What that means in real-world terms, basically, is searching for files and folders in Drive became more simple. Looking for an esoterically named expense report tucked deep within a sub-subfolder? With Drive’s new natural language support, there is no need to remember what you called it. Simply type “find my budget spreadsheet from last week” or “show me recent Sheets documents” and Drive will do as directed, surfacing all document types you indicated saved within the time frame you specified. And it gets better with each query, Google said.
That is not the only manifestation of Drive’s new NLP chops. When you are typing the name of a document in search, it will autocorrect typos and misspelled terms.
Drive also improved in other small but appreciable ways. Documents can be split into multiple columns by choosing the new Columns option from the Format menu. Drive now features automatic backup. When you open, convert, or edit a non-native object in Drive — like a Microsoft Word document or Pages file — a copy of that document will save automatically –they will live within Revision History in Docs, Sheets, and Slides on the web for posterity.
Google said natural language processing will begin rolling out globally, starting Tuesday. They are a web-only affair for now — the company did not specify when NLP would trickle down to the Drive clients for iOS and Android — but if history is any indication, things will not remain that way for long.
Consumer-facing products are not the only form Google’s work in natural language processing has taken. This past summer, the search giant unveiled Cloud Natural Language API, a connected service that provides NLP smarts to third-party apps and tools. It can perform basic functions like entity recognition — like classifying words as names, locations, or expressions — and sentiment analysis — which includes categorizing opinions as negative, positive, or neutral. But the real innovation is its syntax analysis: the API has the ability not only to infer the meaning of words from contextual clues but to identify those words’ linguistic relationships.
Google is not the only one investing manpower and development resources in natural language processing, of course. In June, Facebook took the wraps off DeepText, a learning-based text engine that can interpret with “near-human accuracy” the content of “several thousands posts per second, spanning more than 20 languages.” It is in the early stages, but already, Facebook said DeepText can discern intent, sentiment, and entities from nothing more than a combination of text and images.
