Aki Kunikoshi is a speech scientist at ReadSpeaker and a true enthousiast of language technologies.In the interview with Aki you will find a number of project examples she comes across daily, and a great share of tips and ideas for a career in the field.

How did you become interested in language technologies?

As my father is vision impaired and my eyesight is also weak, I have been using assistive technologies like audio books or screen readers since I was small. I wanted therefore to become one to provide those support, so I decided to study speech science. Currently I am a speech scientist at ReadSpeaker - one of the highest quality text-to-speech (TTS) providers in the world, and working to evaluate the quality of synthesized speech.

I know you are tightly cooperating with the MSc. Voice Technology. In what ways?

I was one of the original members of MSc. Voice Technology development team and was going to teach Speech Recognition I and II. Although my life changed a lot during the two years when the program had been developed and I could not help but to leave the team, I am still happy to be connected to the program. Together with my colleague Dr. Kim, Jaebok, I gave a guest lecture in March and currently help some students for their thesis projects as an advisor. 

Give an example of the projects ReadSpeaker is working on, and problems you are tackling 

Example one: Let’s say, a company made an English TTS with the voice of Morgan Freeman as their custom voice. The company starts a new business in the Netherlands. How to make the TTS speak Dutch with the voice of Morgan Freeman (no Dutch corpus available!)?

Example two: We are continuously improving the TTS quality, but a customer didn’t like the listening test because it is costly or the participants are biased. How to objectively evaluate the performance of the TTS?

Example three: We have a broad list of TTS voices. However sometimes a customer wants a voice which is not in the list. How can we make the voice which the customer describes?

What kind of jobs could MSc. Voice Technology students attempt to pursue at ReadSpeaker and within the industry overall. 

Depending on your interests and skills, you can apply for:

• • a research scientist who applies/develops state of the art methods for TTS
  • a research engineer who transits research outputs to products
  • a computational linguist who develops linguistic materials to build a TTS
  • etc.

What skills and qualities are important for recent graduates in the voice technology field? 

For my speech scientist role, the following ones besides basic knowledge that you have learned in the class are especially important: 

• Curiosity of state-of-the-art speech technologies.
• Good understanding of deep learning algorithms, techniques and best practices.
• Strong programming skills and experience in Python. 
• Experience working with deep learning frameworks such as PyTorch or TensorFlow.
• Strong analytical and problem solving skills

Could you share some tips on how students can sharpen their future career focus during their studies.

• Obtain fundamental understanding rather than using tools as black box.
• Once you learn something, try it. For example, when you learned Tacotron, read the original paper, clone the repo and run it on your laptop. 
• Play with the codes. For example, you can make the codes run faster, easier to edit, or replace a component with another.  
• Think what is missing?
• Apply for an internship! If you are interested in ReadSpeaker, please contact jaebok.kim readspeaker.com and aki.kunikoshi readspeaker.com.