Every free tool started
as a research question.
How do you split a Tamil word into its parts by rule rather than by lookup? What makes two words rhyme well rather than merely rhyme? How should Tamil sound be carried into another script without losing it? KaReFo exists to answer questions like these, and then to give the answers away as tools anyone can use.
The questions we keep returning to.
Tamil words agglutinate: one word can carry what English needs a clause for. Our work isolates the rules that split a word into root and suffixes, which is what makes Piripori and Korpori possible.
Carrying Tamil sound faithfully into other scripts and back. Olingoa came out of this, and now moves text between fourteen writing systems.
Compacting syllables for Tamil text-to-speech, and mapping visemes to phonemes so a dubbed mouth and a Tamil word can agree with each other.
What words appear together in Tamil lyrics, how similes are formed, how a rhyme can be scored rather than only spotted. Emoni and Paadal both grew from this.
The statistics of a million-word dictionary: letter frequencies, genre classification, the shape of the vocabulary itself. This is the ground Chol stands on.
Scoring models and option-generation for word games, and text analysis of how teachers and learners actually talk to one another. Research that ends up as something a child wants to play.
How Tamil is best taught, and in what order: the sequence in which letters, sounds, words and grammar are met, how stroke order is decided, and how a learning outcome is written so that a book, a song and a game can all be built to serve it. This is the research the Payil curriculum and the Thambaa labs are built on.
Papers, conferences and journals.
Published work by KaReFo researchers and collaborators, in international conferences and journals, from 2013 onward. Filter by topic or search by author.
Looking for a full text? Write to inbox@karky.in and we will point you to the publisher's copy.
Research that became a way to teach.
Language research is only half the work. The other half is deciding what a five-year-old should meet first, and what a fifteen-year-old should be ready for. The Payil curriculum is our answer: a single ladder of Tamil language skills running from LKG to Level 12, with each rung defined by what a learner can do at the end of it, not by how many pages were covered.
A ladder, not a syllabus
Each level sets out its own learning outcomes across listening, speaking, reading, writing and word knowledge. Textbooks, e-books, songs, games and podcasts are then written to serve those outcomes, rather than the outcomes being written to justify a textbook.
Levels LKG through 5 are published and in classroom use. Levels 6 through 12 are in preparation. Each one takes a team of linguists, writers, illustrators and editors to carry it from an outcome list to a finished book.
Work done with others.
Some of the research only exists because another institution brought a need, a corpus or a platform to it.
Tamil Virtual Academy
Tamil part-of-speech and semantic-role tagging, built with the Tamil Virtual Academy, along with a Tamil chatbot. The tagger runs today as Kurivi in the Labs, where anyone can put a Tamil sentence in and see it labelled.
Sorkuvai
A Tamil etymology project tracing where words come from and how they changed. KaReFo's dictionary and morphology work feeds it, and Ari authors are pointed to it for Tamil technical terms.
Kindle Tamil Dictionary
Tamil dictionary data for Kindle, so that a reader can look a Tamil word up inside the book they are reading rather than leaving it. The same lexical work that powers Chol.
Karky Tamil Academy
KaReFo is the research wing; Karky Tamil Academy takes the results into schools as the Payil curriculum and the Thambaa Tamil Labs. Our research on how Tamil is read, written and learned goes straight into those classrooms.
These summaries describe the work as it stands in our own tools and records. If you are one of these partners and would like the description sharpened, write to inbox@karky.in.
Work that only a person can do.
Software can carry a dictionary. It cannot decide what a word means, or which words belong together. That part is still done by people reading, arguing and writing it down, and there is far more of it than our team can finish alone. Everyone starts with training. After that you are given real work on a live project, at whatever pace you can keep.
The dictionary
After training you are given a shared sheet and a set of words: a definition to write, a usage example to find, a technical term Tamil does not yet have a word for, an old word from a text no dictionary has recorded. You work on a department that suits what you know.
Suits anyone who reads Tamil closely — a teacher, a writer, a researcher, or someone who knows one field deeply and its vocabulary with it.
Volunteer for CholThe semantic web
Valai records how Tamil words relate to one another: what is a kind of what, what is part of what, what a thing is used for, what causes it. After training you connect words in the database along those relations. Every link has to be judged by someone who knows the language.
Suits people who enjoy precise distinctions — anyone drawn to taxonomy, to grammar, or to arguing about where one meaning ends and the next begins.
Volunteer for ValaiTools and internships
If you write software, there are always Tamil language tools waiting to be built. We take school and college students on internships too, and the work is not token work: parts of several tools running on this site were written by interns, and the eN number tool had a contributor who was in the tenth standard at the time.
Suits developers of any age, and students looking for an internship with something real at the end of it.
Build or intern with usAll three go to inbox@karky.in. Tell us which project, what you study or do, and roughly how much time you could give. There is no minimum. The work is unpaid, and what you add stays free for everyone, with the rest of the labs.
Bring us a hard question.
We collaborate with universities, government bodies and companies on Tamil language technology, and we supervise student research. If you have a corpus that needs analysing, a Tamil problem that needs a tool, or a project that needs a research partner, we would like to hear about it.