Enhancing Communication Equity: Evaluation of an Automated Speech Recognition Application in Ghana

Title of the Paper

Enhancing Communication Equity: Evaluation of an Automated Speech Recognition Application in Ghana

Paper Information

• Subject Area: Human-Computer Interaction (HCI), Speech Recognition, Assistive Technology, Speech Rehabilitation, and Social Accessibility
• Keywords: Speech Impairment, Automatic Speech Recognition, Amplification Theory, Accessibility, Ghana, Global South, Assistive Technology

Research Background and Issues

• Identified Problems or Challenges:

  1. In Ghana, individuals with speech impairments face barriers to interacting with others, potentially leading to social isolation and identity challenges.
  2. The availability of speech therapy and assistive technologies (e.g., speech substitution devices) is severely limited in the Global South, including Ghana.
  3. The practical value of Google Project Relate, an automatic speech recognition (ASR) application, has not yet been validated in the context of the Global South for individuals with speech impairments.

• Significance:

  1. Enhancing the communication abilities of individuals with speech impairments is a crucial means of promoting social inclusion and identity recognition.
  2. ASR technology has the potential to reduce the communication burden for individuals with speech impairments. In developing countries, such technologies could help bridge gaps in resources and services.

• Research Motivation and Related Work:

  • The disabled community in Ghana faces severe service shortages, with only a limited number of speech-language therapists (SLTs), and rehabilitation services are not included in universal health coverage.
  • Existing research primarily focuses on developed countries, neglecting the applicability and socio-cultural constraints of ASR in the Global South.
  • Amplification theory suggests that technology, without adequate infrastructure support, may exacerbate rather than reduce social inequalities.

Solution

• Methods and Steps:

  1. Introduce and evaluate the potential application of Google Project Relate in Ghana, which can create personalized speech models for non-standard speech (e.g., individuals with speech impairments).
  2. Using amplification theory, analyze differentiated access, capabilities, and motivations for ASR adoption in Ghana through qualitative and quantitative data (including observations, interviews, and self-reports).
  3. Design and implement a six-week trial study involving training for speech-language therapists and 20 participants with speech impairments.

• Innovations:

  1. The first study to investigate the practical impact of ASR tools on individuals with speech impairments in the Global South.
  2. In-depth exploration of how differentiated access, capabilities, and motivations influence ASR usage outcomes.
  3. Propose customized technology usage recommendations that integrate local language and socio-cultural contexts.

Research Findings

• Specific Results:

  1. Out of the 20 participants with speech impairments, 18 successfully completed the recording of 500 sentences for personalized speech models.
  2. The study identified multiple factors influencing the use of the Google Relate application, including device accessibility, network connectivity, language adaptability, users' literacy levels, and support systems.
  3. Highlighted limitations in the technology's design, such as limited support for non-English local languages and insufficient user training and design adaptability.

• Advantages Over Existing Solutions:

  • In the absence of customized speech recognition functions, Google Relate represents a cutting-edge advancement in current technology, and its free availability enhances potential adoption.

• Experimental or Evaluation Results:

  1. The project results indicate that ASR can significantly improve communication experiences in certain contexts, particularly for individuals with mild to moderate impairments.
  2. Participants adapted by using features such as the "custom card" function and alternative strategies (e.g., offline mode) to overcome device and language limitations.

• Limitations and Future Directions:

  1. Limitations:
     • Scope of Application: The current application only supports English, neglecting the complexity of Ghana's multilingual environment.
     • Technical Barriers: Not all participants were able to complete the 500-sentence recording, especially those with severe speech impairments.
     • Social Identity: Due to social stigma, some participants found it challenging to use the technology outside their family circles.
  2. Future Directions:
     • Improve ASR support for local languages (e.g., Ghanaian dialects or hybrid English).
     • Promote community support and social awareness training to enhance technology acceptance.
     • Enhance the cultural adaptability of AI-generated models, particularly in optimizing customization for specific languages and phrases.

Summary and Recommendations

• The study underscores the importance of integrating localized needs and user contexts into technology design.
• Designers and researchers should recognize the instrumental nature of ASR technology and support user-centered development that addresses diverse needs.
• Advocate for systemic changes from technology to social systems to achieve broader communication equity.