No abstract

Intelligent personal assistants (IPAs) such as Amazon Alexa, Google Assistant and Apple Siri extend their built-in capabilities by supporting voice apps developed by third-party developers. Sometimes the smart assistant is not able to successfully respond to user voice commands (aka utterances). There are many reasons including automatic speech recognition (ASR) error, natural language understanding (NLU) error, routing utterances to an irrelevant voice app or simply that the user is asking for a capability that is not supported yet. The failure to handle a voice command leads to customer frustration. In this paper, we introduce a fallback skill recommendation system to suggest a voice app to a customer for an unhandled voice command. One of the prominent challenges of developing a skill recommender system for IPAs is partial observation. To solve the partial observation problem, we propose collaborative data relabeling (CDR) method. In addition, CDR also improves the diversity of the recommended skills. We evaluate the proposed method both offline and online. The offline evaluation results show that the proposed system outperforms the baselines. The online A/B testing results show significant gain of customer experience metrics.

Abstract. The collaborative creation of value is the central tenet of services science. In particular, then, the quality of a service encounter would depend on the mutual expectations of the participants. Specifically, the quality of experience that a consumer derives from a service encounter would depend on how the consumer's expectations are refined and how well they are met by the provider during the encounter. We postulate that incorporating expectations ought therefore be a crucial element of business service selection. Unfortunately, today's technical approaches to service selection disregard the above. They emphasize reputation measured via numeric ratings that consumers provide about service providers. Such ratings are easy to process computationally, but beg the question as to what the raters' frames of reference, i.e., expectations. When the frames of reference are not modeled, the resulting reputation scores are often not sufficiently predictive of a consumer's satisfaction. We investigate the notion of expectations from a computational perspective. We claim that (1) expectations, despite being subjective, are a well-formed, reliably computable notion and (2) we can compute expectations and use them as a basis for improving the effectiveness of service selection. Our approach is as follows. First, we mine textual assessments of service encounters given by consumers to build a model of each consumer's expectations along with a model of each provider's ability to satisfy such expectations. Second, we apply expectations to predict a consumer's satisfaction for engaging a particular provider. We validate our claims based on real data obtained from eBay.

Abstract-Contracts are legally binding descriptions of business service engagements. In particular, we consider business events as elements of a service engagement. Business events such as purchase, delivery, bill payment, bank interest accrual not only correspond to essential processes but are also inherently temporally constrained. Identifying and understanding the events and their temporal relationships can help a business partner determine what to deliver and what to expect from others as it participates in the service engagement specified by a contract. However, contracts are expressed in unstructured text and their insights are buried therein. Our contributions are threefold. We develop a novel approach employing a hybrid of surface patterns, parsing, and classification to extract (1) business events and (2) their temporal constraints from contract text. We use topic modeling to (3) automatically organize the event terms into clusters. An evaluation on a real-life contract dataset demonstrates the viability and promise of our hybrid approach, yielding an F-measure of 0.89 in event extraction and 0.90 in temporal constraints extraction. The topic model yields event term clusters with an average match of 85% between two independent human annotations and an expert-assigned set of class labels for the clusters.

Intelligent personal assistants (IPAs) such as Amazon Alexa, Google Assistant and Apple Siri extend their built-in capabilities by supporting voice apps developed by third-party developers. Sometimes the smart assistant is not able to successfully respond to user voice commands (aka utterances). There are many reasons including automatic speech recognition (ASR) error, natural language understanding (NLU) error, routing utterances to an irrelevant voice app, or simply that the user is asking for a capability that is not supported yet. The failure to handle a voice command leads to customer frustration. In this article, we introduce a fallback skill recommendation system (FROST) to suggest a voice app to a customer for an unhandled voice command. There are several practical issues when developing a skill recommender system for IPAs, i.e., partial observation, hard and noisy utterances. To solve the partial observation problem, we propose collaborative data relabeling (CDR) method. To mitigate hard and noisy utterance issues, we propose a rephrase-based relabeling technique. We evaluate the proposed system in both offline and online settings. The offline evaluation results show that the FROST system outperforms the baseline rule-based system. The online A/B testing results show a significant gain of customer experience metrics.