Page 21 - AI Ready – Analysis Towards a Standardized Readiness Framework

AI Ready – Analysis Towards a Standardized Readiness Framework

3.5 Case Study-5: Regional Customizations

This case study involves use cases that may require taking generic solution pipelines, customizing
inferences, and applications impacting regions, e.g. accent training for voice-based solutions.

Khmer telemedicine chatbot from Neak Pean HealthTech [2] attaches great importance to
tuning the AI solution to the needs of the local community and letting the local people benefit
from the technology.

Accessing quality healthcare poses significant challenges in rural and remote areas. Long wait
times, inefficient patient preliminary assessment, and limited access to medical advice hinder
healthcare delivery and patient outcomes. According to Cambodia’s HealthTech Roadmap
(MISTI, 2022), Cambodia faces a shortage of healthcare professionals. It is estimated that there
are 6.9 nurses and 1.9 doctors per 10,000 patients, which is the lowest rate in the region. With
the innovative application of AI, Neak Pean HealthTech, the Khmer telemedicine chatbot is to
revolutionize healthcare accessibility in Cambodia.

Neak Pean HealthTech integrates advanced technologies like natural language processing
(NLP) and Khmer automatic speech recognition (ASR) to facilitate efficient communication
between patients and healthcare providers. This bridges the gap between patients and
healthcare providers, especially in remote and rural areas by reporting symptoms remotely,
detecting keywords, generating summary reports, scheduling appointments, storing medical
records, and accessing medical advice in Khmer.

The challenging part of this use case lies in processing audio data in the local language with
various accents. Khmer has a complex script, making typing difficult for many Cambodians. As
a result, voice input is more commonly used. Patients can talk or text in Khmer with the chatbot
of the Neak Pean platform to report their symptom or their health concerns as a preliminary
assessment before meeting doctors, which is believed to be patient-friendly. In some cases,
as the local community does not speak the language the model or the algorithm is using,
the indigenous people cannot benefit from the technology. By using the local language, this
platform bridges the pervasive gap in education level or socioeconomic status, reducing the
inequalities in the distribution of medical resources.

To successfully facilitate the use of local languages with different accents, this platform employs
keyword detection algorithms to extract crucial information from patient inputs. Then it will
automatically generate a comprehensive summary report for health for healthcare providers’
review. Patients can then schedule appointments seamlessly, with the platform facilitating
efficient communication between patients and healthcare providers. Importantly, patients have
access to medical data in Khmer, empowering them to make informed decisions about their
health and well-being.

14