Golden Retriever

information retrieval engine for your text documents

About Golden Retriever

Searching for relevant information in text is a common challenge for many organizations. Organizations collect lots of information over time in sprawling documents. Finding an answer to a specific question in these documents require human labour and time to read and understand them. 

Golden Retriever allows you to search existing documents (e.g. Frequently Asked Questions, Terms and Conditions, etc.) with distinct sections using natural language queries. You can feed documents to Golden Retriever, make queries, and have it return you the most relevant answers. 

Golden Retriever can be further fine-tuned with questions specific to your documents to improve accuracy.

Key Features

Semantic Similarity Matching

Perform semantic similarity matching between queries and terms

Conversational Language

Ask questions in natural, conversational language

Integrate with Chatbots

Can be integrated into existing chatbots to collect questions

High Coverage

Return one best response or multiple responses for higher coverage

Improve Accuracy

Fine-tune questions to improve accuracy

How it Works

  1. Prepare the document. Document should be distinctly sectioned with spaces between each chunk of information. 
  2. Load the document into Golden Retriever.
  3. Choose the number of answers you want per query (we recommend 1 – 3 responses).
  4. Make queries and receive answers!
  5. Model can be improved with a list of questions tagged with correct responses.

Automatic responses for FAQs

Automatically give responses when site visitor types questions

Search through T&Cs

Assist officers in answering customer queries regarding T&Cs

Searching through contracts

Query contracts, and look for corresponding terms between multiple contracts

Use Cases

Source Repositories

For developers