Automatic Comprehension of Customer Queries for Feedback Generation

Abstract

One major challenge in customer-driven industries is the response to large volumes ofqueries. Inresponsetothisbusinessneed,FrequentlyAskedQuestions(FAQs)have been used for over four decades to provide customers with a repository of questions and associated answers. However, FAQs require some efforts on the part of the customers to search, especially when the FAQ repository is large and poorly indexed or structured. Thisevengetsdifficultwhenanorganisationhashundredsofqueriesinits repository of FAQs. One way of dealing with this rigorous task is to allow customers to ask their questions in a Natural Language, extract the meaning of the input text and automatically provide feedback from a pool of FAQs. This is an Information Retrieval (IR) problem, in Natural Language Processing (NLP). This research work, presents the first application of Jumping Finite Automata (JFA) — an abstract computing machine — in performing this IR task. This methodology involves the abstraction of all FAQs to a JFA and applying algorithms to map customer queries to the underlying JFA of all possible queries. A data set of FAQs from a university’s Computer and Network Service (CNS) was used as test case. A prototype chat-bot application was developed that takes customer queries in a chat, automatically maps them to a FAQ, and presents the corresponding answer to the user. This research is expected to be the first of such applications of JFA in comprehending customer queries.