Decoding the Hidden Value of Unstructured Text Data

January 25, 2019 by Jason Rodriguez

Data science is about supporting decisions with insights, advice or predictions based on rigorous analysis of prior business facts, decisions and outcomes. One of the biggest challenges for insurers in becoming more data-driven is getting access to a sufficient volume of reliable data in a usable form (Figure 1).Executive SummarySentiment analysis and latent semantic indexing are two of the text mining techniques that can help claims handlers unlock the hidden value of unstructured text data, improving prediction accuracy and creating decision-making engines that match closer to human performance.

Executive Summary

Sentiment analysis and latent semantic indexing are two of the text mining techniques that can help claims handlers unlock the hidden value of unstructured text data, improving prediction accuracy and creating decision-making engines that match closer to human performance.

The most usable form of data is a table where each row represents an observation (i.e., a policy, a vehicle, a submission, a claim) and each column contains understandable and useful attributes of those observations or an outcome. Data in this form is commonly referred to as structured data. Unstructured data refers to all other formats and includes text documents, images and sensor data.