Multitask Unified Model – a unified technology that Google is developing for answering complex search queries.
Although not live just yet, Google gave us a sneak peak into how MUM is being developed and the role it will play in delivering helpful responses through increased language understanding.
The premise behind MUM is to equip search engines with the sophistication to answer a question the way an expert would. What does this mean for the user? Simple – less searches to get things done.
Here are the highlights from Google’s announcement.
Removing Language Barriers
There’s no doubt that language can be a significant barrier when it comes to accessing information. Google understands this which is why MUM is being developed with the potential to break down these barriers through the transfer of knowledge across languages.
MUM achieves this by transferring knowledge from sources across languages and makes use of those insights to find relevant results in your preferred language.
Information Understanding Across Types
According to Google, MUM is multimodal, meaning it has the ability to simultaneously understand information from all different formats such as images, webpages and more.
“Eventually, you might be able to take a photo of your hiking boots and ask, “can I use these to hike Mt. Fuji?” MUM would understand the image and connect it with your question to let you know your boots would work just fine. It could then point you to a blog with a list of recommended gear.”
No Simple Answer? MUM’s Got You
Similar to BERT, MUM is built on a transformer architecture, but it’s far more powerful. With the ability to not just understand language but also generate it, MUM is trained across 75 different languages and many different tasks. This allows it to develop an advanced understanding of world knowledge and information.
“Take the question about hiking Mt. Fuji: MUM could understand you’re comparing two mountains, so elevation and trail information may be relevant. It could also understand that, in the context of hiking, to “prepare” could include things like fitness training as well as finding the right gear.”
Google is constantly working on enhancing the search experience by integrating many different tasks together in order to create one cohesive experience. If MUM succeeds, information will be become more accessible across language barriers and mediums, bringing the world closer together