AI-CourseEvaluations

This is a Proof of Concept (POC) demonstrating how AI can enhance the safety and efficiency of processing free-text course evaluations by detecting potential threats and categorizing feedback for manual review.

Project Overview

This POC offers two approaches to classify course evaluations:

• Azure Logic App: Orchestrates text classification using Azure Language Services and SharePoint lists. This method is better suited for non-developers.
• Azure Function: Classifies text using Azure Language Services. This approach is ideal for developers who prefer handling input/output directly.

Select the approach that best fits your needs. The Logic App/SharePoint approach is better suited for 'non-developers', whereas the Function App only processes the data and leaves the input/output to you.

Note: Azure Content Safety produced inconsistent results for texts in Swedish, so a translation option has been added. Sentiment analysis is performed on the original text.

Setting Up the Logic App Workflow

Prerequisites

• An Azure subscription with permissions to create resources.
• Access to a SharePoint site with permissions to create lists.
• A service account or similar credentials to authenticate the Logic App or Azure Function with the SharePoint site.

Initial setup

1. Create the necessary Azure resources as outlined in the prerequisites section.
2. Set up SharePoint lists as described in the SharePoint Lists section.
3. Deploy resources following the instructions in the Deployment section.

SharePoint Lists

Import the templates to your SharePoint site:

• MainListTemplate.csv: Main list template with example data.
• SettingsTemplate.csv: Settings list template containing default values.

Modify the settings list as needed to match your requirements.

Deployment of the Logic App and the necessary resources

Note: The parameters below are sample values.

1. Create a resource group to deploy the Logic App and related resources:

   az group create --name course-eval-ai-rg --location swedencentral
   

2. Deploy using the main.bicep file. You can set parameters directly in the file or via the --parameters flag.

   The parameters needed are:

   • namePrefix: Prefix for the names of the resources. Default is 'course-eval-ai'.
   • sharepointSiteUrl: URL of the SharePoint site.
   • sharepointListMainName: Name of the main SharePoint list.
   • sharepointListSettingsName: Name of the settings SharePoint list. Default is 'Settings'.

   Deployment Commands

   • If parameters are set in the main.bicep file:

     az deployment group create `
         --resource-group course-eval-ai-rg `
         --template-file main.bicep 
     

   • If parameters are not set in the main.bicep file:

     az deployment group create `
         --resource-group course-eval-ai-rg `
         --template-file main.bicep `
         --parameters `
             namePrefix=course-eval-ai `
             sharepointSiteUrl=https://somesharepoint.com/sites/somesite `
             sharepointListMainName=CourseEvaluations `
             sharepointListSettingsName=Settings
     

3. Post-Deployment Steps:

   1. In the Logic App, go to the Connections tab and authenticate the connector with a user that have access to the SharePoint lists.
   2. Enable the Logic App.

Usage

Check the Wiki.

Implementing the Azure Function for Text Classification

For detailed information on setting up and using the Azure Function, refer to the readme.md for the Azure function.