Today the Lucidworks team is very happy to announce the release of Lucidworks Fusion 4.2, the latest version of our platform for building, deploying, and optimizing AI-powered search and data applications. Fusion 4.2 will alleviate the burden that merchandisers feel maintaining complex business rules that require constant adjustment as their business changes. Fusion 4.2 gives these merchandisers advanced machine learning to stay ahead of trends, automating much of their tedious maintenance work, improving conversion rates, and freeing their teams to focus on the retail innovation that machines can’t do — and where human expertise matters most.
Lucidworks Fusion 4.2 sends a strong signal to these digital commerce teams: ML gives you more control over your business. Product owners and domain experts can easily deploy a more powerful digital shopping experience to delight customers and grow revenues. We’ve also added a new DevOps center for easier maintenance and management of your Fusion deployments. We’re also excited to introduce new NLP and deep learning capabilities in this release.
Here’s what’s new:
Fusion 4.2 brings together several different query rewrite methods into one interface. Business leaders using Fusion 4.2 for digital commerce can provide the best possible shopping experience with improvements in traditional methods like business rules in combination with newer ML-driven technologies like synonym and phrase detection. With this set of tools, merchandisers can configure, preview, and commit changes to their Fusion apps quickly and safely.
Add, Configure, and Test Business Rules
Fusion’s business rules manager is an intuitive interface that makes it far easier to add and modify rules and trigger specific actions as certain search queries come in. It provides an extensive list of frequently used rules, and that list is completely customizable. Merchandisers can always create new rules. When your business needs a very specific query rewrite (such as treating “i pad” the same as “ipad”) you can send individual shoppers to the best page for a particular brand.
Rewrite Zero Results and Underperforming Queries
Some queries return few results — or no results at all — turning an enthusiastic shopping experience into disappointing dead end. Fusion 4.2 trains machine learning models on signals from shopper queries and clicks to automatically identify underperforming queries and proactively suggest improved queries to the shopper. As one part of Fusion’s integrated AI engine, we’ve seen conversion rate boosts of between 12-15% (even over a very large base of historical purchase data).
Automatically Detect and Correct Misspellings
Fusion 4.2 maps misspellings submitted in user queries to their corrected spellings. When Fusion receives a query containing a known misspelling, it rewrites the query using the corrected spelling and returns relevant results, instead of showing the shopper nothing and asking them to try again.
Find Common Phrases and Word Pairs
Phrase detection identifies phrases in shopper signals and then uses intelligence from those signals to boost results from matching phrases. For example, the query “ipad case” is rewritten (before running the query) as “ipad case”~10^2, meaning that if “ipad” and “case” appear within 10 characters of each other, Fusion boosts that result by a factor of two.
Automatically Detect and Augment Synonyms
Your shoppers call things by names that aren’t necessarily exactly what’s written in your product catalog data. Synonym detection allows you to automatically replace or augment user search terms with those that are used in your collection of documents about your products.
Also Built For the Digital Workplace
Although we’ve been focussing on Fusion 4.2’s powerful new predictive merchandising suite for digital commerce applications, the Query Rewriting dashboard is also fully optimized and ready for deployment for your internal search and data apps that help your employees collaborate and serve customers more productively. The same ML and business rule features in 4.2 benefit those Digital Workplace use cases in the same ways they improve Digital Commerce.
New NLP and Deep Learning Capabilities
We’ve introduced new natural language processing index and query annotator stages, leveraging the popular open source John Snow library from Spark to allow our customers to perform named entity recognition (NER), sentence detection, and parts-of-speech (POS) tagging.
Fusion 4.2 also introduces new deep learning capabilities by integrating with Google’s popular TensorFlow library, enabling Fusion to use pre-trained TensorFlow models to calculate runtime predictions such as sentiment analysis within query and indexing pipelines.
Improved DevOps Center
Our customers working in DevOps want to rapidly implement multiple use cases for highly complex data and search applications on the Fusion platform. In 4.2, the new Fusion DevOps Center delivers that functionality and makes sophisticated management, monitoring, and debugging capabilities available through an intuitive, easy to use UI.
DevOps center is built for small operations teams managing complex deployments at scale. It makes every stage of that workflow easier, from data acquisition, storage and data processing, to the backend through machine learning, AI algorithms, and complex querying that benefit the end user. Through dedicated dashboards, DevOps teams can monitor the entire cluster, or drill down on the performance of specific hosts, services, or data sources. A log viewer supports full export for further analysis.
Other updates include forked queries for parallel processing, new query stages on both the indexing and query side, and enhanced NLP capabilities. Connectors and data sources for Dropbox, SharePoint, OneDrive, JDBC have been updated and we have a new Sitecore connector for indexing content managed with the popular CMS.
Fusion 4.2 works with Apache Solr versions 7.5 or later and Apache Spark 2.3 or later.
Contact us today to learn how Lucidworks can help your team create powerful search and discovery applications for your customers and employees.