Delivering multilingual support to customers or across support supply chains between operators still involves strategic choices. SpeakLike offers a semi-automated solution for short form communications.
Do you set up costly local contact centers in your geographical markets to handle the language? Or do you depend on the availability of language specialists in one centralized organization? And how about the general shift to online support, where there is a rapid increase in chat, email and other ‘short-form' communication methods for CS and supply chain management?
At TAUS' Berlin Summit devoted to exploring synergy between localization and CS agendas, there was noticeably more interest in the challenge of CS chat than in the automatic translation of knowledge base articles in response to customer-driven requests. Now the US-based firm SpeakLike has come up with a solution to multilingual online chat that, as usual today, combines automation with human editing.
"MT falls apart with short form or special context content," says Sanford Cohen, SpeakLike's CEO, "while mainstream translation services cannot handle the near real time turnaround at a sufficiently low cost as it is dependent on a longer translate-edit-proofread workflow."
"Our solution is to embed the translation process into the communication. This means using proprietary SMT to translate chats and emails, combined with a team of online human editors to vet quality. This lowers the final latency level of multilingual chat to typing speed. The technology speeds up the work and lowers the cost for end customers, but we deliver a business quality service. Our service costs 10 to 15% of what a traditional translation service would ask.
As a user example, a US software company that builds health care workflow solutions handles Cs messages between 72,000 doctors, technicians and patients, some of whom communicate in Spanish. The firm plans to extend to the service to French and China in line with its globalization effort.
Another consumer-focused CS customer streams the SpeakLike web service through its CRM so it can record all communications in both languages while remaining transparent to users.
To address the inevitable quality issue, SpeakLike takes a "quality by consensus" approach. "We leverage the crowd of translators to determine what the quality should be, using quality metrics that assess the whole chain from translation to customer feedback" says Cohen. This means that over time, a growing portion of CS requests will be translated automatically as the crowd agrees that a particular response is OK. The system uses various other tools and canned responses to speed up the process and enable translators to work more efficiently.
The TAUS take: SpeakLike innovates in a number of key areas to achieve time and cost targets. It addresses the issue of multilingual CS head-on, enabling users to speak their language at a lower cost than is normally possible. It deploys automation to speed up each translation, eventually leading to more translated content that can be recycled over time to simplify the underlying process. It measures quality by using both community and automation, rather than creating a central quality control bottleneck. And the SMT solution could benefit largely from access to open parallel data.