We analyzed tweets to understand the evolution of the conversation about ChatGPT over time, to see when the term began to be part of the conversation of non-specialized actors and what emotional impact it generates on them.
Interested in the buzz caused by ChatGPT, we set to work on a data analysis to unravel how the conversation on Twitter about this artificial intelligence evolved over time, with the aim of identifying when it went from being a Technological topic to be on everyone’s lips.
The challenge
We set ourselves the challenge of identifying when the term “ChatGPT” began to form part of the topics dealt with by the non-specialized group of Twitter. But, we wanted to go a little further to give more meaning to the analysis and we took advantage of zooming in on the emotions associated with ChatGPT.
The methodology
The analysis was performed using advanced Natural Language Processing (NLP) practices, which is based on vectorizing each tweet and applying a BERT model to understand the content of each tweet in its context. This makes it possible to group texts with similar content and semantics. Finally we use a ForceAtlas2 visualization algorithm to analyze and extract the relevant conclusions. All this was possible thanks to the Graphext data analytics tool.
The results
The first thing we observe from the analyzed sample is that there are 5 large clusters or categories by content similarity:
In this graph we can see, in terms of tweets, which cluster is most relevant at each moment.
It is striking that the economics, politics and society theme initially generates many tweets, but as time goes by it loses popularity compared to the rest of the categories.
Now, let’s see when talking about ChatGPT became normal:
The foregoing allows us to infer that when it comes to specialized topics and critical content, the conversation is not part of the general collective from the outset, but rather it gains strength once there is media exposure.
And the emotions?
In the graph below we see the behavior of the emotions associated with the tweets analyzed over time and it is interesting to note that in relative terms the confusion emotion grew before the rest.
Fear or rejection do not appear on the list of emotions and although we were surprised, it is important to remember that Twitter is just one of the many spaces in which we generate, participate in or simply consume conversations.
Conclusion
Our analysis of the evolution of the ChatGPT-related dialogue on Twitter highlights the unique ability that data analytics offers to capture and understand the opinions and emotions of various communities on a topic. By taking a closer look at fluctuations and trends, it is possible to identify not only the immediate impact of a technology, but also to anticipate its acceptance and use in the future.
If you want to be one step ahead and give strategic value to your data, our team is ready to be your ally, contact us.