The Linguist

Chartered Institute of Linguists AUTUMN 2026 The Linguist 21 FEATURES Pocketalk evaluation Pocketalk is a truly likeable AI device due to its familiarity and responsiveness. When it was first demonstrated to me, my reaction was childlike – not easy to capture as evidence, but the profound experience stays with me. The formal evaluation process happened in two stages. The first was a three-month mini- research project to ascertain the accuracy of translation. We chose two languages (Arabic and Polish), two MA/DPSI-qualified interpreters and a sample of pre-booked medical appointments. After a brief explanation of how the device works, clinicians and patients were asked to use it during the appointment. The interpreters were present to assess the accuracy and quality of the translations, and to intervene if required. Arabic Pocketalk translated mostly accurately, but there were moments when the device did not pick up the sentences, so the interpreter had to repeat them. On several occasions nonsense was delivered, but after the sentence was repeated once or twice, Pocketalk was able to produce an accurate translation. Overall it performed well. When it came to the Polish version, the interpreter observed that, on occasions, the translation was flawed, especially when a sentence was long. For example, the rendition of 'If their temperature goes above 38 degrees celsius, you need to seek urgent medical advice' omitted the temperature details, giving only the meaningless 'above c'. Some parts of the conversation flowed; other parts were simply repeated untranslated. Examples of misinterpretation included: jablka ('apples') for both 'poos' (kupa) and 'pee' (siusie); wyczuwalne ('perceptible') for 'uncomfortable' (dyskomfort) when discussing discomfort in breasts; dyskietka ('floppy disk') for 'feels floppy' (wydaje się wiotkie); and uraz glowy ('head injury/trauma') for 'headaches' (silne bóle głowy) in the sentence 'It is also important to report any severe headache'. During this stage, we did not obtain enough data or opinion-based information to warrant a roll-out to all maternity units. However, we knew more than we did prior to the pilot, could make an educated prognosis for the future, and could recommend the way forward for Pocketalk utilisation within NUH. Pocketalk is not suitable for longer bookings with an additional level of complexity or more obscure vocabulary/phraseology; where time is of the essence, as sentences often have to be repeated or rephrased to get the correct translation; or where there is a resistance to try a new device on the spot. It could do well during shorter, straightforward appointments; when telephone and video interpreting has failed completely; and out-of-hours on wards, where there might be a need for a short conversation with a patient or their family. 4 Future developments The second stage of our trial was longitudinal in nature. In a 12-month period from April 2025, we advised maternity staff to use 14 devices during their daily encounters with pregnant women in all situations where Pocketalk does well. This proved to be a success. SCALE Innovation representatives and a Consultant Midwife trained staff on how to utilise Pocketalk, including speech rendering, clarity of utterances, and monitoring patients' facial expressions to ensure understanding and reduce the risk of misinterpretation. Piloting this technology provided insights into digital and AI capabilities to support LEP patients. We expanded NUH ITS capabilities and responsiveness to demand where a professional interpreter is not available and clinical intervention is time critical to ensure linguistic assistance is always present. It is my belief that the next five years will see a new approach to language services for our healthcare clients. Thanks to NUH Midwifery Matron Debrah Neale for her enthusiasm while trialling CardMedic; NUH Community Midwife Pavanpreet Bahra for not giving up on Pocketalk; CardMedic and Pocketalk reps for their ongoing support; everyone who saw the project through to implementation; and Dr James Dalgleish for his editorial input. Notes 1 Ockenden, D (2026) 'Findings, Conclusions and Essential Actions from the Independent Review of Maternity Services at Nottingham University Hospitals NHS Trust', 125 & 321 2 Buxton, S (2026) 'CardMedic Pilot Evaluation Report: Maternity and Neonatal Service Nottingham and Nottinghamshire LMNS. Unpublished evaluation report', Nottingham: Nottingham and Nottinghamshire ICB 3 Ibid. 4 Towlson, R (2024) 'SCALE Innovation and Nottingham University Hospitals NHS Trust Pocketalk Pilot. Unpublished evaluation report', Nottingham University NHS Trust © SHUTTERSTOCK

• Cover