Summarizing Conversations Content using ChatGPT or Google Palm API

Bharat Dwarkani
SaaS Product Engineering
2 min readJul 15, 2023

One of the most valuable applications of the ChatGPT API is content summarization, which can save countless hours and enhance the productivity. The concept of summarization can be applied to various applications, and in this article, we will explore an example within a Help Desk Software ticketing system.

Either ChatGPT or Palm API can be used for summarization purpose.

There are numerous use cases for integrating ChatGPT into customer service HelpDesk software. One such use case is summarizing a ticket conversation. This article aims to provide engineering notes on implementing the summarization feature using ChatGPT, highlighting how it simplifies a previously challenging task.

Summarization of content using ChatGPT or Palm API

The implementation of this feature is relatively simple, and the following steps outline the process:

When a user clicks, a option in UI to summarize a ticket perform these steps in backend.

  1. When a user clicks on an option in the user interface (UI) to summarize a ticket, perform the following steps in the backend:
  2. Fetch the conversation details of the specific ticket, which includes the interactions between the customer and the agent (messages exchanged).
  3. Pre-process the input. This step is crucial as it helps optimize token usage. If the ticket data is passed raw to the ChatGPT API without pre-processing, it may result in excessive token utilization.

Some items to consider while pre-processing

  • Normalize Text — Change whole content to lower casing
  • Remove HTML content and extract plain text.
  • Eliminate stop words such as “a,” “an,” “the,” “of,” etc.
  • Remove punctuations
  • Reduce words to base form by stemming text

4. Employ prompt engineering. Instead of directly passing the raw input by appending the content of the entire conversation, add context to the content. Format the content in the following manner:

Customer:
Agent:
Customer:
Agent: …

5. Call the ChatGPT Completion API with the following prompt:

Prompt: Summarize the following customer agent interaction :
>>>Ticket Conversation<<<

Note: ChatGPT 3.5 API is good enough for this purpose, it is not necessary to opt for ChatGPT 4 API. It will reduce cost if ChatGPT 3.5 is used.

By utilizing the ChatGPT API, the implementation of complex features like conversation summarization has become much simpler, streamlining the process for developers and users alike.

Start with a basic implementation using ChatGPT first. Later, you can experiment with custom or other models.

In conclusion, the ChatGPT API offers a powerful tool for content summarization, particularly in the context of HelpDesk software. By leveraging the capabilities of ChatGPT, support teams can enhance their efficiency and reduce the time spent on manual summarization tasks.

The steps outlined in this article provide a clear roadmap for implementing the summarization feature effectively. This feature can be used to summarize any lengthy articles too.

With ChatGPT, what used to be a challenging endeavor is now a straightforward process. Isn’t this simple !!!

--

--

Bharat Dwarkani
SaaS Product Engineering

Tech Enthusiast, Full Stack Software Engineer, Product Manager, Engineering Enterprise SaaS product