The evolution of enterprise search relevance
Today, individual employees still spend at least 400 hours each year searching for information, resulting in millions of dollars in operational costs and lost business opportunities. While enterprise search engines have reduced that effort with the advent of NLP technologies applied to the text found in documents, we are still in the early stages of AI-driven information discovery. The next stage of this evolution will come from applying deep learning models to enable machines to understand language intent and present more relevant information to employees, known as Natural Language Understanding (NLU). With NLU, the shift from employees finding information to information finding employees will be accelerated, further unlocking productivity and innovation.
Sinequa Neural Search - the next generation of search relevance
Sinequa’s Neural Search provides the most sophisticated engine for discovering enterprise information assets available on the market today. By combining state-of-the-art deep learning language models with the best NLP and statistical techniques, employees and customers spend less time searching for information and more time developing insights to drive decisions and solutions.
The benefits of Sinequa Neural Search include:
- The best relevance, four state-of-the-art deep learning models and hybrid search results retrieval
- Easy to deploy, no models to train
- Fast and efficient, optimized for performance using most advanced cloud GPU processing technology
"Sinequa significantly differentiates itself through its use of deep learning (artificial neural networks), and how it uniquely applies multiple deep learning models to provide more accurate search results,"
Alan Pelz-Sharpe of Deep Analysis
-
Frequently Asked Questions
What is neural search?
Neural Search is a new approach to retrieving information based on the application of advanced machine learning for Natural Language Understanding (NLU). Specifically, it refers to the use of deep neural nets (there’s the “‘neural”), a specific form of machine learning, to aid in information retrieval. These machine learning techniques search for results with a greater appreciation for the meaning of phrases and their context, instead of searching for the presence of specific words or word forms.
What use cases do we expect Neural Search to benefit?
Neural Search will improve the search result relevance for any use case where the intelligent search is already being used. Text that provides context - full sentences and long phrases - stands to show the most improvement. Examples of places where neural search augments employee abilities may include:
- Research and development, in areas such as drug discovery and product engineering
- Direct answers for customer service, compliance or risk management
- Employee onboarding
- Predictive analytics and chatbots/virtual assistants
How is Sinequa’s Neural Search different?
Unlike any other search solution available on the market today, Sinequa’s Neural Search uses four models to improve relevance: Meaning Encoder, Passage Ranker, Query Generator, Answer Finder.
Does Neural Search replace Sinequa’s existing statistical search?
No, the two methods are complementary, which is why Sinequa is taking a hybrid approach. We are adding Neural Search to statistical search, and any customers who purchase Neural Search will be using a hybrid - a combination of statistical and language-based (neural) methods.
Is Neural Search processing-intensive?
Yes. Deep learning is a computationally-intensive process. That computation is required both in training the model (which Sinequa has already done) and running the model. We have made great efforts to create the smallest viable models that produce high-quality results, but even small models will significantly increase infrastructure/compute requirements.
Due to the computational processing intensity of Neural Search, we recommend using Microsoft Azure with the latest Nvidia Tensor Core GPU.
Will Neural Search be included in the standard license ?
Once generally available, customers must specify the number of “Neural Documents” (in the same way they specify the number of Documents or a number of Records) and purchase a license accordingly. After doing so, they will receive a separate license key that will activate the neural search features.
Is it available now?
Neural Search is currently available in beta, and only existing Sinequa customers can access the beta version. We anticipate that neural search will be released as GA - general availability - in late Q4.
