Google showcased the development of a new AI-based model that can support up to 1,000 languages. The state-of-the-art ULM model boasts Automatic Speech Recognition (ASR) technology and can detect popular languages like English and Mandarin and lesser-known dialects like Amharic, Cebuano, Assamese, and Azerbaijani. Despite limited supervised data, the model has achieved a less than 30% word error rate on average across 73 languages. The ULM model, which supports two billion parameters, has been trained on 12 million hours of speech and 28 billion sentences of text spanning over 300 languages.
The new model is a significant development for Google, which has been working on expanding its language capabilities. Currently, YouTube uses the ULM model to generate closed captions on millions of videos in several languages, but this is limited to around 100 languages. With the new ULM model, Google hopes to support the 1,000 most-spoken languages worldwide.
The ULM model’s ability to detect and recognize lesser-known dialects is particularly noteworthy, as this has been a challenge for many language models. By training the model on a vast amount of speech and text data, Google has overcome this challenge and created a more comprehensive language model.
The ULM model’s ASR technology is also a significant advantage, allowing for more efficient and accurate speech recognition. This technology has already been used in various Google products, including Google Assistant and Google Translate, and has been instrumental in improving the accuracy and speed of these services.
Despite the ULM model’s impressive capabilities, there is still work to improve its accuracy and expand its language capabilities further. Google has acknowledged that the model is still in development and will require ongoing refinement to achieve its full potential.
The new ULM model significantly impacts various industries, including education, healthcare, and business. With the ability to support up to 1,000 languages, the model could improve communication and access to information for people worldwide. In education, the model could be used to develop language learning tools that are more accurate and comprehensive. It could improve communication between doctors and patients who speak different languages in healthcare. In business, it could improve communication and collaboration between colleagues who speak other languages.
Developing Google’s new ULM model is a significant AI and natural language processing breakthrough. With its ability to support up to 1,000 languages and detect lesser-known dialects, the model can improve communication and access to information for people worldwide. While there is still work to refine and expand the model, this development represents a significant step forward in creating more comprehensive and accurate language models.