University of Southern California Researchers Leverage AI for Translating Bible into Rare Languages

TL;DR:

  • USC researchers have developed the Greek Room, an AI-powered technology to aid in translating the Bible into extremely rare languages.
  • The technology focuses on very low-resource languages not included in the top 500 and aims to streamline the translation process.
  • The Greek Room includes tools for spell-checking, world alignment, and detecting improper characters in a script.
  • The researchers promote a local church-driven model for Bible translation, prioritizing languages without a written system or existing resources.
  • The project aims to revolutionize the traditional Western missionary model and encourage translations into people’s heart languages.
  • The Greek Room’s current version emphasizes quality control, with plans to incorporate better translation suggestions in future iterations.
  • The researchers hope that their open-source codes and data will encourage further innovation and integration into existing translation tools.
  • The initiative is supported by the Wycliffe Bible Translators USA and aligns with Every Tribe, Every Nation’s goal of making the Scripture available in every language by 2033.

Main AI News:

The translation of the Bible into numerous languages has been a remarkable feat, with over 700 translations accomplished out of the 7,100 languages spoken worldwide. Nonetheless, there remains a significant challenge in translating the Bible into the remaining languages, many of which are exceptionally rare. In response, two scientists are harnessing the power of artificial intelligence (AI) to bridge this gap.

Joel Mathew, a research engineer, and Ulf Hermjakob, in collaboration with the University of Southern California’s Information Sciences Institute, have recently unveiled the Greek Room—a cutting-edge AI technology designed to streamline the intricate process of biblical translation. Their ultimate goal is to reach all languages spoken on Earth.

Drawing upon Hermjakob’s extensive expertise in natural language processing and Mathew’s profound understanding of Bible translation, the researchers have developed this technology to specifically target “very low-resource languages that are not even in the top 500,” according to Mathew.

The Greek Room encompasses three primary tools: spell-checking, which ensures accuracy; world alignment, which guarantees consistency in translation; and Wildebeest, a tool employed to detect improper characters within a script.

The partnership between Mathew and Hermjakob originated in 2015 when Mathew joined USC to pursue his master’s degree in computer science. It was at the Information Sciences Institute’s AI division where he crossed paths with Hermjakob, and their shared ardor for languages and Christian faith laid the foundation for their collaboration.

Having grown up with parents who were Bible translators, Mathew witnessed firsthand the challenges associated with manual translation carried out by local church members. In his hometown of New Delhi, he took note of the tasks that technology could assist with. Spell-checking, in particular, demanded significant manpower and time. In the case of translating into rare languages, only qualified local church members were available, but lacked the necessary technological support.

These are not trivial problems; these are very hard problems. But big companies are not interested in solving them; it’s not their business model to target very rare languages,” explained Mathew.

When Mathew shared the difficulties faced by Indian translators with Hermjakob, the latter eagerly seized the opportunity. “I always had this feeling to know how, at some point, I could apply my skills to my faith,” expressed Hermjakob, who holds a PhD in computer science from the University of Texas.

Their project aspires to address languages that lack a written system, grammar codes, dictionaries, or spell-checkers. “We are thinking of languages like Uyghur or Oromo,” Hermjakob stated, with Oromo being spoken in Ethiopia and northern Kenya.

Recently, the duo received a request from an Indian consultant interested in utilizing the spell-checking and world-alignment tool for Bible translation in Kolami, a language spoken by 130,000 native speakers in western India.

The Greek Room also seeks to revolutionize the conventional model of Bible translation, which was historically spearheaded by Western missionaries, limited to working on just two languages throughout their lifetime. In contrast, Mathew and Hermjakob advocate for a local church-driven approach.

Local churches and local language communities are asking for translations of the Bible in their heart language,” Mathew explained, emphasizing that in multilingual contexts, the heart language is the one in which individuals express their deepest emotions and is typically their native tongue.

The initial iteration of the Greek Room prioritizes quality control, enabling translators to allocate more attention to tasks that require subjective judgment, such as finding appropriate translations for concepts that do not exist in a given language. In subsequent versions, the researchers aim to enhance the tool’s capability to suggest better translations.

With their codes and data now available on GitHub, Mathew and Hermjakob hope that other users will integrate their research into the existing tools and drive further innovation.

Their initiative, supported by the Wycliffe Bible Translators USA organization, is part of a broader program directed by Every Tribe, Every Nation, which aims to make the Scriptures accessible in every language by 2033.

Conclusion:

The use of AI in Bible translation for rare languages has significant implications for the market. This innovative technology addresses the challenges faced by local church members and translators, enabling them to overcome language barriers more efficiently. By focusing on low-resource languages and promoting a local church-driven model, the Greek Room opens up new opportunities for businesses in the language translation industry. The project’s open-source approach fosters collaboration and potential partnerships, facilitating the integration of their research into existing translation tools and driving further advancements in language technology. Overall, this development showcases the potential of AI in language translation and highlights the importance of catering to diverse linguistic needs in the market.

Source