Katia Starenka

Katia Starenka
Grammarly
United States

ExLing 2024 Paris exhibition
AI-Powered Features at Grammarly

At Grammarly, our mission is to improve lives by improving communication. One of the key technologies driving our innovations are Large Language Models (LLMs). These advanced AI (Artificial Intelligence) systems are designed to understand, generate, and refine human language by processing extensive corpora and employing sophisticated algorithms. Our objective is to make communication assistance more intuitive, accurate, and contextually intelligent.

In this presentation at the ExLing 2024 exhibition, I will showcase how Grammarly leverages LLMs to refine integrated feature sets, focusing on both revision and composition suggestions. LLMs help us not only detect and correct grammatical errors but also provide users with feedback on the quality of their writing. LLMs also power our suggestions for improving clarity, conciseness, and engagement, ensuring that the text aligns with the intended audience and communicative goals. Finally, they enable Grammarly to summarize long passages to provide concise overviews, making them invaluable for tasks such as drafting emails or writing reports.

Computational linguists play a vital role in developing and refining Grammarly’s LLM-powered solutions. They design the interactions between user input and the AI, handle text processing, analyze the quality of AI-generated suggestions, and identify and correct inaccuracies to ensure a high-quality user experience.

Bio
Katia Starenka is a computational linguist at Grammarly with a strong background in applied linguistics. Katia has worked extensively across various areas of applied linguistics, specialising in developing linguistic solutions that address grammatical variability, lexical alternations, and paragraph structuring to enhance the text’s clarity, tone, and fluency. Her work also focuses on developing sophisticated and audience-aware solutions to make communication assistance more accurate and contextually intelligent.