Vés enrere

Thesis Defence: Alp Öktem

Thesis Defence: Alp Öktem

29.03.2019

 

Alp Öktem

Incorporating Prosody into Neural Speech Processing Pipelines. Applications on automatic speech transcription and spoken language machine translation.

Supervisors: Dr. Mireia Farrús and Dr. Antonio Bonafonte (UPC)

25/02/2019

 

Thesis Brief Description:

In the current era where humans build more and more interaction with machines, prosody still is a less-focused phenomena in systems with natural language understanding. Automatic speech recognition systems give focus on "what is said?" and disregard "how it is said?", which is many times vital element of human communication. In my thesis, I focused on the recovery of the prosodic information which deals with the "how" an utterance is delivered in speech processing pipelines. I developed my methodologies on two principal use-cases, automatic speech transcription and spoken language machine translation, and used state-of-the-art neural methods as baselines. My experiments showed the enhancing effect of representing spoken language together with its prosodic features. 

Experience as a PhD Student:

My PhD journey has been full of rewards and learnings. I consider myself very lucky to have had the opportunity to discipline myself for the search of something that is hardly definable at the beginning. It is a journey that requires a strong motivation first, and then the right people and resources to guide you. I am thankful for the many talented people I met, be it researchers, developers or artists in UPF and nearby universities, in Barcelona and the conferences that I visited. I am glad to have ended this journey with great motivation to keep searching and making things that will benefit others. 

Multimèdia

Multimedia

Multimedia

Categories: