Language is probably one of the most natural skills for human beings. Companies have spent years making their language learning process as easy and simple for people as possible. In recent years, innovative technologies have revolutionized numerous areas of our lives and language acquisition is no exception. However, it turns out that it is the not-only-ML (Machine Learning) approach that can really change the way we learn a language or acquire speaking skills.
How we learn languages today
Humans are unique in having complex language capabilities. We are able to communicate emotion and attitude "musically" using tone and inclination. Cultural and community attitudes are communicated via language. In addition, vocalization is combined with body and facial movement to communicate information and emotion.
Or, even, why do we actually learn languages and get to the edge - for what do we have them? One would say, we have languages like any other living beings do. Birds, cats, even viruses and bacteria have sort of a protocol of communication. For us, humans, the story goes much deeper. Not only we're "protocoling" to each other using words. We're actually communicating emotions and setting up the whole community flavor using tone and inclination of a particular phrase. This could be one or two words you thought over the public channels, or something personal you decided to communicate to your neighbor. Shouting to your dog? No worries, they understand our emotional context better than most primates including... eh, humans.
Ok, convinced in the common truth - languages are important. Thank you. How to learn them?
Well, you don't need to 'learn' speaking skills. They are, in fact, pre-built into most parts of the brain. Instead, we need to figure out an approach on how to activate these mechanisms. That's where, as you would smartly expect, the language learning services, some dating as old as humanity known, exist. In the epoch of the Internet and machine-learning, the amount of companies who offer the Best Language Learning experiences today is probably higher than the amount of people learning these skills. Most of them are not even an app, could be a video channel or a voice-based private chat with a teacher in your best messenger.
How to break into this market? Who gets closer to what's needed instead of expected?
Voccent spoken language acquisition is key
Our approach is similar to learning music and engages those parts of the brain that handle tone and timbre ( the unique tonal structure of an instrument). This allows learners to acquire accent and phrase structure as tonal units.
...and here we go, Voccent is the company that tackles the field from the inside of your skull. Well, they are not going to drill and put some wires... yet, but it's interesting. It turns out, that the biggest problem right now is as simple as that: repetition and neuro-feedback contribute to the maximum level of success one could achieve while learning new auditory experiences, and in particular spoken languages, by relying on the oldest mechanism in our brain: Frequency-following responses (FFRs). Voccent deliberately doesn't distinguish between a myriad of languages and dialects combinations on the software level, it's so blind in a good way of saying that to the audio inputs, that you can try repeating after a musical instrument, like piano or guitar, or what any other you just thought of. Based on that Voccent's CEO Victor Brunko states that internally within the team and with the early adopters of the product the agreement is to minimize usage of the words: 'learning, teaching, linear, lesson' in favor of the words: 'acquiring, mentoring, non-linear, challenge'.
Technologies behind Voccent
So, as of today, the startup is in a growing phase. Which means that all the approaches and technologies are being mapped to meet the overall paradigm. Not without challenges of course. For example, you still need to put an effort. Listen to the samples. To cycle over Repeat-Compare-Improve-Repeat. Mitigate privacy requirements.
To make all this happen, a team of 8 people including two co-founders are busy building and supporting a variety of services. Crucial services like speech-processing and analysis are hundred percent homegrown. The frequent question for the young companies that people often ask is - why not to rely on something other big leaders in the speech industry provide as a service? Well, the answer is not short to that. Though the concise idea is that even the combination of different APIs you would find today is not meeting the requirements to achieve this main idea: Voccent is a dialect-agnostic language-acquisition service that allows people to practice their speaking/listening skills based on immediate feedback. And the feedback may contain a number of variants: pronunciation, prosody, energy, pitch, pace, etc. One more question is obvious - privacy and security. At Voccent all the audio files and non-public references are being encrypted. Unless a person would like to share anything, all the information is secured and hidden behind the walls. Depending on the use-case, for example, such as schools, government or regular customers - privacy and security requirements are all different. The rule of thumb for the Voccent engineering team is quite simple - even engineers having highest access level to the infrastructure shall not be able to listen to private audios. Sounds good as a minimum requirement.
How many people do you need to write 2 million lines of code? Well, it depends on the focus and motivation. At Voccent Engineering it took 8 people. With the leaders who prefer to warn how not to do instead of simply telling what to do, everyone in the team matters. Team doesn't practice standard approaches most of the companies do these days. In other words it looks more like a scientific club rather than one-more-startup-in-the-field. It has cons and pros, but on the good side is the feeling people share - a prosperous future with the product that does something others don't and something which can improve communication between humans.
Currently, the product is in the pre-production testing phase with a limited number of users. Market penetration strategy team decided to go with simplifying a user experience for the audio and video streaming based lessons created by language teachers. It's achieved by adding the link to the language learning channel at the description on any other platform and get the customers practicing the target language on the Voccent platform. First results and feedback show it's more efficient for both: content creators (*tubers, *grammers, *tokers) and people who acquire new language skills. Visit https://voccent.blog/ to find more about Voccent.