Designed for non-programmers
A web-based interface lowers the technical barrier for researchers who need to inspect and interpret qualitative text at scale.
Multilingual NLP · Vietnamese & English · Free-text analytics
FreeTxt-Vi
A benchmarked Vietnamese-English toolkit for segmentation, sentiment, summarisation and corpus exploration.
FreeTxt-Vi brings corpus linguistics and modern NLP together in an accessible web-based environment for building, exploring and interpreting bilingual free-text collections, without requiring users to write code.
About FreeTxt-Vi
Making bilingual free-text analysis more accessible
Qualitative responses from surveys, questionnaires, interviews and feedback forms contain valuable information, but analysing large collections manually is slow and difficult to scale. FreeTxt-Vi provides an integrated environment for exploring Vietnamese-English text with corpus methods and NLP.
FreeTxt-Vi extends the Welsh-English FreeTxt research platform to Vietnamese-English data. The project develops a unified bilingual workflow for preparing text, analysing sentiment, generating summaries and exploring patterns in free-text collections.
The research places particular emphasis on Vietnamese, a widely spoken language that remains comparatively underrepresented in NLP resources and evaluation. The toolkit is designed for researchers and practitioners who work with qualitative text but may not have specialist programming or machine-learning expertise.
From raw responses to interpretable evidence
The aim is to make bilingual text analysis reproducible and practical by combining language-aware preprocessing, benchmarked NLP components, corpus exploration and interactive visualisation within a single workflow.
Bilingual by design
Vietnamese and English are handled within one research workflow rather than as disconnected monolingual pipelines.
Evaluated components
The project benchmarks its core language-processing components rather than treating the toolkit as an unevaluated application layer.
Toolkit capabilities
A unified environment for analysing bilingual free text
FreeTxt-Vi combines established corpus-linguistic methods with transformer-based NLP so users can move between close inspection of text and scalable automatic analysis.
Vietnamese-English segmentation
A language-aware segmentation pipeline prepares Vietnamese and English text for downstream analysis, including a hybrid VnCoreNLP and Byte Pair Encoding strategy for Vietnamese.
VnCoreNLPBPEBilingual preprocessing
Sentiment analysis
Transformer-based sentiment modelling supports scalable interpretation of opinions and attitudes expressed across Vietnamese-English free-text collections.
Transformer modelsFine-tuningEvaluation
Abstractive summarisation
LLM-based summarisation helps users condense larger bodies of responses into shorter, interpretable summaries while retaining the key content of the source collection.
Qwen2.5LLMsBilingual summaries
Corpus exploration & visualisation
Concordancing, keyword analysis, word-relation exploration and interactive visualisation support deeper inspection of recurring themes, language patterns and relationships.
ConcordanceKeywordsVisual analytics
Research design
From bilingual text collection to interpretable analysis
The project treats FreeTxt-Vi as a set of language-processing components that can be individually evaluated, integrated and used within a broader corpus-analysis workflow.
CollectImport bilingual free-text responses
SegmentPrepare Vietnamese-English text
AnalyseSentiment and corpus-level patterns
SummariseGenerate concise abstractive summaries
Segmentation benchmark
Tests the reliability of the bilingual preprocessing pipeline, including Vietnamese word segmentation and tokenisation choices.
Sentiment benchmark
Measures classification quality against established baselines to assess whether the toolkit provides robust bilingual sentiment analysis.
Summarisation benchmark
Evaluates the LLM-based summarisation component in Vietnamese and English rather than relying solely on qualitative demonstration.
Project outputs
Research, software and reusable methodology
FreeTxt-Vi produces both an applied bilingual analysis environment and research evidence on the language-processing components that underpin it.
LREC 2026Project publication
FreeTxt-Vi: A Benchmarked Vietnamese-English Toolkit for Segmentation, Sentiment, and Summarisation
Language Resources and Evaluation Conference (LREC), 2026. The paper presents the FreeTxt-Vi architecture and a three-part evaluation of segmentation, sentiment analysis and summarisation across Vietnamese and English.
Open-source toolkit
A web-based environment for creating and analysing Vietnamese-English text collections through corpus and NLP methods.
Benchmark methodology
A reproducible evaluation setup for testing segmentation, sentiment and summarisation as distinct components of a bilingual NLP toolkit.
Vietnamese language resources
The project contributes practical methods, evaluation evidence and reusable workflows for a language that remains underrepresented in mainstream NLP.
Research team
FreeTxt-Vi team
The project connects multilingual NLP, corpus linguistics and language-resource expertise across VinUniversity, Lancaster University and Cardiff University.
Mo El-Haj
Principal Investigator
VinUniversity, Vietnam
Multilingual NLP, low-resource language technology and LLM-based text analysis.
Profile
Paul Rayson
Co-Investigator
Lancaster University, UK
Corpus linguistics, NLP evaluation and language-resource development.
Profile
Dawn Knight
Co-Investigator
Cardiff University, UK
Corpus linguistics, digital language resources and the original FreeTxt research programme.
Profile
Nguyen Huy Hung
Research Assistant
VinUniversity, Vietnam
Vietnamese-English NLP pipeline development, experimentation and benchmarking.
LinkedInCollaboration
Institutions and research groups
FreeTxt-Vi develops Vietnamese-English language technology through collaboration across VinUniversity, UCREL at Lancaster University and Cardiff University.
Why it matters
Supporting research with multilingual qualitative data
FreeTxt-Vi is designed for settings where rich qualitative text is common but difficult to process at scale. By reducing technical barriers, it supports reproducible bilingual analysis and the development of Vietnamese NLP resources.
Education & social researchAnalyse open-ended survey responses, feedback and qualitative research data across Vietnamese and English.
Digital humanities & cultural heritageExplore bilingual collections using corpus methods, visualisation and language-aware NLP components.
Inclusive language technologyStrengthen tools and evaluation practices for Vietnamese within multilingual NLP research.
Get in touch
Project lead: Dr Mo El-Haj
Address: CECS, VinUniversity, Hanoi, Vietnam
Email: elhaj.m@vinuni.edu.vn
Research group: NLP @ VinUniversity