What we learned cleaning a Sindhi text corpus before training a tokenizer
Low-resource language models live or die by data quality. Lessons from auditing a large Sindhi corpus, filtering it, and training a 32K Unigram tokenizer on the…
Helping businesses grow, scale and digitize.
The Text Tech Solutions blog
What we learn from building websites, automations, language datasets and software products, written for business owners, researchers and teams.
Low-resource language models live or die by data quality. Lessons from auditing a large Sindhi corpus, filtering it, and training a 32K Unigram tokenizer on the…
A practical guide to automating customer conversations on WhatsApp: the replies worth automating first, when you need the Business Platform, and how to keep a human…
Where money and fuel go missing in a paper-based petrol station, a step-by-step checklist for moving to software, and how PetroOS is being built to make…
Work with us
Websites, automation, language AI or custom software. Tell us the problem and we will suggest a practical way forward.
Newsletter
Be the first to hear about new product launches, research insights from our language AI work, and new articles on the blog.
Occasional emails only, no spam. We will ask you to confirm by email, and you can unsubscribe at any time.