Smart Compose in Gmail, a function to guess what people want to write in an e-mail and fill it out as an option, later gets an option to emulate their own language. In this way, the suggestions should become more personal.
The team behind Smart Compose calls that ‘personal language models’, appears from a blog post from Google. It is still unknown when the function will come in Smart Compose, Google only says that the team is working to integrate it. Further explanation about the new function is missing, except that it is ‘designed to emulate the writing style of an individual more accurately’ and to process it in the system.
Google announced the autocomplete function for e- mail last week during developer conference I / O. The feature works at least in English, but in the blog post the search giant specifically refers to the system having to work for all 1.4 billion Gmail users. As a result, it appears that Smart Compose will work in all languages supported by Google.
The blog post also explains that Smart Compose should show autocomplete suggestions within 100 milliseconds to ensure that users save time through the feature. In addition, training the algorithm was complex because they did not have access to user e-mails and therefore had to use other expressions of language to train the algorithm. In addition, the developers attach to the prevention of genanous situations by means of a bias in the suggestions, for example by proposing undesirable items via autocomplete. Smart Compose should be available later this month for all Gmail users.