November 11, 2015 | Written by: Francis Friedlander
Categorized: Real-time analytics
Share this post:
Just like any other social network, Twitter is a free and abundant source of insight into what people want, how they feel and what they need. But Twitter is also special because it is highly versatile.
What people tweet about at any given moment is what they are thinking at that moment. Traditional polls and surveys measure static and coarse-grained indicators. Twitter data, however, is available in real time and is continuously updated. It is multi-dimensional and much more nuanced than polls and surveys. And unlike answers picked from a list, tweets help us to detect unforeseen trends.
Tweets are timestamped, geo-localized real-time events. If you apply the right technology to them, they will help you understand when and where trends are initiated and how they evolve. You can then use those insights to predict risks or opportunities, and take the right action in response.
There are many situations where making sense of tweets in real-time can be useful:
• A government transportation agency is expecting a cycling event downtown. How big will this event be, and how will it affect traffic? A large number of positive tweets mentioning the event indicate that many will join and that the event could result in congestion. The agency can alert people about possible congestion as soon as it’s predicted and help redirect them, based on where they live and work.
• Epidemiologists can detect a flu outbreak or trace a massive food poisoning episode by watching tweets that describe symptoms. Based on where these tweets are propagating, hospitals can be notified.
• A substantial number of users have started to complain on Twitter about a technical problem in a recently-released device, indicating there might be a flaw in the manufacturing process. The manufacturer can be alerted as soon as the number of complaints reaches a significant threshold, so a recall can be issued and the brand’s reputation can be protected from considerable impact.
These examples show how useful Twitter can be for detecting trends across populations, and it can also help us gather insights at the individual level. For example, one of our clients, a retail bank, is watching what its customers are tweeting on a voluntary (opt-in) basis. If a customer sends multiple tweets mentioning tuition fees and university applications, the bank can be notified that one of the customer’s children will likely be going to college next year and that there is an opportunity for the bank to offer the customer a student loan.
To use Twitter data in this way, two technological challenges must be addressed: First, you must filter out any irrelevant tweets. Then you can extract significant and structured information from the tweets that are left. The technology required to complete these tasks ranges from keyword detection and text mining to cognitive computing in cases where you need to extract sentiment.
Second, because of the spontaneous nature of Twitter, you shouldn’t act on every single potentially relevant tweet. Instead, you must correlate individual relevant tweets with other tweets and events to identify when to take action. The triggering conditions and thresholds must be accurate, because you want to be alerted before it is too late. But at the same time you don’t want to be alerted before a trend is shaping up.
Finally, you must alert the right individuals. For example, if you are detecting or predicting a pollution spike, you only want to notify people who suffer from lung disease and who live in an area that is likely to be affected. Unless the alert is critical, you don’t want to send a text message at 2 AM. To turn tweets into relevant insights, you will need to process them both in real time and in their proper historical context. Converting insight into relevant and customer-centric actions will require tweets to be considered in the context of each individual customer’s circumstances.
To decide when and how to act, you will need to implement sophisticated, versatile and business-driven logic. That challenge calls for a platform that provides real-time, in context and rule-based decisions like IBM Operational Decision Manager (ODM) Advanced. Take a look at this smart paper to learn more about how ODM Advanced insights can help you detect and act on risks in real time.
Join us at IBM InterConnect in Las Vegas from February 21-25. Don’t miss the following exciting sessions:
BDM-4336 : Detect trends on Twitter with cognitive computing, analytics and ODM Advanced and act in real-time, Francis Friedlander, IBM & James Casey, IBM
BDM-2672 : The patient journey leveraging IBM ODM insights and IBM ODM rules at Kaiser Permanente, Francis Friedlander, IBM & David Herring, Kaiser Permanente
In this session we will present and demonstrate how a healthcare organization can leverage Twitter, cognitive computing and ODM insights in order to be ready for major events such as a flu outbreak.
BDM-2961 : New trends in insurance with Twitter, weather forecasting and real-time actionable insight, Francis Friedlander, IBM & Elena Litani, IBM
Be sure to look for our other related sessions around the topics of ODM rules, ODM insights, Watson, Bluemix and the Internet of Things.