Elastic Case Study

As the leading provider of industry intelligence for anyone seeking to understand where the future lies, it only made sense that MarketResearch.com wanted to take advantage of the machine learning features provided in the Elastic stack.

Their unique business meant that there were unique challenges to consider.  With over 350 publishers covering every sector of the economy and emerging markets, the results that MarketResearch.com is curating is the most comprehensive collection of market reports and services that are updated daily.

This meant that downtime wasn’t an option, natural language was key, and scalability was at the forefront of the project.

Achievements and Recognitions:

"The latest release of Elastic Search enables us to combine vector, keyword, and semantic techniques for better results. At a time when customers have so much choice, it will help us stand out against the competition."

Javier Zamudio, CTO MarketResearch.com

Here’s how we made it happen for them:

Elastic Services & Knowledge Required:

  • Elastic Cloud for Hosting

  • Elastic ELSER model for creation of dense vectors

  • Upgrading from 7.x to 8.x 

  • Expertise of Elasticsearch mappings / field types to plan strategy for breaking content down into small chunks for vectorization

  • Elastic’s ELSER model and machine learning nodes to vectorize content

  • Heavy use of aggregations to provide data for various UI functions

  • Utilization of search templating to make Elasticsearch queries easy to adjust independently of integrated services when necessary.

Results:

  • Implemented machine learning-powered semantic search on Profound.com, which scales to more than 60 million documents

  • Boosted relevancy of search results: our implementation of ElasticSearch combined vector, keyword, and semantic techniques that return more relevant search results 

  • Upgraded  to Elastic stack 8.x.

  • Utilized Elastic Cloud for hosting and the Elastic ELSER model for creating dense vectors for content analysis

  • Integrated semantic and keyword-based searches using Reciprocal Rank Fusion to enhance search results

  • Launched a “beta” version of search on Profound.com allowing customers to search using natural language

Third-Party Integrations:

  • Ingest Process: Python

  • Web Integration: .net and C#

  • Azure Services and Elastic Cloud on Azure

Previous
Previous

Opala

Next
Next

SAE