Sirvi Autor "Luhtaru, Agnes, juhendaja" järgi

Nüüd näidatakse 1 - 3 3

Estonian Synthetic Error Generation by Prompting for Grammatical Error Correction
(Tartu Ülikool, 2024) Vainikko, Martin; Luhtaru, Agnes, juhendaja; Fišel, Mark, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Eesti keele grammatiliste vigade parandamise jaoks ei ole piisavalt andmeid, et tõhusalt treenida autokorrektorit. Hiljutised edusammud suurte keelemudelite vallas on aga avanud uusi võimalusi sünteetiliste andmete genereerimiseks. Genereerime OpenAI GPT mudelitega (GPT-3.5-Turbo, GPT-4-Turbo ja GPT-4) lausetesse grammatilisi vigu. Hindame genereeritud andmehulki manuaalselt hulki märgendades ning treenides hulkade peal transformeripõhiseid autokorrektoreid. Me järeldasime, et laused, mis pärinevad grammatiliste vigade korpusest ning kuhu GPT on vigu genereerinud, saavutavad automaathindmisel võrdväärseid tulemusi inimandmetega. Märgendamise tulemusena märkasime, et GPT genereerib probleemseid muutusi lausetesse. Töö avab palju uusi suundi edasisteks uurimusteks.
Grammatiliste vigade parandamine sageduspõhise sünteetilise andmestikuga
(Tartu Ülikool, 2022) Univer, Jakob; Luhtaru, Agnes, juhendaja; Fišel, Mark, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
In this thesis we introduce a grammatical error correction method with a neural network trained only on synthetic data. The method is useful for languages without big corpora for training a grammatical error correction model, like Estonian. From a smaller human corrected corpus, we found the probabilities of word deletion, addition, substitution and changing word order mistakes in the text. With the help of these probabilities we created a bigger synthetic corpus and we trained a neural network for grammatical error correction on the synthetic data. The author found that the probabilities of mistakes do not have to be very precise and the trained neural network can correct spelling mistakes as well as grammar mistakes.
Suhtlusvõimekuse arendamine sotsiaalsele humanoidrobotile SemuBot
(Tartu Ülikool, 2024) Unn, Albert; Kruusamäe, Karl, juhendaja; Luhtaru, Agnes, juhendaja; Kuulmets, Hele-Andra, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Social robots have been developed for decades, but creating the ability to have natural conversations with humans without strict rules has been a significant challenge. Approaches for communication between humans and robots have relied on pre-programmed dialogue options, which limits interaction and forces users to follow strictly defined rules. The rapid advancement of large language models offers a promising solution to this problem, enabling significant progress to be made in the quality of social robots. SemuBot is a student project to develop the first Estonian-speaking social humanoid robot, and this work focuses on exploring various solutions to achieve its ability to have conversations with people using a large language model. The study explores the use of three key components: speech recognition, large language models, and speech synthesis. The goal is to find optimal solutions for these components in the context of social robotics. As a result, a hardware and software solution for the social humanoid robot SemuBot is developed, enabling the robot to engage in natural conversations with people in Estonian.