May I take your order? A Neural Model for Extracting Structured Information from Conversations

Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers, pages 450–459, Valencia, Spain, April 3-7, 2017. c⃝2017 Association for Computational Linguistics May I take your order? A Neural Model for Extracting Structured Information from Conversations Baolin Peng1, Michael L. Seltzer2, Yun-Cheng Ju2, Geoffrey Zweig2 and Kam-Fai Wong1 1The Chinese University of Hong Kong, Shatin, N.T. Hong Kong 2Microsoft Research, One Microsoft Way, Redmond, WA, USA {blpeng, kfwong}@se.cuhk.edu.hk {mseltzer, yuncj, gzweig}@microsoft.com Abstract In this paper we tackle a unique and im- portant problem of extracting a structured order from the conversation a customer has with an order taker at a restaurant. This is motivated by an actual system un- der development to assist in the order tak- ing process. We develop a sequence-to- sequence model that is able to map from unstructured conversational input to the structured form that is conveyed to the kitchen and appears on the customer re- ceipt. This problem is critically differ- ent from other tasks like machine trans- lation where sequence-to-sequence mod- els have been used: the input includes two sides of a conversation; the out- put is highly structured; and logical ma- nipulations must be performed, for ex- ample when the customer changes his mind while ordering. We present a novel sequence-to-sequence model that incorpo- rates a special attention-memory gating mechanism and conversational role mark- ers. The proposed model improves per- formance over both a phrase-based ma- chine translation approach and a standard sequence-to-sequence model. 1 Introduction Extracting structured information from unstruc- tured text is a critically important problem in nat- ural language processing. In this paper, we attack a deceptively simple form of the problem: under- standing what a customer wants when ordering at a restaurant. In this problem, we seek to convert the conversation between the customer and the order taker, i.e. the waiter or waitress, into the structured form that is conveyed to the kitchen to prepare the food, and which appears on the customer receipt. Hi, how can I help you ? We’d like a large cheese pizza. Any toppings? Yeah, how about pepperoni and two diet cokes. What size? Uh, medium and make that three cokes. Anything else? A small Caesar salad with the dressing on the side Sure, is that it? Yes, that’s all, thanks. Waiter: Customer: Waiter: Customer: Waiter: Customer: Waiter: Customer: Waiter: Customer: Figure 1: A conversation example of an order- taking interaction at a restaurant. Item Size Qty Modifiers Pizza large 1 add pepperoni Caesar Salad small 1 side dressing Diet Coke medium 3 Table 1: An example of the structured data record corresponding to the conversation in Figure 1 We develop this system to analyze real-time in- teractions with the aim of discovering errors in the order-entry process. Note that the objective is to analyze the interaction and suggest corrections to the human order-taker. Thus, we take both sides of the order-taking interaction as input, and are not attempting to predict the order-taker’s side of …

Content truncated to 500 words