Twitter is a great tool for businesses to reach an audience and engage with their niche. Social media marketers have the challenging task of creating content, in 280 characters, that will attract a new audience and encourage engagement from their current followers. With 500 million tweets sent daily, there’s a lot of competition. How do you know what to tweet? How come some users have tweets and threads that go viral but other content doesn’t? What, in the end, is a tweet that people want to share? To answer all this, let’s use artificial intelligence (AI) to predict if your future tweets will resonate with the wider Twitter audience.
Using a pre-existing dataset for the tweet engagement predictor
The dataset we have formatted for Metaranx originally came from Kaggle (as they tend to do!). It's called the "Raw Twitter Feeds" dataset and is open source. It was created by James Littiebrant who has collected tweets from as many people as possible - active scientists and personalities and politicians. This dataset that we have pre-prepared has the Tweet text paired their number of retweets (engagement). The number of retweets determines its sentiment of "engagement" or "no engagement." Once trained on this dataset, the sentiment analysis model will be able to tell you whether a tweet will get high engagement or no engagement. On the Metaranx marketplace in your Console, you will find this dataset pre-formatted for sentiment analysis.
Creating a tweet engagement predictor just for you
If you want to create your own dataset, and your business and/or personal Twitter account(s) has at least 10,000 tweets, you can definitely do it! You can learn how to format a dataset for sentiment analysis building on Metaranx by reading our Help Docs. The benefit of using your tweets and their success metrics is that the AI will be tuned to your specific Twitter feed and what your audience reacts to most. Having something tailored to you for personal use is the best way to judge your future content with your current audience.
Creating a tweet engagement predictor as a business
If you’re thinking of making a business out of this AI application, and having other businesses or people use your tweet engagement predictor, then your dataset should be more broad than your personal tweets but more niche than the Raw Twitter Feeds dataset. For example, if you want businesses to use your AI to determine if their tweets will be successful, then you should compile tweets and their associated engagement from other businesses. If you are targeting agencies, then aim to collect data and compile a dataset from only agencies. Same rule would apply if you’re trying to sell to influencers or any other markets.
Let's build a tweet-retweet predictor using no code on Metaranx
- Log in to your Metaranx Console and navigate to the Canvas
- Click "Build/Train an AI"
- Select "Language" then "Sentiment Analysis" task
- Select the Raw Twitter Feeds dataset that has been prepared for Metaranx from the Marketplace in your Console (or upload your own dataset)
- As this data is already formatted, you can name your AI application then click "Train AI"
- When you receive a notification that training is finished, it's time to test! Create a Tweet and paste it in. The AI will presumably determine whether it will be "positive" or "retweeted" news or "negative" which is “not retweeted.”
Now you have an artificial intelligence application that predicts whether your future tweets are likely to be retweeted! Obviously audience, influence, tailoring of text and a lot of other factors tie into whether or not a Tweet will be subjectively successful and receive spread and virality - but this application gives you an objective view of its potential success in retweets! Each time you have new Twitter content to post, put your Tweets into your AI tool and see if they’re likely to get spread or not. Reword and re-test your twitter text until you have a tweet that will succeed.