Converseon Advances AI-Based Text Analytics

Converseon, a provider of social listening and voice-of customer technology, is rolling out a new version of its AI-powered SaaS text analytics solution, Conversus.

According to the company, Conversus 4.0 puts the power of machine-learning in the hands of users via an intuitive dashboard to allow them to generate actionable insight and business results by separating meaningful signals from the noise associated with unstructured social and VoC data.

Conversus allows users to clean and normalize unstructured data, but also build machine-learning-powered custom classifiers to unlock the value of social listening data. Examples of these classifiers include sentiment, emotion, intensity, advocacy, customer care, purchase intent, brand health attributes, customer care, predictive analytics and customer experience analysis. The active learning approach enables users to go beyond basic text analytics analysis capabilities to custom-tune the analysis to client requirements with human-level precision, but at the speed and scale of software.

In addition to social listening data, other sources for Conversus analysis include survey verbatims, call center/focus group transcripts, and customer reviews.

Built upon Converseon’s proprietary training corpus of millions of human-coded training records gleaned over the last decade, Conversus says it also offers users access a range to off-the-shelf custom classifiers for verticals, including pharma, financial services, hospitality, consumer products, entertainment, B2B, and retail. Automated F1 performance scoring allows users to understand the precision and recall of their classifiers, so the data can be used with confidence for reporting and modeling.

Conversus-generated data is being used in quantitative models, including market-mixed modeling, predictive analytics, and brand health. According to the company, for customer care, the solution was able to improve upon the data relevancy of an automotive company-built customer classifier by more than 300% in less than 2 weeks compared to a traditional Boolean-based approach.