Brigham Tomco, CEO of Emmersion, joins SlatorPod to talk all about automating language evaluation and the underlying tech behind it.
Brigham begins with his journey into the language education and tech industry, where he started two venture-backed companies while working toward his law degree and MBA. He recounts the early days of Emmersion, where they offered language tutoring and how they evolved into building language assessment tech.
The CEO shares the importance of language learning in the age of AI with language tech like machine translation and speech-to-speech breaking barriers every day. He identifies the different key client segments, from call centers and BPOs to academia and LSPs.
Brigham unpacks the language tech and patenting process behind language testing products, such as TrueNorth Speaking Test and WebCAPE. He outlines how they handled the rapid demand for remote hiring as a result of Covid and Emmersion’s roadmap for the future.
First up, Florian and Esther discuss the language industry news of the week, with Simon Lee, CEO of language data platform Flitto, sharing how the company has integrated the metaverse into their everyday work.
The duo talk about the state of the data-for-AI market as Appen’s share price fell by 28% after releasing full-year 2021 results on February 24, 2022. Meanwhile, in Australia, captioning and translation provider Ai-Media grew 29% to AUD 29.6m in the first half of 2022.
Esther then gives an update on the Language Industry Job Index, which climbed three points to 176.6 in March 2022, up from 173.4 the month prior.
Florian: What is Emmersion? Give us the extended elevator pitch to set the scene because most of our listeners come from the LSP and localization buyer side.
Brigham: Emmersion certifies the language ability of job seekers, employees and learners. We have developed an AI adaptive assessment engine that allows us to assess people’s speaking ability as well as other skills in language in about 10 to 15 minutes and the real breakthrough is we can give immediate results. As soon as the assessment is done, you have the score reports and are able to make decisions whether it is placing a student in a course, or does this interpreter or translator have the ability to work in this language, or can this call center rep speak in English or Mandarin or whatever language? We initially started in English, but we have expanded and have Spanish, French, German, Portuguese, Italian. Then we have expanded into some of the Asian languages and released Tagalog, Japanese, Mandarin and we now have Korean and Russian under development. Again, we have been able to shorten the process and the accuracy to certify someone’s language ability for those different use cases. The key to that process is keeping it short and simple so that people can take it anytime, anywhere.
The weekly language industry podcast. On Youtube, Apple Podcasts, Spotify, Google Podcasts, and all other major platforms. Subscribe Now.
Esther: What about your professional background? You have got extensive experience in business. How did you come to decide to found a language education company of all things?
Brigham: It is an interesting process. I joke with my friends sometimes that I am a little bit of an overeducated entrepreneur. I studied accounting and then I got my law degree and got my MBA. I started two venture-backed companies during my studies and one of them was an education technology business. It was one of the first online high schools ever created and it still exists. I was always interested in education but right out of high school, I went and lived in Thailand and I was a missionary and became fluent in that language. I had studied Spanish in high school and I used that in my work and so early on language was something important. As I left some of those early entrepreneurial opportunities, I managed a large investment fund, about a billion dollars and we have investments throughout the world. As I looked for the next stage of my career, I knew I wanted to do something international. I loved languages. I loved culture and people. I founded a business in the Philippines where we did offshore staff leasing. We built teams for companies all throughout the world, primarily focused on digital marketing and then in later years in software development.
A lot of the initial pains that I felt in that business and evaluating English as a second language is what led to Emmersion. We had a particular experience early on. I only had two recruiters and I had a client come to me and say, I need 500 English-speaking call reps and I need them placed within three months so we had to screen about 5,000 applicants with two recruiters who did not speak English as a first language. I remember sitting, this was about a decade ago, and thinking there has got to be a better way to assess the language ability of job applicants because there were not any automated and price-conscious options at the time and that was the seed of my side of the business. My Co-founder Jacob Burdis, the AI technology that we have in our patents was his dissertation research. He had been running a language learning school that trained about 30,000 people a year in 60 different languages and he was having trouble assessing their speaking ability, so we commiserated and realized I have this problem in business, he had that problem in education and we came together and developed Emmersion to solve that problem.
Esther: What do you think are some of the key milestones and challenges that you have faced along the way since founding the company? Where are you at in terms of the leadership team and the size of the business now?
Brigham: We have about 750 customers now, so it has grown dramatically. About 600 universities and about 150 large multinational corporations. We did a lot of language tutoring at the beginning and a lot of our customers continued to ask for an assessment and so that is when we started building our assessment technology. We first created an assessment that worked but then we needed an assessment platform for admins to be able to integrate with platforms and so that is where things have evolved. From a single test to what we call our assessment engine that allows us to create other languages and deliver data and let our customers manage that data because some of our customers have over 3 million applicants a year and so you have to be able to organize that data, so it is both the front end and back end. Our team is mostly on the engineering side of edtech businesses. A lot of people are passionate about what we do because we feel we are helping people gain opportunities by proving their ability to use the language, whether it is to get more education or get a job and, as we have grown, we have recruited out of other SaaS technology startups. Primarily education and HR tech-type companies are the people that have been attracted to what we do.
