Appen recently sponsored the IEEE International Conference on Acoustics, Speech, and Signal Processing (iCASSP) in Brighton. Our VP of Business Development in Europe, Dorota Iskra, shares some insights from her experience there.
With an extensive background in speech and language technology, Dorota holds a PhD in Computer Science and Speech Recognition from the University of Birmingham. After focusing on the research side of this field for many years, she redirected her efforts to applications and language resources. Dorota has worked in the telecom and software development industries and, as Appen’s VP of Business Development in Europe, helps Appen’s customers improve their speech, natural language, and AI products with high-quality training data.
Hi Dorota, can you share a little bit about your background, the work you do at Appen, and how you became involved in the industry?
I discovered speech technology when I was working on my university degree. It seemed like a perfect combination of linguistics and math, two disciplines that I always found hard to choose between. I was so excited by the topic that I wanted to understand it better and decided to work toward a PhD in that field.
After that, I took various roles that gradually moved my focus from research to applied research, then to applications, and finally to commercial projects. Now I’m in a role that is completely commercial. For the last seven years, I’ve been looking after Appen’s European customer base, serving our existing clients and growing our presence in the market. . So, over 20 years I’ve worked for several organizations in a variety of roles with one common denominator: speech and language technology, and more broadly, artificial intelligence (AI). I’m still fascinated by this technology – maybe even more so now, thanks to a number of technological breakthroughs and seeing its applications in daily life.
How many years have you attended iCASSP? What trends or shifts in the market have you noticed over the years?
Strange as it sounds, this was actually the first time that I have attended iCASSP, but I have attended its sister conference, Interspeech, ever since my PhD days. The most interesting aspect of iCASSP to me is that it is about signal processing in general, not just speech, so it was fascinating to learn about applications related to other types of “signal” such as music, sounds we make when asleep, breathing, and other sounds made inside our lungs.
With the expansion of neural networks, we see the range of application areas for signal processing growing rapidly. Also, the availability of development toolkits makes it possible to use this technology without having to be an expert. The challenge now is how to improve it, and one way of doing that is by using more data or better-quality data, which is what we at Appen like to hear!
What are the most exciting new applications of speech & signal processing you’ve seen lately?
Although not new, I’m still very excited by virtual assistants (VA). Their broad range of functionalities is continuously expanding and their performance is constantly improving. I had put off buying one since they aren’t available in my mother tongue, but my kids’ English is good enough by now, so we have a new addition to the family (I won’t mention the name though!) It feels like having a new pet that everybody needs to meet! Our favourite use case is requesting music and challenging our assistant with foreign titles and artist names. And we do have a few languages we can choose from!
Do you have any top takeaways from the conference?
Spending much of the conference talking to customers and prospects, many of the highlights are one-on-one conversations with people who visit Appen’s booth, as well as the sessions I attend. This year I heard two very interesting ones. One was by Sir David John Spiegelhalter from the University of Cambridge on “Handling Uncertainty.” This world-famous statistician explained how surveys and other statistical studies can be subject to interpretation. He illustrated in a very entertaining way how one of his own studies went to live a life of its own all over the British press, based on selective quoting and free extrapolation of numbers forward and back in time. I had never heard an academic audience laugh that loud in a presentation on statistics!
Another interesting keynote was by Corinna Cortes from Google on “Federated Learning and Structured Prediction.” In the times of GDPR and widespread concern around data privacy, federated learning allows the improvement of machine learning models while keeping data on device — which seems a very effective solution addressing these concerns.
Where can we find you next?
The next conference I’m planning to attend is the AI Summit in London (12-13 June). Appen is one of the sponsors, so we’ll have a booth there. I’m looking forward to learning about a whole new range of use cases for AI in various vertical markets. After that is the Chatbot Summit in Tel Aviv (25-26 June), where I will also be discussing how Appen we helps improve chatbots and virtual assistants with high-quality data.
At Appen, we’ve helped leaders in machine learning and AI scale their programs from proof of concept to production. Contact us to learn more.