams accuracy, flexibility, efficiency, and transparency while making morphological resource creation easier to adapt across languages.

Outputs

Inflectional wordforms from lemmas or wordforms Complete morphological features for generated forms Rule-application output for transparency Synthetic datasets for NLP model training

Signals

10+ Indian languages Inflectional wordforms Morphological features Rule transparency

morphology synthetic data nlp rule based indian languages lexical resources

Original project page

Project stories built around LiFE.

BoLI

SpeeD-IL

Multilingual Education

Swiss Tamil Project

MorphGen