Text analytics reaches new territory
The global text analytics market exceeded $5 billion in 2019 and is predicted to reach nearly $15 billion by 2025, representing a growth rate of more than 17%, according to Mordor Intelligence. Adding to predictions for strong growth, Technavio forecasts the text analytics market will grow by more than $8 billion between 2020 and 2024, expanding at a rate of more than 20% during the time period.
The increasing number and diversity of channels producing unstructured information, including blogs, chats, product reviews, emails, and multiple social media channels, provide a rich array of sources for such analytics. Major brands are making effective use of these channels for customer engagement and personalization. Although the market predictions do not take into account the economic impact of COVID-19, some sectors such as healthcare and pharma are understood to be turning to text analytics for insights related to vaccine research.
The retail sector has been a leading user of text analytics, but the technology is becoming more widely used across many industries as well. Educational institutions, financial services companies, and government organizations are among the groups that are extracting key information from sources that were previously dormant in terms of the value they offered.
Fostering engagement and support
Michigan State University wanted to build relationships with its alumni to foster ongoing engagement and support. The university had quantitative data about who had made donations, but in order to get all alumni involved in the university and its activities, the business development office needed to have greater insight into alumni’s interests and feelings. With more than half a million alumni, more spark involvement on the part of the ones who were likely to become active in university projects and motivated to donate.
The bulk of the information about the alumni was in unstructured form. The university saw text analytics as the best way to gain these more nuanced insights. Data sources included social media and records such as emails and documents that showed involvement by alumni in events that reflected their interests and concerns. The university identified IBM SPSS Modeler as the best match to harvest this information and turn it into actionable insights.
Much of the information was scattered throughout the university and was not recognized as a valuable resource. “Information was available from emails and phone call notes, committee and board meeting records, and many other sources,” said Priya Krishnan, director of offering management for IBM SPSS Modeler. “Targeting donation requests to match alumni interests and involvement is much more effective than simply looking at donation history.” By building relationships that are more personalized, engagement is increased.
As a result of using IBM SPSS Modeler to evaluate alumni sentiment and behavior, the university was able to create an affinity score that measured the individual’s attitudes and feelings. It found that up to 85% of alumni were likely to donate if affinity scores were sufficiently develop new approaches to encouraging alumni to contribute to the many projects the university wanted to maintain.
SPSS Modeler was acquired by IBM in 2009, along with the SPSS statistical package. It is now a part of Watson Studio. The SPSS text analytics tool is integrated into IBM SPSS Modeler. It uses natural language processing, so it can break down text for semantic and sentiment analysis. The fact that text analytics is integrated into SPSS Modeler and the latter into Watson Studio is critical because this configuration allows for easy analysis of structured and unstructured data together. This capability is particularly helpful for predictive analytics. “The more data you can include, the better the predictions,” said Krishnan. (More on understanding sentiment on the next page.)
Watson Studio can be used on the desktop, in the public cloud, or in a private cloud. It allows users to organize data (removing outliers, for example), and visualize it in order to see patterns. Instead of being a stove-piped function, text analytics can be fully integrated into analyses that are enriched by the additional data but can also utilize quantitative, structured data for a full picture of the issue being explored.
Text analytics is not needed for all analyses of unstructured content, Krishnan pointed out. “If you just want to sort through documents and classify them, text analytics may be overkill,” she commented. “For that type of activity, a natural language classifier would be more appropriate. However, if you are interested in understanding the meaning of the text for purposes such as concept extraction, then text analytics is the best approach.”
Much unstructured data remains unexploited. “One of the things that organizations often don’t recognize is that there is a lot of unstructured data that is just sitting there, ready to be analyzed,” Krishnan emphasized. “It is not that difficult to get started. Our solutions are coded for specific industries, so they can get immediate value and then refine over time to be more customized.”