Florian: How hard is it to add a language and how do you make the selection of languages?
SlatorCon Remote June 2023 | Early Bird Now $120
A rich online conference which brings together our research and network of industry leaders.
Brigham: Supply and demand, so English was first as it is a broad language for business and for the ability to access education. A lot of our early traction as a business was Latin America so Spanish and Portuguese were added pretty quickly and then French is another global business. What drives the growth of those other languages is usually customer demand, so if we have large customers and multiple of them need a certain language, then we will proceed forward with those. Tagalog is probably the most random one out of the list. We had two large customers, one a large education institution, and another one a large BPO that had a big need for Tagalog, and three of our management team members speak Tagalog as well so we also had a unique skill set and background to be able to do that. As we evolved as a business, we developed the first language, but as we developed our platform, we specifically did it in mind knowing that we would want to develop other languages. The general way that the algorithms and scoring work is the same across languages and we have a dialed-in process of what is the minimum number of items or questions we need to start the test. How many people do we need to calibrate it? How do we remove bias by getting people of different ages or gender? What is their base language? It is a lot of work if you did it from scratch. We built the platform in a way that as much as we can, we have a process. One of the hardest things is going from the traditional romance languages to the Asian tonal and character languages. We did Japanese first because we had a lot of skill sets within the company at once. We were able to figure out Japanese, Mandarin and Korean were much easier after that because we figured out some of the nuances of using those languages.
Florian: Do you ever get the question on the language learning side that people are saying, with all this fancy AI technology, machine translation, speech to speech, why is there a need to learn a language in the first place?
Brigham: We definitely do because similar to the AI technology that is coming out in the translation, interpretation space in call centers and customer support, chatbots have become much more prevalent. At some point are chatbots just going to replace humans? The answer is it probably will replace it in some low-level capabilities similar to the way that machine translation will replace a lot of the basics. Then you get to an area where the machine translation is not accurate and it has to be absolutely accurate in areas of legal and healthcare and these other areas. There is often still an overlay of human review and involvement. Particularly in what we do, interacting within the office in that language is important, not just that we have these chatbots to do it, so there is always going to be that. Human empathy needs to be involved with jobs that require language ability. Some of the low-level things will be replaced, but I view AI and technology as a way to enhance the human component of what we do so that highly trained, highly paid people could focus on more important but high-level activities. I view it as AI is human augmentation as opposed to human replacement with a lot of these overall processes, at least.
Esther: How does what you do interact with some more of the traditional or standardized certifications that might exist? How does that tie in with the solution that you are providing at Emmersion?
Brigham: We would call those assessments high stakes, TOEFL, IELTS, TOEIC. You have a handful of those. On our score report, we give you a crosswalk to those tests. We are not trying to replace those, but if someone is preparing for an IELTS, TOEFL, etc., you can take our assessment and get an idea of where you are at and it is much cheaper and much quicker to do that. I will not say that competing with those is not in our future, but it is definitely not in our one to two-year roadmap. With 600 universities using us in their language courses for placement of progress, we are getting tons of data. What is going to set us apart in AI is data and the trust that these organizations have. I imagine as language departments trust us, admissions departments will trust us as well down the road as we choose to focus there. Our goal on that side is to help people know how they are progressing towards their IELTS and TOEIC. We do have a few universities that use our assessment as the entrance exam, but that was by their choice, not by us intentionally marketing in that direction. You have something that is accurate, it is much more inexpensive for students and much faster for students with immediate results. It is naturally going to enter that market where you have seen companies like Duolingo start to take TOEFL on their English entrance exam.
LocJobs is the new language industry talent hub, where candidates connect to new opportunities and employers find the most qualified professionals in the translation and localization industry. Browse new jobs now.
Esther: Who else makes up the key client segments for Emmersion. Who are you targeting?
Brigham: Call center and BPO would be a subset of what we call pre-employment screening so anywhere where large organizations need to assess language ability. We have a lot of language learning companies that either resell our assessments to their students or use them for placement of progress. There is academia and then we also work with language service providers in a couple of different ways. They use us for recruiting and hiring their translators and particularly interpreters where you need to know their ability in their L1 and L2 or both languages that they are working in. They can assess them on both sides and then companies use us for promotion, so they will regularly assess their employees because certain sales jobs or outward-facing jobs will require better language or communication skills. Then finally you have a variety of our partners. Both language companies and LSPs resell us to their customers because as an LSP most of their customers view them as the experts on all things language, so if they have a problem with anything language, they say, Hey, how do we solve this? We are starting to have more and more LSPs add us to their toolbox so when customers ask them about assessing language ability, they have a tool they can offer either directly through us or through a white label type arrangement.
Florian: Did you think that you were going to work with LSPs when you started this? Or was there one company that came, they had a need and, there are all these others out there now?
