- Example business use cases
- Getting started
- Smart Answers workflow
- Smart Answers jobs
- "Question-answering" pipelines and stages
Fusion Smart Answers brings the benefits of a versatile, scalable enterprise search platform to virtual assistants and chatbots.
This robust Q&A system makes use of advanced deep learning techniques to match question/answer pairs and enable self-service for employees and customers, increasing search relevance and saving cost on answering incoming queries.
These features bring traditional search relevancy development and data science together into an easy-to-use configuration framework that leverages Fusion’s indexing and querying pipelines.
Even if you don’t have an existing set of recorded interactions for question/answer pairs, you can still build a robust Q&A system using our cold start solution, which improves relevancy by using machine learning (dense-vector semantic search) based on content documents.
Example business use cases
Embed this system as a self-help feature on your Help page or Contact Us page to reduce the call center load.
Make it available to your customer support team to find the answers to already-solved problems.
E-commerce websites can use this system to search "how to" content, product user manuals, or existing product questions. For example, amazon.com provides a search function on questions about each product.
You can achieve fast knowledge extraction by applying this solution to these types of knowledge repositories.
The cold start solution utilizes word embedding methods to capture semantic and contextual information for long queries or natural language questions. The cold start solution doesn’t require FAQ training data. Instead, it can use word vector training combined with Solr scores using our provided query pipeline.
If signals are available, Smart Answers can also join signal data with catalog data to get query-to-product-name (question-to-answer) pairs, then train a model to help with recall at query time.
To get started, you need a trained machine learning model. There are two methods for generating a model, depending on the kind of data you already have:
|The FAQ solution||The cold start solution|
Use this solution when you already have a collection of questions that have been asked before, and their answers.
See FAQ Solution for details.
Use this solution when you have no historical FAQ data, or fewer than 200 question/answer pairs.
See Cold Start Solution for details.
Each method requires a different model training procedure, but the model deployment procedure is the same for both.
The system combines the power of Solr and neural dense vectors search from deep learning. Neural dense vector retrieval transfers documents into a digital vector space to incorporate semantic and contextual information into query understanding.
Smart Answers workflow
To implement Smart Answers, you’ll follow this workflow:
Train or install a machine learning model.
Configure the index and query pipelines.
Fusion includes default pipelines to get you started. See Configure The Smart Answers Pipelines.
Evaluate the query pipeline’s effectiveness.
Fine tune your query pipeline configuration by running a job that analyzes its effectiveness. See Evaluate a Smart Answers Query Pipeline.
Smart Answers jobs
These jobs provide the machine learning features that drive Smart Answers:
Machine learning model training:
Machine learning model deployment:
This job analyzes your configured Smart Answers query pipeline to provide data about its effectiveness so that you can fine-tune your configuration for the best possible results.
"Question-answering" pipelines and stages
Once you have trained and deployed your model, you can use one of the default pipelines that are automatically created with your Fusion app. Both pipelines are called
"-question-answering" and "-question-answering-dual-fields" index pipelines
"-question-answering" and "-question-answering-dual-fields" query pipelines