Case Study: Lingit achieves faster, cost-effective language model building for dyslexia support with Amazon Web Services

Lingit - Customer Case Study

Lingit builds assistive writing tools for people with dyslexia, addressing shortcomings of ordinary spell checkers that miss badly misspelled words or context errors. To create more accurate statistical language models they needed to process terabyte-scale text corpora, so they partnered with data-mining firm Atbrox to handle the heavy data and algorithm work.

Atbrox implemented a four‑phase MapReduce pipeline on AWS—uploading data to Amazon S3, running parallel tokenization, token normalization, sentence grouping, and n‑gram extraction on Amazon EMR, then writing results back to S3. The cloud solution cut costs and processing time compared with buying dedicated hardware, let Lingit experiment quickly with different models, and accelerated product improvements while avoiding large upfront infrastructure investment.

Lingit

Torbjørn Nordgård

CEO