BERT
A transformer language model Google applied to search from 2019, improving understanding of word order, prepositions and context within…
AlgorithmsMultitask Unified Model
MUM
Google's multimodal, multilingual model announced in 2021, able to transfer knowledge across languages and formats to answer complex comparative questions.
In full
Google described MUM as far more capable than BERT and trained across 75 languages and multiple modalities, letting it apply information found in one language to a query in another. Its earliest visible applications were in COVID-19 vaccine information, Lens multimodal search and 'Things to know' features rather than general ranking. It is a precursor to the reasoning behind AI Overviews and AI Mode.
Example
A question phrased in English about a hiking route draws on Japanese-language sources that have no English equivalent.
Related terms
AI Overviews
Google's AI-generated summaries shown at the top of many search results, synthesising information from multiple web pages with links to…
Semantic search
Retrieval based on the meaning of a query and documents rather than literal keyword matching, using entities, context and learned…
Google ranking systems
Google's documented list of the systems that generate rankings, distinguished from one-off 'updates' to those systems, and including…
Multilingual SEO
Optimising a site that serves several languages, ensuring each version is fully indexable, correctly annotated and not mixed within a…
Where to read more
- Attention Is All You Need · Landmark Post · arxiv.org
- Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer (T5) · Landmark Post · arxiv.org
- MUM: A New AI Milestone for Understanding Information · Landmark Post · blog.google