In the insurance industry, timely and effective interaction with customers are at the core of everyday operations and processes that are key for a satisfactory customer experience. These interactions often result in sequences of data derived from events that occur over time. Such recurrent patterns can provide valuable information that can be used in a variety of ways to improve customer related work-flows. In this paper we demonstrate the application of a recently proposed algorithm to uncover such time patterns that takes into account the time between events to form such patterns. We use temporal customer data generated from two different use-cases (satisfaction and fraud) to show that this algorithm successfully detects patterns that occur in the insurance context.

Insurance companies manage a large number of claims on a daily basis as new claims are reported and existing claims are serviced. A key component for servicing a claim is the ability for Claims personnel to enter in raw text, aka claims notes. Claims notes contain invaluable information often beyond that of structured data, capturing this information in a machine learning setting offers remarkable benefits to many downstream tasks in a Claims department. The ability to leverage claims notes enables an insurance company not only to make data-driven and insightful decisions while handling claims, but to create value through working more efficiently and serve their customers more effectively. To best leverage the information contained claims notes, we develop insurance-based language models (IBLMs) by further pre-training existing general domain language models (ULMFiT and BERT) on a large number of claim notes with enhanced vocabulary. Furthermore, we tested these IBLMs against three downstream binary classification tasks: (1) identification of auto claims with attorney retention, (2) bodily injury prediction, and (3) auto claims fraud investigation detection. We train different classifiers based on claims notes available on day 1 and through day 10 from when the claim was reported. We found that IBLMs show a significant improvement over the traditional classification approaches. Further, we provide practical insight into how an insurance company might use these models through the analysis of volume (capacity) thresholds.

We are presenting a method for the enhancement of speech in speaker dedicated speech communication systems. The proposed procedure is fundamentally different from most state-of-the-art filtering approaches. Instead of filtering a distorted signal we are re-synthesizing a new ldquocleanrdquo signal based on its likely characteristics. These characteristics are estimated from the distorted signal. We present a successful implementation of the proposed method for a communication system for which speaker enrollment and noise enrollment are feasible. Forty minutes of clean speech training data is usually sufficient for successful denoising. The proposed method compares very favorably to other state-of-the-art systems in both objective and subjective speech quality assessments.