Torah study accessibility: OU

The Arizona State University Artificial Intelligence Cloud Innovation Center (AI CIC), powered by Amazon Web Services (AWS), collaborated with the Orthodox Union (OU) to create a solution that modernizes Torah study and accessibility. The OU, one of the largest Orthodox Jewish organizations in the United States, partnered with the AI CIC to enhance the transcription and searchability of Torah texts, making them more accessible to a broader audience. This project focused on developing a custom vocabulary model for AWS Transcribe and used a Large Language Model (LLM) with Retrieval-Augmented Generation (RAG) to improve how users interact with and explore Torah topics.

The OU faced two significant challenges: 

1.Some ancient Torah texts in Hebrew and Aramaic were not translated to English and that created barriers to learning; 

2. The OU has a large video and audio library of primarily English lectures intermixed with Hebrew or Arameic parts, and they desired a multi-language search of this content. 

The opportunity to enable a targeted learning approach was significant. Existing transcription and search methods were often inadequate for capturing the nuances of these lectures in the context of the texts they talked about, limiting the ability of individuals to explore and study them. As a result, the OU sought a solution that would enable accurate transcription and efficient content retrieval, helping users navigate and understand Torah topics more easily.

To tackle this challenge, the AI CIC team partnered with the Orthodox Union to develop a two-part solution. They first created a custom vocabulary model for AWS Transcribe to ensure accurate transcription of Hebrew and Aramaic Torah texts. The next step involved using a large language model leveraging the RAG approach, enabling the identification and categorization of Torah topics on OU's digital platform. This approach used AWS services including AWS Lambda, Amazon Bedrock, and S3. The innovative approach resulted in a scalable, AI-driven solution that significantly improves access to and engagement with Torah study materials.

To develop this solution and advance shared learning, the ASU CIC student workers teamed up with students from varying universities that were interning at OU. The combined team met weekly to discuss solution progress and ultimately create the technical proof of concept. Special thanks to the University students for their hard work: 

• Matthew Brummund, ASU
• Arpita Mandal, ASU
• Esther Wachspress, Touro University, Lander College for Women
• Yair Lew, Touro University: Lander College for Men
• Arley Bornstein, Yeshiva University, Stern College for Women
• Chaya Perelson, Touro University
• Shoshana Szanzer,  Touro University

The success of this project demonstrates the potential for broader application across other religious and educational institutions that manage large volumes of complex, multilingual content. By leveraging AI and machine learning, organizations can develop custom solutions that make their resources more accessible, ensuring that users can engage with their content in a more meaningful way.