Brigham: Definitely the latter. We initially started at the very beginning working with what we call intensive English programs, so schools around the world who were teaching and learning English. We had a few LSPs whose customers were asking them how do we evaluate the language ability of our call center employees or our hospital staff, so they came to us for that solution. It started with a couple of companies in the industry, like Akorbi and CyraCom, and then just grew from there because once the need was known, we went and shared that need and it was a universal pain for a lot of these LSPs.
Esther: You have got The TrueNorth Speaking Test and WebCAPE. Can you tell us as much as you can about the underlying technology and also a couple of patented solutions? How does that work?
Brigham: TrueNorth overall is our speaking assessment scale and our methodology for assessing speaking. WebCAPE stands for computer adaptive placement exam so it is a particular product for academia and for placement into those courses. Our real breakthrough has been in automating speaking assessment scoring and doing it accurately. A lot of what was out there was essentially taking speech and running it through a speech recognition engine like a Watson or a Google and just getting pronunciation feedback because that is what the data those systems give you. Pronunciation is not fluency and is not an ability score so we use standard front-end speech recognition engines. In many ways, it is a commodity for us on the very front end. We get our initial data either from an IBM Watson or a Google Voice, depending on which platform has a better model for the language that we are working in and then our algorithms come after that. Without going into what the actual algorithms are, each question or item is calibrated based on language ability as well as grammar and other principles. From each question we ask, we get a lot of data, a lot more than just pronunciation. We know that that question or item is at a certain level on the scale so that if you perform to a certain level, it is inferred that you can perform there. It also tests different grammar principles in that process as well and the overall process we use for most of our speaking assessments, called an elicited imitation. It is basically listening and repeating but we are able to get the data through our AI models using that system.
Pronunciation is a component. It is not the only one, so if you think about work in the workplace, intelligibility is what you are going for. If my Mandarin accent is different from your Mandarin accent but the people that speak Mandarin can still understand us, then we should all be able to get that job and work that job. One of the important things in language bias with AI is that accents in certain languages or certain L1s, your primary language, have heavier accents going to different languages. We have chosen to leave the aperture really wide on pronunciation so we focus on intelligibility. If the model can understand you, then you should be able to do the job or you can go to that school and then perform in those courses.
Florian: For pricing and the business model, it seems now it is very much enterprise sales in bulk? Is that true and are there any plans to go B2C? Or selling those tests on a test by test basis by the internet?
Brigham: It is an evolution that we are a part of. As we started the company, we sold tests in bulk batches. If you buy 100 assessments it is going to be a little bit cheaper than 10 per test and that is how we started. As we have evolved, we have tried to move a little bit more to a SaaS model where it is a subscription either monthly or annual based on volume. We have focused on B2B. For us, it was important to have credibility. With academia, we get a lot of credibility and also with the large organizations we are with but naturally, we have started having individual learners and people interested in our tests. Our first foray into a consumer model is what I call B2B to C, meaning we have language companies who have consumers who use them and now resell directly to the consumer. You could technically find a way to buy a single test on our website. It is just not the promoted primary path but for me, I look at our long-term play. If companies are using us and universities are using us, individuals or consumers are going to want to be using our tests and practicing because that is what is going to help them get those jobs and get those in the future so I would imagine in the next couple of years we will have a division that focuses on B2B as well as B2C.
Esther: What has the impact from COVID been, if any? We have spoken about remote hiring trends, general remote trends in the industry, but has there been any accelerated demand as a result of COVID?
Brigham: The whole industry with translation, interpretation has been riding this wave of globalization. Even small companies are now very global. There were solutions for large companies but all these small businesses had to figure this out as well and then the digital transformation from COVID only increased the globalization trend that was already happening, so you had the big wave. It broke through barriers. I ask a lot of people, would you ever have hired this position remotely two years ago? We would have never thought of doing that. Would you hire an executive that you would never meet face to face? No way and now we all do it every day. Half my company has never worked from our home office now or even been to the home office. What COVID did is break through a lot of those barriers and so we are willing to hire people in positions we were never willing to hire before and when you make a job posting for a software engineer, it is not in the UK or Switzerland, or even in London or Geneva. I do not care where you are in the world, if you can speak the language I need you to do, and you can perform the job then you can have it. We feel that our ability to get in there and certify the language ability of all of these opens up the world so then they can focus on the other skills that are required for the job.
Florian: Tell us what is on your priority list over the next 2 – 5 years? What are the top initiatives?
Brigham: Over that timeframe, we will continue to expand languages to get more and more coverage. With the languages, we have covered 95 or 98% of the business need, but we will continue to expand there. Integration is key, so data is great. It is important to have it at your fingertips in the systems that you are already using. The final one is in our assessments, we have the ability to deliver subscores based on the different capabilities within language that we can deliver just from a single speaking assessment. Finding ways to deliver more data in a shorter period of time for these language assessments will be key to us as we continue to evolve.