Moving into adjacent markets can be a powerful growth strategy for language service providers (LSPs).
At SlatorCon Remote December 2022, a panel on Adjacent Markets took a deep dive into three services adjacent to translation — transcription, captioning, and linguistic validation — with Michael Rosman, VP of Corporate Strategy at Verbit and Elizabeth McCullough, Senior Director of Linguistic Validation Operations at RWS.
Rosman described how the steady rise in demand for transcription and captioning is driven by accessibility regulations and increasing volumes of video content. Growth varies by customer segment, however. “Media is pretty much penetrated already,” Rosman said. “We expect to see more growth in corporate.”
Verbit’s AI-enabled transcription and captioning proposition attracted record-breaking funding in 2021 and Rosman explained the appeal to investors, saying, “The idea of having a larger player emerge based on technology is very attractive to them.”
Rosman outlined the expert-in-the-loop approach used at Verbit: automated speech recognition (ASR) engines produce raw transcriptions (with the help of context-specific glossaries), and are then reviewed by a human editor. The ASR engine improves over time, learning from the editor’s changes.
Rosman also tackled the question of whether open-source ASR engines (such as Open AI’s Whisper) have the potential to disrupt the market for transcription services. “For consumer use, perhaps such engines will be enough, but not for high-profile and B2B use cases,” he said. “I think the age of collaboration between humans and AI is definitely here for the foreseeable future.”
Turning to captioning, Rosman shared some of the intricacies associated with caption production (such as the handling of sound effects, speaker identification, and caption placement). He also highlighted scalability as a potential barrier for LSPs entering the market.
Acquiring sufficient regulatory knowledge may also be a challenge. “Accessibility services are driven by regulations, and these differ between countries,” he said.
Assessing Synergies and Risks
The panel also explored linguistic validation (LV), a multi-step service designed to produce language versions of COAs (clinical outcome assessments) — medical questionnaires used to collect information from patients in clinical trials.
LV is generally more profitable than other language services, according to RWS’ financial reporting. Demand is also increasing due to an increase in clinical trials globally, and the more frequent use of COAs.
However, LV workflows are complex. “The primary and toughest barrier in developing an LV service is the level of expertise required in order to do it well,” RWS’ McCollough told the SlatorCon audience.
Furthermore, the risks involved can be significant. “There is a real level of responsibility associated with delivering a patient’s voice,” said McCullough. “If there was a problem in the translation of a COA, all the data [elicited from clinical trial patients via that COA] has to be thrown out.”
Slator Pro Guide: Finding Growth in Adjacent Services
LSPs can, however, take advantage of some synergies between LV and life sciences translation, according to McCollough. “There’s a lot of overlap in the demand for LV with other translation services for pharmaceutical and medical-device clients,” she said.
Existing linguist pools can be leveraged in some LV steps. “We develop what is called a harmonized translation,” McCullough explained. “Two linguists translate the COA independently and then collaborate to come up with a harmonized draft.”
On the other hand, the “cognitive debriefing” stage of LV — which involves interviewing in-country patients to test the translations — requires “a really heavy investment in training of resources in order to do it well,” McCullough said.
She added, “We need people who can do the recruiting, understand how to do a good interview, and record the results in a way that we can then look at and use as a basis for decisions.”
Your information will not be shared with third parties. No Spam.