The content sources are the centrepiece for creating Digital Twins. This feature allows the user to upload content of multiple file types, like- Documents, Knowledge Base articles, Custom Question and Answer pairs, External website URLs, and Assets. The twin then uses content present in these files to answer user queries, using a technique called Retrieval-Augmented Generation (RAG).

Furthermore, you can create a Golden Test Set to validate the model's responses against the ideal answers for a set of queries. You can also generate bulk predictions for a set of queries.

Navigating to Training Content

1. Click View next to Training Content within Step 1· Establish Identity.

   ​

   ​

   ​

2. You will be navigated to the Training Content window.

   ​

   

Capabilites under Training Content

Within the Training Content feature, you have access to a wide range of capabilities aimed at managing and optimizing your content sources effectively.

1. Create, update, and delete content sources.

2. Sync data from the Sprinklr Knowledge Base utilizing filters, upload files of various types, add custom question-answer pairs, import external content by specifying a URL, which automatically scrapes the content from that link, and import assets from Digital Asset Manager.

3. Generate reports detailing similarity and ambiguity in content across different sources.

4. Upvote and downvote content types based on relevance.

5. Other capabilities include discovering content through FAQs, creating and managing a Golden Test Set for performance evaluation, and triggering and downloading bulk predictions from an Excel format for further analysis and insights.