The Customer

A leading auto manufacturer benefits from being able to quickly and accurately catalog customer issues and related solutions into its central management system. This allows for efficient identification of problems and solutions based on past experiences. However this classification task requires being able to sift through large amounts of customer service records and internal memos, and then sorting them into relevant groups for the catalog. While sorting is typically not a problem for AI, the prevalence of different text jargon from wildly different sources causes major issues.

The Solution

Text data with varying syntaxes poses a serious challenge for AI. A mechanic and a customer may describe the same problem using different words and phrases, and it is critical to know that these two accounts belong in the same class. tackles this problem through the use of Deep Learning. Because the platform has been pre-trained using a large corpus of knowledge, it is capable of looking beyond syntax through a variety of techniques like using word embeddings, Word2Vec, Char-level CNN, and analyzing textual features (e.g. editing distance, BM25, etc.).

This Deep Learning approach helps the generated AI model find common concepts between different syntaxes. These building blocks are then put together in Meta-learning to produce a highly optimized and accurate model. In this case the actual model created by performed with 98% accuracy.

The Benefits

An increasingly accurate catalog also leads to increased productivity and a better customer experience as customer cases are handled faster, and a centralized source of information allows accurate statistics to be gathered, providing insights into customer issues and quality control trends. Additionally, when it comes time to update the system, the platform nature of allows a new classification model to be generated within hours.

Create Your Free Account

Share this Post