Online cloud saving website Dropbox has replaced its platform-specific search engine with a new version called ‘Nautilus’ that is capable of full-text searches.
Search has always been a challenge for Dropbox since its such a large platform, especially given searches must be personalized for each of the service’s over 500 million registered users.
Dropbox’s search engine needs to personalize its results differently than a web search engine like Google. The Nautilus engine parses through and index the user’s files while also taking into account search preferences and behaviours.
The Nautilus engine also accounts for changes in documents, says Dropbox. Since Dropbox is a collaborative online workspace, the content of a document, for example, could change significantly over a short period.
The engine is able to crunch all of this data with the help of Dropbox’s machine intelligence platforms through a few key stages along the search pipeline. There’s content-specific machine intelligence for image recognition, as well as learning systems that rank search results based on the user’s previous preferences.
In order to be as smart as it is, the Nautilus engine is made up of two sub-systems — one for indexing and the other for serving content to the user. Indexing processes the files and user activities before extracting content and metadata to create the search index. The serving system uses the information that the indexing system has learned to present search results to the user.
Since Nautilus has taken over as the platforms dominant search engine, the company has seen improvements in the time it takes to index and update content. The team claims that it’s done working on the core of the Nautilus project, but it’s still working on features to add value or improve search quality.
The team has a breakdown of all of the technologies that were used to create the search engine on its blog for anyone interested in the minutia of how a modern search engine works.