Researchers at New York University have developed a machine learning model that mimics how children learn language. Using video and audio recordings from a young child’s perspective, the model successfully learned to match words with corresponding pictures.
The strong performance, twice that of much larger models, suggests we are close to understanding how children come to understand and use language. For now, we are moving into the realm of unconfirmed, but only probable theories.
– There are many different theories about how children acquire language. On the one hand, we have the theory of natural language acquisition - many researchers argue that children are born with special elements or knowledge built into the brain that uniquely allow us to learn language. On the other hand, there is a theory based on education that states that children learn language primarily through sensory experience. We are not talking about innate abilities, but about the experience that we gain every day, which allows us to learn, for example, the English language. explains Wai Kin Wong of New York University’s Data Science Center.
The scientists based their work primarily on the second of these theories. In developing the model, they also wanted to gain a deeper understanding of the process of early language acquisition.
The database for this work was collected using a head-mounted camera worn by one child. She recorded video and audio recordings while accompanying children ranging in age from 6 to 25 months. The data set included 600,000 people. video frames combined and 37.5 thousand transcribed statements. It is fundamentally important that the research takes place not in laboratory conditions, but as close as possible to the natural environment.
– This is the largest model of its kind ever created. There are already datasets containing recordings of people talking to their children, but none have covered such a long period of time or included video footage of what the child sees at any given moment. Our data provide a unique perspective on language acquisition. Additionally, we combined this data with a multimodal neural network. The model is quite general in terms of how learning occurs. Data is transferred and a simple update of the rules on the basis of which training occurs. The fact that learning occurs opens up a new perspective on early language acquisition that differs from previous approaches. Much more emphasis is placed on learning and what can be achieved from it. – emphasizes Wai Kin Wong.
The performance of the model was assessed based on categorization by assigning words to corresponding visualizations. The classification accuracy was 61.6%. For comparison, tests were carried out on the CLIP image-text contrast neural network, trained on a much larger data pool. At the same time, the classification accuracy was only 34.7%.
– For the first time, we confirmed that our model can learn these words from children’s real-life experiences. This learning opportunity supports the idea that this may be one way for children to learn such concepts. We cannot say this categorically, but this is a very promising and interesting direction for further research. – announces the researcher.
According to Precedence Research, the global machine learning market will exceed $38 billion in 2022. income. By 2032, its value will rise to more than US$771 billion.
Source: Wprost
I am George Brown, author at Daily News Hack. I mostly cover economy news and I have been doing this for quite some time now. I have a lot of experience in this field and I’m always looking for new opportunities to learn more.
