Language I/O Solution for Multilingual Customer Support

Abstract We describe Language I/O’s multilingual customer support solution. By combining intelligent selection of machine transla- tion vendors with a self-improving trans- lation process, we enable support teams to become multilingual in less than 24 hours, while maintaining ISO-27001 cer- tification and general data protection reg- ulation (GDPR) privacy standards. 1. Introduction Support is a key business process to provide the best customer experience and fuel growth. It re- quires that customers interact with global corpo- rations in their own language through any chan- nel, and the requests need to be solved efficiently and as soon as possible. This involves a signifi- cant capital expense as well as ongoing opera- tional costs. There are many pieces required to optimize customer support operations, including but not limited to chatbots and automation, live agent assistance based on machine learning, in- sights derived from data in real-time, and self- improving machine translation. Enabling a support teams to have high-quality conversations via translation technology requires the capability to learn from customer interac- tions, i.e. self-improving machine translation. This can be a difficult goal to achieve without compromising privacy considerations. In this paper , we will describe the main compo 1 - nents of the Language I/O solution that enable translation quality improvements, without com- promising on data privacy considerations. 2. System overview Figure 1 contains the key pieces in our workflow. Language I/O customers use a variety of cus- tomer relationship management systems (CRMs) like Salesforce, Zendesk, and Oracle Service Cloud. Agents can use the chat, social messag- ing, and ticketing (e-mail) solutions to communi- cate to customers whenever they get in touch with the company. When an agent receives a message in a foreign language, it will be auto- matically displayed in the agent’s own language, and she will be able to communicate with the customer to solve the issue. We can add a dis- claimer for customers to know they are reading machine translated text if needed. Language I/O takes several steps to create a seamless translation experience for both the agent and the customer. First, the integration with CRMs is essential so that agents and cus- tomers do not need to use a third-party software. We provide machine translation as a feature in the CRM interface directly to minimize training requirements and interface switching costs. Second, the best machine translation engine will vary depending on the content type, lan- guage pair, and customer, among others. Lan- guage I/O solution selects the best engine with- out any customer intervention so that the best possible output is achieved. Currently, we sup- port Google, Microsoft, Amazon, DeepL, and Systran. We select which engine is best suited for a customer by analyzing agent feedback on trans- lation quality over time. By gathering this feed- back, we are able to adjust the machine transla- tion engine to adapt to customer needs, without needing to collect sensitive chat transcripts. All of our engine integrations are vetted to ensure that the vendor does not store customer transla- tion …

Content truncated to 500 words