As a VoC (Voice of Customer) provider your main mission is to help your customers be in sync with their market. Since most feedback data created is unstructured, it has become crucial for VoC platforms to be able to convey unstructured customer feedback alongside the available structured data. The only way to achieve this is by integrating a text analytics component into the survey- or VoC platform. In this blog post we focus on technological aspects that VoC platforms need to take into account before making a decision on a partnership.
Currently on the market there are two different approaches to doing text analytics. Broadly speaking there’s the established way, keyword- or dictionary-based technology, used by current market leaders and the new way, leveraging machine learning and deep learning that’s based on the most recent research coming from academia.
In this blog post I’ll be discussing the following topics:
Types of text analytics core technology
Comparison of integration with VoC platforms
Spread of technology types across vendors
Traditional Text Analytics: Dictionary- & Rule-based
In the last couple of years, AI has become a very trendy topic. However, AI and its subdiscipline Text Analytics, have been around for a long time. These systems rely on programmed rules and predefined keywords, which are built up based on industry verticals and then customized for each client. Integrating with these systems can take months to execute and is costly because it requires specialized expertise. Any changes such as the use of a new word in the market will require the system to be reconfigured. This traditional approach to text analytics can be adequate for companies that operate in very stable industries where there is not much variation expected in the feedback provided by clients.
Next-Generation Text Analytics: Machine Learning and Deep Learning
For companies operating in dynamic marketplaces, predefined lists of keywords leave too many blind spots (unknown words are just ignored by the system). For these kinds of companies, the ideal text analytics solution should be able to infer meaning from context on the go. Used in this way, text analytics becomes a powerful early warning system, allowing companies to take action before issues become critical. For this kind of flexibility, next-generation text analytics rely on machine and deep learning methods to build complex NLP (Natural language processing) models. This involves training algorithms to “understand” language in a manner that is more similar to a human, by extracting meaning from context.
Integrating VoC Platforms with Text Analytics
For ease, I cover a comparison of how the two types of technology handle VoC platform requirements.
Machine Learning Core vs. Add-ons
The recent rise in popularity of AI in the common awareness is due to breakthroughs in machine and deep learning. The established text analytics market scrambled to sprinkle the buzzwords on top of their marketing messages. As a buyer, it’s important to establish if the core technology is running based on these modern technologies, or if they are just add-ons.
Learn more about this topic by listening to Dan Flagella’s podcast, where he’s interviewing our CEO Narjès Boufaden on machine learning.
Assessing Performance & Accuracy
Ultimately you can’t really know how well a text analytics system will do until you test it on your datasets. Make sure to use the kind of datasets that you would encounter most often. In the case of surveys, it’s important to understand that not all surveys are suitable for text analytics research. Learn more about designing better surveys for humans and AI alike. Comparing the analysis results of several systems against the same dataset is the only way of establishing with certainty which system is the right one for your company. If you’d like to learn more about benchmarking and accuracy, get in touch or ping us on twitter @KeatextAI.
SPREAD OF TECHNOLOGY TYPES ACROSS VENDORS
Recent developments in artificial intelligence has enabled the creation of a new generation of text analytics methods. However, this new technology hasn’t been accessible to established companies, since they would have to scrap their existing systems. Innovative text analytics techniques are usually encountered in smaller companies such as Keatext, with strong ties to the academic world.
Download the brochure if you’d like to learn more about our partnership program. Keatext uses machine learning and deep learning to infer what matters to customers and business, by extracting contextual insights in real time. This technology requires no human intervention, making it highly adaptable, future-proof and accessible.