All Collections
What's New?
November 2023
Introducing Summarization on TRAC Narratives powered by GPT 4
Introducing Summarization on TRAC Narratives powered by GPT 4
Dominick DiCesare avatar
Written by Dominick DiCesare
Updated over a week ago

Following the successful launch of Pulsar 3.0 and the innovative introduction of Narratives on TRAC, we are thrilled to announce our latest advancement: Narrative Summarization on TRAC. This enhancement is a direct integration with the capabilities of Chat GPT 4.0, designed to transform how you comprehend and interact with the narratives in your TRAC searches.

Why Are We Doing This?

If you’ve been following our journey over the past 18 months, you know that we’ve been rolling out Pulsar 3.0, which is focused around several major revolutions. The continuous integration of AI tools to help with producing relevant insights, is key to our Extreme Relevance revolution. The focus is on helping you get what is relevant to you and your search in the most convenient way possible. This initial launch of Narrative Summarization is merely our first step in the expansion of our analysis methods on TRAC leveraging the latest in cutting-edge AI tech.

Narrative Summarization strikes at the heart of why being able to isolate noise from signal in social listening is important. When we run searches, we try to understand the heartbeat of a conversation around a given topic, or glean insights into a community of individuals interacting with each other as a community or audience. The Narratives on TRAC functionality draws from the content that makes up these conversations and interactions, and the summarization helps you to understand what the focus of the content is at a glance.

How Does it Work?

Narratives Summarization on TRAC brings a new layer of convenience and efficiency to how you can interact with, and understand the various narratives within your TRAC searches. Each narrative identified in a search will now be accompanied by a brief 20-word summary. These summaries appear at the top of each narrative in the 'Narratives' tab of your TRAC searches, and the summary is crafted to provide a clear and concise overview of the topic being discussed within a narrative. This means users can now quickly grasp the essence of each narrative without reading through the entire content.

When you go into the narratives tab, you have the option to click into any individual narrative.

From here, the summary of that narrative will appear at the top of the page, giving you a chance to quickly understand the narrative. Words that help to segment the narrative from others in the search are highlighted as a part of the summary, giving you a perfect caption for any charts or reports you might want to build.

The Narrative Summary is built using the top content found in the specified narrative. Using a specialized prompt in combination with this content, ChatGPT crafts a single-sentence summary keeping the focus of the search (the title) and the major connective points of the given narrative (the segment keywords) in mind. Because GPT 4.0 is Non-Deterministic, changing around your filters, changing the date range, and even coming back to your narrative later can result in a completely new summary that reflects any slight changes that can occur within the narrative.

What Do I Need to Do?

Nothing! This feature works without any interference from the user. However, it's worth pointing out a few helpful tips to get the most out of your narrative summaries:

  • Make sure your search's name is short and to the point, and accurately reflects the main topic of your search (i.e. if your search is focused on baseball teams in the USA, try naming your search "US Baseball Teams" rather than "Search: Baseball in the United States Topic Search #1")

  • Try looking at longer date periods, or filtered datasets that ensure you have at least 100+ results to analyze.

This isn't the end of our potential integrations with GPT. This feature is definitely in beta, so if you notice any irregularities, have any questions, or want to suggest any new features, please don't hesitate to reach out to us via our support channels.

Did this answer your question?