Amazon was able to train an AI model to improve product searchers in multiple different languages. The new AI model is named as a multitask model, where functions overlap in order to improve outputs. In order to accomplish this Amazon researchers trained the system with a number of different languages simultaneously. Nikhil Rao, an Amazon applied scientist, stated that training the system with two languages at ones allows the system to fill gaps with another language.
Improving language functions is important for the e-commerce giant, a it currently has services available across 14 different countries. Additionally, Rao spoke about the potential future applications fo multitask models, "In ongoing work, we are continuing to explore the power of multitask learning to improve our customers’ shopping experiences.”
Image Credit: Shutterstock