How to Create a Knowledge Base for a Question-Answering Experiment

This tutorial shows how to create a Knowledge Base (KB) for a Question-Answering experiment in Cogniflow.

In this example, a Question-Answering experiment is created to use pre-trained models capable of answering questions on a Knowledge Base.

1. Create the Knowledge Base

Create a KB with text documents containing information you want to ask questions on and place it in a single folder.
Think your KB as a repository of information about one or more topics of your interest.
Supported file formats to upload include most common document types: txt, docx, pptx, pdf, etc.
Important: Tables and lists are not fully supported yet. For best performance, we recommend using paragraph-based documents.

A small toy KB consisting in docx, pdf and txt files about No-Code AI is shown below:



2. Create the ZIP File

Select the KB folder and compress it into a single ZIP file. Now the KB is ready to be uploaded to Cogniflow.

3. Upload your Knowledge Base

When creating your experiment you will reach the step when a KB has to be uploaded. Click on "Browse your files" and upload the ZIP file generated before.





When the file upload is complete, click on "Next step".

4. Check if Everything is OK

After uploading your KB, the QA model generation process is started. In the "Dataset" tab you can see more details about the uploaded KB. Also, it is possible to download the data.



Example Datasets

No-Code AI KB: Create a QA experiment to ask questions about No-Code AI. The KB has four files, one .docx, one .pdf and two .txt files.

English Small AI KB: Create a QA experiment to ask questions about AI (English). The KB has three text files, with data extracted from Wikipedia.

Spanish Small AI KB: Create a QA experiment to ask questions about AI (Spanish). The KB has three text files, with data extracted from Wikipedia.
Was this article helpful?
Cancel
Thank you!