Workshop on Textual Customer Feedback Mining and Transfer Learning

Introduction to Workshop

For any business and organization, customer feedback, as the most direct representative of user experience, is essential for improving the quality of its products or services. The long history of breakthroughs and advancements in communication technology and social media has been playing a profound role in enhancing the capabilities of companies and organizations to understand their customers, from speech to ancient cave paintings to all sorts of social medias such as Skype, WhatsApp, Facebook, Twitter, etc., from speech interviews to paper surveys to call center services to online reviews or chats to feedback systems to social media discussion platforms. Thus, mining information and gaining insights from customer feedback through a variety of data sources like in-client systems, surveys as well as publicly available data like online reviews and social media discussions is crucial to the emerging tasks for department stores and commercial companies in understanding and optimizing users’ thoughts and attitudes towards their products and services. As increasingly high volumes of electronic customer feedback become available to companies daily in different forms such as elicited surveys, unsolicited comments, suggestions, criticism as well as kudos, either in product or in social media like Twitter/Reddit, more intelligent and automated big data analytic mechanisms in customer feedback become critically in need. In addition, textual customer feedback data is extremely noisy with lots of slang, dirty words, incorrectly translated texts, and etc., due to the fact that users all over the world have freedom to leave feedback. Text mining and sentiment analytic technology has been a powerful weapon to solve these problems.

Furthermore, in order to rapidly respond to the customer feedback and direct the issues to the appropriate production and engineering departments inside the companies, it is of great importance for them to understand what topics are people talking about their service or products and how are people feeling regarding different topics. The first problem is generally addressed by various text mining techniques, such as clustering, classification, topic modeling, text rule/ lexicon based approaches, and so on. The second problem is mainly applied by some sentiment analysis techniques, including latent semantic analysis, support vector machines, and so on. Consequently, the emerging demands and challenges facilitate the development and advancement of text mining and sentiment analysis technology for customer feedback.

Recently, transfer learning has been widely discussed on the textual customer feedback mining. Although widely applied on lots of scientific research, conventional statistical machine learning revolves on a simplified assumption that the training data, from which the algorithms learn, are drawn i.i.d. from the same distribution as the test data, to which the learned models are applied. This assumption, being broken down by numerous real-world applications nowadays, especially with the emergence of large-scale data from the private internal data, or the public Internet, has fundamentally restricted the development of practical learning algorithms. For example, intelligent recognition systems are trained to recognize the identity of faces, to classify the category of objects, or to understand the customer feedback; however, when deployed in the new environment, they may confront strange faces with significantly different facial appearances, objects in different shapes, colors, textures with different background, or feedback for different products from different groups of customers. Such issues attract substantial research attention in the era of “Big Data'', as indicated by the National Academies report on Frontiers in Massive Data Analysis:

“Data may have been collected according to a certain criterion, but the inferences and decisions made may refer to a different sampling criterion. This issue seems likely to be particularly severe in many massive data sets, which often consist of many sub-collections of data, each collected according to a particular choice of sampling criterion and with little control over the overall composition”

Target Audience: The goal of this workshop is to bring together a diverse group of practitioners, researchers, engineers and data scientists who are actively involved in processing user feedback data based on data mining and machine learning techniques to share latest findings in the field, exchange ideas on how to improve the strategies, address real-world problems, and explore a vision for the future direction and new research areas in customer feedback brought by advancement in big data analytics, data mining, machine learning and statistical learning. Moreover, the workshop will review the status of transfer learning and domain adaptation and, to discuss the challenges that we are facing, especially given enormous weakly labeled source/auxiliary data for learning tasks on the target data, and to explore future directions particularly in the unconstrained social environments, such as social media data in the cloud, Facebook and YouTube applications, and textual customer feedback mining.

Workshop Motivation

The rapidly increasing attention to customers’ satisfaction by department stores and commercial companies and development in social media as well as online review systems has promoted production and research in topic detection from customer feedback, text-based sentiment analysis, feedback analysis of NPS (Net Promoter Score), knowledge discovery from review web pages, etc. It also demands big data analytics incorporating latest advancement in data mining, machine learning, especially related to transfer learning. It is meaningful to hold this workshop to provide a platform for professionals, researchers, engineers, and data scientists to share opinions and exchange ideas, so as to enhance the research and production in text mining, sentiment analysis together with transfer learning techniques for customer feedback, the advancement in the communication between customers and companies, and the improvement of the quality of the world-wide products and service.

Primary Contact:

·       Xin Deng, xinde@microsoft.com

·       Ming Shao, shaoming533@gmail.com

Alternate Contact:

·       Y. Raymond Fu, yunfu@ece.neu.edu

·       Amrita Ray, amra@microsoft.com

·      Ross Smith, rosss@microsoft.com