Elastic Gets New Vector Search and NLP Capabilities
Customers should get more relevant search results when using an Elastic-powered search engine thanks to the addition of vector search and NLP capabilities in Elastic 8.0, the company announced last week.
Vector search techniques based on neural networks are one of the hottest areas in search engines. Instead of using basic keyword searches, vector search uses deep learning techniques to match the input term to a vector created from an array of features.
The vector search approach–which is also related to neural search approaches–is heralded as enabling more context to be extracted from the search term, and therefore to return better matches than what basic keyword-driven search can yield.
Elastic 8.0 enables users to bring custom or third-party natural language processing (NLP) developed in PyTorch directly into Elasticsearch. Elastic says the addition of native NLP support with vector search “enables users to perform inference within Elasticsearch, resulting in faster and more relevant search results.”
This release also brings support for approximate nearest neighbor (ANN) search, which will enable customers to query large quantities of unstructured data, such as documents, images, and audio files.
This technology was originally created for searching for image and text content, but now it’s being brought into the world of business data, and that benefits customers, Elastic says. Vector search with NLP support can “deliver faster, more relevant customer support information, improve customer shopping experiences with unique product alternatives, and enhance search accessibility by providing unique audio and visual search results,” the company says.
“Elastic is doing the heavy lifting for its customers and giving them the on-ramp they need to gain seamless value from machine learning applied to search,” Mike Leone, a senior analyst with Enterprise Strategy Group, says in a press release.
“It’s easy to get lost in the technical jargon of areas like user-behavioral ML, dense vector fields, and nearest neighbor algorithms, but at the end of the day, whether they know it or not, organizations need to apply these types of technologies to their enterprise search deployments,” Leone continues. “And many of those same organizations don’t have the time or staff to build it themselves.”
Elastic 8.0 brings several other features to the Elastic Search Platform, including new default security settings to secure data, network, and user information in self-managed clusters. The software now auto-generates tokens and certificates, which helps to streamline the setup of good security configurations, the company says.
This release also brings a more simplified Elastic Cloud on AWS onboarding experience. This includes the new Elastic Serverless Forwarder, which is designed as an AWS Lambda application and enables users “to simplify their architectures and streamline data ingestion without the overhead of provisioning virtual machines or installing data shippers,” Elastic says.
Related Items:
Are Neural Nets the Next Big Thing in Search?
Update: Elastic Shifts Licensing Terms, Citing Amazon Moves