Twitter is one of the world’s most active social media platforms and customers often turn to it to express their happiness about a company. They also vent their frustrations, which regularly go viral. To keep up with the incredibly large number of tweets, modern companies need efficient solutions for opinion mining - they need a way to automate sentiment analysis in real-time.
Fortunately, new no-code tools make automating sentiment classification easy, so now every company can be AI-enabled. The best way to get started is to combine prebuilt machine learning text analysis models with no-code automation tools to monitor social media and Twitter data and flag when you get negative mentions or feedback.
Automation platforms like Zapier, when paired with Akkio’s no-code machine learning and natural language processing (NLP) models, can unlock the power of cutting-edge sentiment analysis automation for any business.
You will need to create a free Akkio account here and enter your access token into the Akkio Zapier App to enable the sentiment workflows. Each Akkio account comes with a pre-built sentiment model created with training data of 1.6 million tweets (sentiment140 dataset). The model will predict sentiment (positive or negative) given any tweet, and will also output a “polarity” or sentiment score - the probability the tweet is positive or negative.
Here are 6 ways you can use AI-powered sentiment analysis:
Twitter metrics matter. If a brand fails to respond to negative customer sentiment online, its image quickly sours and it can be difficult or impossible to recover. On the other hand, Twitter users appreciate timely, empathetic replies.
For instance, one Twitter user complained that Adidas sent an email with the insensitive subject line: “Congrats, you survived the Boston Marathon!” Adidas was quick to apologize, limiting the fallout.
In contrast, the US Army tweeted “how has serving impacted you?”, and took multiple days to give any response to the tsunami of negative tweets that flooded in. The negative replies were featured in outlets like NPR, New York Times, Slate, Time, Business Insider, and more. Talk about negative publicity.
You can use Zapier to automatically pull in new Tweets from the Twitter API that mention your company’s name (using the trigger called “Search Mention in Twitter''), and use Akkio to build an AI-driven text classifier to see if the text has a positive or negative tone. This will classify the full text of the tweet, not just the hashtags.
You can set up a Zapier “filter” to send a direct message in Slack for any tweets that have a negative sentiment. Your tweet sentiment filter can fine-tune the detection threshold - only sending along tweets where the model has an 80% or greater confidence the tweet is negative.
Just as you want to be aware of any negative sentiment about your company online, you want to hear the good news too!
If someone gave you a compliment and you didn’t even reply it’d be kind of awkward, wouldn’t it? The same holds true online and you want to acknowledge positive customer feedback. Doing so shows that you care about your customers and it is also a great opportunity to increase the visibility of positive feedback.
You can use Zapier to automatically pull in new tweets that mention your company name and use the same prebuilt sentiment model from Akkio to recognize positive tweets, just like with the first Zapier Flow in this list. This time, when a Tweet’s text or hashtags contain positive words, you can flag it for a thank-you response or even automatically retweet it in real-time.
Again, you can easily set a threshold to avoid taking action on neutral tweets - telling the workflow to only proceed if the model has a 90% or greater confidence the tweet is positive.
When someone leaves your company a review on G2 Crowd or Trustpilot, you can connect it to a text classification model in Akkio to see if the review has a positive or negative tone.
If the review is negative, you can have Zapier post a notification about it on Slack so you can respond immediately.
Facebook, like Twitter, is an important source of customer conversations.
You can use Zapier and Akkio to predict the sentiment of new Facebook posts on your page’s timeline. If the sentiment is negative, you can automatically get notified about it on Slack. Just as on Twitter, it’s vital to quickly address customer concerns and show that you care about your customers.
You can even add each new post’s sentiment score to a Google Sheet so you can monitor and create a visualization of your brand health over time.
Stock trading is notoriously difficult. While investing in a diversified portfolio over several decades has always been profitable, the same can’t be said for day trading where an estimated 80% of day traders lose money.
That said, it can be wildly profitable for those who find factors that impact markets. Elon Musk’s Twitter account, for instance, has reliably moved stock prices time and time again. When Musk tweeted “Gamestonk!!”, GME shares “rocketed as much as 157%.” After tweeting, “I kinda like Etsy,” ETSY stock jumped around 10%. Of course, correlation isn’t causation, and these tweets weren’t the only factor in these stock rallies, but there’s no doubt that Elon has a tremendous impact on the markets.
That said, Elon Musk is constantly tweeting, with around 14,000 tweets to date - the vast majority of which have no relation to stocks. Using our tutorials on trading stocks based on text sentiment, you can directly place stock market orders based on the sentiment of any Elon tweet that mentions an asset.
Just as it’s possible to place stock market orders with Zapier, through integrations like Alpaca, you can also place cryptocurrency market orders with the Cryptowatch integration.
By again using the prebuilt sentiment classifier, you can deploy a model that predicts the sentiment of tweets that mention any cryptocurrency and then place trades as the momentum of opinion swings the price.
This can be even more profitable than trading traditional assets given the immense volatility of crypto. Of course, this also means a lot more risk, so don’t trade any money that you can’t afford to completely lose - and we recommend pressure testing any trading model with paper trades before using real money.
The following workflows will require you to train a custom machine learning model - this process should take about 10 minutes in Akkio. You can connect Akkio directly to your existing data locations or simply upload a dataset via csv file to train a new model.
Financial firms are inundated with millions of credit applications. In fact, one bank, Capital One, lost 100 million credit applications in a data breach.
With over a billion credit cards in the world, an incredible amount of resources must be dedicated to approving or rejecting applications.
Using Akkio, you can build a credit approval model to (semi) automate the process of credit approvals, by learning the patterns of credit risk from historical metrics and save countless hours.
No-show appointments are a costly hassle in a number of industries. In the healthcare industry, patient no-shows can lead to adverse health outcomes and increased costs for providers and payers. For hotels, airlines, and the like, no-shows can lead to losses, especially when partial or full refunds are given and seats go empty.
Akkio can predict these no-shows, giving decision-makers the power to pro-actively rearrange schedules, implement mitigations, and more.
A full lead pipeline is a good sign, but not all leads are made equal. If you’ve gotten to the point where you need to prioritize leads, then you can train an Akkio machine learning model to calculate the likelihood of conversion. Akkio already includes an example lead scoring dataset pre-uploaded, but for this application, you will want to train a custom model with your own data.
Instead of assigning salespeople to spend an equal amount of time on all leads, you can now put your sales efforts where they matter most.
Every time a lead form is submitted on your Squarespace site, you can use Akkio to calculate the probability of conversion, add the lead to Hubspot with the probability of conversion, and kick off the sales engagement based on priority. The same integration would work for a Typeform, a Google Form, and a number of other form tools.
If you’re anything like us, your email account is constantly being flooded with messages, making it hard to stay on top of deeply important messages.
With Akkio, you can train a machine learning model to detect emails coming from potential investors or clients and use a Zapier flow that classifies incoming emails. When an email from a potential investor or client is received, it will send a notification to Slack and add the contact to a new row in Google Sheets for engagement tracking.
Note that this requires a sizable email account, with preferably hundreds (or thousands) of emails to train a model on. Virtually every email service has simple ways to export data. Here are the instructions for Gmail and Outlook.
At Akkio, we believe in a future where AI works for everyone so we are working hard to make machine learning easy - no data science or Python coding skills required. We also believe you should never have to sacrifice performance for ease of use - so we maintain open dataset benchmarks against Amazon, Microsoft, and Google’s AutoML platforms.
These 10 examples are only scratching the surface of near-limitless use cases. From natural language processing models to custom machine learning-powered Zapier flows, now anyone can automate complicated tasks and leverage AI to drive efficiency throughout your business.