Empower Nepal: Advancing Women, Youth, and Marginalized Representation in Media Using AI

Background

Omdena partnered with The Asia Foundation in Nepal to address the lack of credible information and diverse narratives in the media. As part of USAID CSM M&E’s initiative, the project focused on improving the representation of women, youth, and marginalized (WYM) groups. The goal was to enhance the diversity and reliability of media content in Madhesh and Lumbini provinces by employing AI-driven quantitative methodologies, including natural language processing (NLP) and machine learning (ML) algorithms.

Objective

The project aimed to:

• Develop AI-based tools to classify and analyze media content, focusing on WYM representation.
• Measure diversity in media content using the OIND 3.1 indicator.
• Create scoring algorithms for diversity ratings in Nepali media, including social media and radio content.
• Build language models tailored to the Nepali language.

Approach

The project utilized advanced AI techniques, including:

1. Data Analysis: Collected and organized data from mainstream and social media, and Nepali radio.
2. Natural Language Processing (NLP): Developed models for Named Entity Recognition (NER) and Sentiment Analysis specific to the Nepali language.
3. Machine Learning (ML): Built classification models to rate media content based on diversity and WYM representation.
4. Interactive Dashboards: Designed and deployed visual dashboards to present findings, integrate backend databases, and improve user interaction.
5. Collaboration: Engaged cross-functional teams, including data engineers, NLP engineers, and frontend developers.

Tools and techniques included MongoDB, Python, NLP frameworks, and interactive visualization libraries.

Results and Impact

• Enhanced Representation: AI models accurately classified and rated media content based on diversity metrics, improving WYM visibility in Nepali media.
• Custom Language Models: Tailored NLP models addressed the unique linguistic challenges of Nepali, enhancing accuracy in NER and sentiment analysis.
• Scoring Algorithms: Scaled media content evaluation using the OIND 3.1 indicator, enabling reliable diversity measurements.
• Interactive Dashboards: Delivered user-friendly dashboards for stakeholders to access and analyze results, streamlining policy decisions.

This initiative contributed to fostering inclusivity and equitable representation of underrepresented groups in Nepal’s media landscape.

Future Implications

The project’s findings can inform future policy-making and advocacy for inclusive media practices. The tools and methodologies developed have the potential to be scaled across other regions and languages, addressing representation gaps globally. By driving evidence-based changes, this project sets a foundation for further research and action in media diversity and equity.