MUM is a machine learning model that detects nuances of speech in searches to better filter the information provided to the user. In this case, Google aims to identify a wider range of moments of crisis.

Google wants to do more for the users who turn to its search engine in times of trouble. To that end, the company is integrating its latest machine learning model, MUM, into the platform to detect a wider range of personal crises. The idea is to provide contextual information, related to the original search, that can help the user.

MUM will serve to unravel, in a detailed way, the various nuances of a given search. According to Anne Merritt, product manager in Google's health department, the system is "able to help us understand longer and more complex sentences, such as 'why did he hit me when I told him I didn't love him'".

In conversation with The VergeMerritt explained that "it may be obvious to a human that this research would be about domestic violence, but without the right AI, natural language is difficult for our computers to understand.

Other examples of how MUM can provide more appropriate answers include searches such as "most common ways to commit suicide," an instance where, previously, the engine would understand it as a simple search for information; or "places where most suicides are committed in Sydney," a search that would previously bring up links about tourist areas in the Australian city.

Currently, when Google detects a moment of crisis, the technology responds with a contextual information window where help links and contacts are provided.