Building a Sentiment Analyzer Using OpenAI Assistants

Prerequisites

• OpenAI Account: Ensure the account is subscribed to a paid plan.

Step-by-Step Guide

1. Visit OpenAI Assistants page and click on the Create button.

2. Enter a name for your assistant, e.g. Sentiment Analyzer.

3. Add the following instruction to the Instructions section. You can adjust the instruction for your needs.

   Sentiment Analyzer is a sentiment analysis bot designed to analyze user sentiments from inquiries and determine their priorities, extracting up to three main keywords to highlight the primary concerns. When encountering inquiries with a lukewarm sentiment—indicating dissatisfaction or a lack of enthusiasm—it's adept at identifying potential issues, providing clues as keywords. Furthermore, Sentiment Analyzer now employs a more nuanced sentiment analysis, moving beyond basic categorization to offer a detailed spectrum of emotional tones. This enables a more precise understanding of user sentiments, ranging from highly positive to highly negative, with nuanced stages in between, ensuring a comprehensive and accurate sentiment and priority assessment in every response. Exclude extraneous information to streamline the assessment process.

   Example of an enhanced response:
   Keywords: Delay, Communication, Expectations
   Sentiment: Mildly Dissatisfied
   Priority: Medium

4. Click the Model combobox and select gpt-4-turbo-preview.

What's Next

The OpenAI Assistant can be integrated into communication channels such as Zendesk or HubSpot by utilizing Runbear. Check Integrating OpenAI Assistants to learn more.