
Thanaritt Kunajak
Fähigkeiten und Kenntnisse
Werdegang
Berufserfahrung von Thanaritt Kunajak
• Launched 2 new Siri voices for Thai locale to replace the old voice • Fine-tuned feature selection in a deep learning architecture’s training pipeline, resulting in enhanced model accuracy and performance through targeted evaluation and adjustments • Reviewed and debugged TTS responses to fix incorrect pronunciations, unnatural prosody, and inaccurate tonal outputs • Curated, cleaned, transformed, and analyzed text corpora using Python, regular expression, Pandas, HuggingFace, and PyThaiNLP
• Localized, translated, and created Thai dialogue scripts based on American English originals • Created automated tools/scripts to facilitate linguist works using Python and Regular expression
- 3 Jahre und 10 Monate, Juni 2018 - März 2022
Assosicate Linguist
Google
Contractor via Pactera and Accenture • Created, annotated, and reviewed linguistic data, developed context-free grammars (CFG) used for Natural Language Understanding (NLU), and performed evaluation and error analysis for existing and new Google Assistant Conversational AI features • Reviewed linguists’ work to ensure technical integrity and design compliance of the domain • Provided onboarding training sessions to 15 new joiners; recordings are being used as part of the training curriculum
• Implemented real-time anomaly detection on streaming time series data, using an unsupervised learning method called "HOT SAX" (Eamonn Keogh, 2006) to solve the issues of shortage of labelled training data and low computational resources.
- 2 Monate, Juni 2016 - Juli 2016
Research student
Japan Advanced Institute of Science and Technology
• Conducted NLP research on the topic of "Examination of Influence of Context Window Size in Word Sense Disambiguation" The task was to examine the correlation between the context window size and the accuracy of WSD. With “bag-of-words” feature vectors and the SVM algorithm as the classifier, results have shown that the optimal context window size for each word varies depending on each word’s part-of-speech category, which may result from the topicality of each POS type.
Ausbildung von Thanaritt Kunajak
- 3 Jahre und 11 Monate, Juli 2013 - Mai 2017
Computer Engineering
Chulalongkorn University
Relevant Courses: Auto Speech Recognition, Time Series Mining and Knowledge Discovery, Grammatical System, Languages in ASEAN+3
Sprachen
Englisch
Fließend
Japanisch
Grundlagen
Thai
Muttersprache
Spanisch
Gut
Catalan
Grundlagen
XING – Das Jobs-Netzwerk
Über eine Million Jobs
Entdecke mit XING genau den Job, der wirklich zu Dir passt.
Persönliche Job-Angebote
Lass Dich finden von Arbeitgebern und über 20.000 Recruiter·innen.
22 Mio. Mitglieder
Knüpf neue Kontakte und erhalte Impulse für ein besseres Job-Leben.
Kostenlos profitieren
Schon als Basis-Mitglied kannst Du Deine Job-Suche deutlich optimieren.