Why University of Edinburgh


I loved my time as a software engineer at Emmersion/Rosetta Stone. I was super happy about working there because I was super interested in the core technology of the product that we were developing. I loved it when we were creating or maintaining the actual tests. I was very interested and pushing to be a part of the assessments team that would have responsibility over the quality of the tests. I liked being able to answer product questions with data, and I found myself interested in things that maybe fall outside of the role of an engineer in a larger company. I had lots of conversation with my good friend at work, who was the head of assessment, about how we could improve out data pipelines, as well as improve our tests. We were a small lean team, so there wasn't as much opportunity to work on test development as I would have wanted. However, I was able to do some experimentation with different ASR providers and also brainstorming about potential ML solutions for our open response. I was fascinated with the speech technology and the power of the speech models, that we were looking at. I wanted so bad to find a better solution to our open-response grading.


This interest actually gave me the idea that maybe I was pursuing the wrong career. I had a faint idea that maybe more school was the right answer. It wasn't a strong feeling, because I loved my job as an software engineer, especially with such a great team.

I started to feel like the development of language and speech products could be a really interesting way for me to focus and advance my career in a positive and meaningful direction. I started exploring options that exist that would advance my knowledge in ML, AI, Speech, and NLP. One of the fundamental reasons that I was working at Emmersion was because of my interest in the intersection of technology and natural language. So, naturally I was looking into programs and courses that continued to align with those same interests and passions.

At first, I looked up top NLP Master's programs. Unsurprisingly, I say a lot of Stanford, Oxford, and other Ivy leagues near the top of these random online lists. As I started to compile several lists together, a few schools stood out. And one that stood out particularly strong in my mind was the University of Edinburgh. Ever since coming back from my time in the Czech Republic, the thought of studying at a university in Europe would be fun. It seemed exciting and challenging. Another thing that drew my attention to Edinburgh is the fact that Victoria loves the UK. She has always had this dream of living in the UK for a long time. So, with all these factors in mind, I decided to send in an application for a MSc of Speech and Language Processing at the University of Edinburgh. The application was due soon and I didn't have anything to lose, only options to be gained. The application deadline was at the end of April for the upcoming Fall semester, so it was last minute, but I figured "its a top program, so my chances are slim anyway."

The timeline of everything was basically fate telling us what to do, because near the end of May, I got a phone call from my awesome manager to tell me that IXL Learning (the company that acquired Emmersion) was reorg-ing the company. We were all being let go to be absorbed by there Rosetta Stone team in North Carolina. It was actually really bad news, that I took fairly hard, because I really enjoyed my job. I was hoping to be there for a long time, and I was even talking with the Rosetta Stone ASR team about potentially switching orgs myself. No one ever expects a layoff and they suck. This bad news came the very same week that I got an acceptance letter from University of Edinburgh. It kinda felt like a sign that going back to university was a good thing for us to do.

Since being at university, I have learned that Edinburgh was the first university to set up an AI research center. They opened shop in 1963, which is a significant reason that they have such strong and notable AI programs. My program, Speech and Language Processing, is an interesting program because we are technically in the Philosophy, Psychology, and Language Science department because it is speech focused. However, all my classes that I have taken and will likely take are in the Informatics (CS) department.

This program is perfect for me and my interests, It lies right in the intersection of natural language and technology. There are several things that I want from this program:

  1. A deeper knowledge and understanding of the fundamentals of AI and speech technology systems.
  2. A strong network of incredible and smart people, both students and professors, with similar interests and passions.
  3. An unforgettable adventure living as an expat in the great country of Scotland.


Comments

Post a Comment

Popular posts from this blog

Phonetic Posteriorgrams

Image
For my MSc dissertation, I am trying to synthesize speech with variable L2 accent strength. There are a few ways to approach this. The "fastest" and "easiest" way is to have the perfect dataset that has reliable labels for accent strength on each utterance. There are two problems with this approach. First, this dataset would be incredibly difficult and expensive to create. Second, it is difficult to get a trusted label for "accent strength" that is also consistent across the dataset. What does accent strength even mean and how would it be measured non-subjectively? I was talking about this problem with one of the PhD candidates in the CSTR group, and she told me to look into phonetic posteriorgrams. Since then, I have tried to read a bit about them. I wanted to explain them in this post.  What is a Phonetic Posteriorgram? A phonetic posteriorgram (PPG) is well described in it's name. Phonetic indicates that we are talking about phonemes, or speech cate...
Read more

The Hard Part About Speech

Image
The rise of Large Language Models (LLM) in recent years has certainly made a lot of people see the world differently. I believe that many people are more interested in statistics, machine learning, computer science, and linguistics because of these fantastic advancements, myself included. As I have studied intensely at the University of Edinburgh over the last year, I have realized more about what these kinds of models do, and how they relate to the bigger picture of modeling functions in general. All of these models are doing some form of function approximation/prediction task. Language models like GPT or BERT are simply predicting the next word token given the context, Automatic Speech Recognizers (ASR) like Whisper are predicting the text output given a waveform, Text to Speech (TTS) models like Amazon Polly are trying to predict the spectrogram or waveform given the textual input and who it should sound like. There is always a task, a dependency, data, and a metric to measure that ...
Read more

Robotic Voices and Speech Synthesis: Experiments with Unit Selection

Image
This is the paper that I wrote for my Speech Synthesis class in semester 2 at the University of Edinburgh. Building a unit selection voice was a lot of fun, especially getting to record our own voices. Learning unit selection help me to understand a lot of the basics and fundamentals for the state of the art technologies because they are all trying to solve similar problems with different approaches. Here are some sample audios from my synthesized voice. "She sells seashells by the seashore." "I scream. You scream. We all scream for ice cream." These are very robotic, but this was state of the art not that long ago and cool that it is built with a database of my voice recordings. Abstract Unit selection in speech synthesis, an aging technique dating back to 1996, remains a fundamental method for synthesizing speech. This paper explores the theoretical principles of unit selection, its practical implementation, and its implications for modern speech syn...
Read more