AI Response

Optimizing the Response Quality of Your Kipps Chatbot

With Kipps, there are several ways to improve the accuracy and quality of your chatbot's responses. Here's how you can fine-tune and optimize them.

Refine the Chatbot's Instructions

The instructions help shape your chatbot's behavior and how it responds to user queries. To make sure your bot answers only based on the information provided, be specific in the instructions. For example, you could say, "You will only provide answers based on the information in [document name]."

The default instruction for Kipps chatbot is as follows:

Ensure Readability of Uploaded Data Sources

The quality of your chatbot's responses is directly influenced by the clarity of the data sources you provide. Kipps generates responses using readable text, so ensure that the websites, PDFs, or documents you upload are text-based. Kipps does process images on PDFs but not on notion or websites. So if your data has text within an image it is better to convert it into PDF and then upload.

Kipps also takes tables and structure of data into account for PDF files.

If your chatbot struggles to answer questions from your website, it might be because the site isn’t easily scrappable. A workaround is to copy and paste the relevant information into text-based data sources or upload it as a PDF instead.

Use the "Revise" Feature and Q&A Data Sources

The "revise" feature is accessible from the dashboard in your conversation history. This tool lets you adjust responses to improve future interactions.

But don't forget to retrain the chatbot after you have revised the answer.

Create a Document Mapping URLs to Page Names

If your chatbot generates incorrect URLs that lead to 404 errors, create a document mapping all valid URLs to their corresponding page names. This can be especially useful if your chatbot is handling queries across a website with multiple pages. By providing this mapping in your data sources, your chatbot will have a better understanding of how to respond to URL-related queries.

For example, you can input a mapping document with URLs and page names like this, add it to the text data source and retrain your chatbot:

https://www.example.com/: Home Page, The main landing page showcasing the latest features, products, and updates from Example.com.
https://www.example.com/products: Products Page, Explore a wide range of products and services offered by Example.com, sorted by categories for easy browsing.
https://www.example.com/about: About Us Page, Learn about the company's mission, values, and history. This page provides detailed insights into Example.com's journey and team.
https://www.example.com/blog: Blog Page, Discover articles, tutorials, and news related to Example.com. Stay updated with the latest tips and trends.
https://www.example.com/support: Support Page, Access FAQs, user guides, and support articles. Users can also submit queries or get in touch with the customer support team for assistance.
https://www.example.com/contact: Contact Page, Get in touch with Example.com through various contact methods including email, phone, or a contact form for inquiries or feedback.