SBIR Phase I: Artificial Intelligence (AI)-Enabled African Language Database

这个小企业创新研究（SBIR）第一阶段项目的更广泛/商业影响是为非洲语言创建一个音调熟练的人工智能（AI）翻译数据库。没有这样的产品或服务可以准确地翻译非洲语言，因为非洲语言传统上一直缺乏西方公司的资源。到2050年，全球近25%的人口将来自撒哈拉以南非洲，目前超过60%的非洲人年龄在25岁以下。预计到2024年，非洲大陆的消费者和企业支出将达到5.6万亿美元，美国将投资超过3.5亿美元，以扩大数字接入和扫盲，并促进美国企业在非洲大陆的投资。通过扩大准确翻译和学习非洲语言的机会，该项目将支持美国和非洲国家的经济增长，并通过促进与非洲裔美国人和新移民的沟通来支持健康和福利。非洲语言非常多样化，整个大陆有2000多种不同的语言。它们对于非母语人士来说很难学习，翻译应用程序也很难正确解释，主要是由于音调和喉音以及轻微的发音差异，使得类似的发音单词具有完全不同的含义。拟议的AI数据库是同类中的第一个。该项目将建立数据处理，模型训练和数据库评估步骤，以生成支持AI的数据库。我们的目标是训练一个数据库来破译这些音调变化，并确保传达正确的含义，从一个包含单词和短语的正确发音音频和视觉示例的大型数据集开始。该项目的主要目标是在数据库中开发入门级、一致的机器学习（ML）核心功能和多模式交互，可用于创建其他基于音调的语言ML/AI数据库。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader/commercial impact of this Small Business Innovation Research (SBIR) Phase I project is creating a tonally proficient Artificial Intelligence (AI)-enabled translation database for African languages. There are no such product or service that can accurately translate African languages, as African languages have traditionally been under-resourced by Western corporations. By 2050, almost 25% of the earth’s population will be Sub-Saharan African and currently more than 60% of Africans are under 25 years old. The African continent is projected to have $5.6 trillion in consumer and business spending by 2024 and the U.S. is investing over $350 million to expand digital access and literacy and promote U.S. corporate investment in the continent. By expanding opportunities to accurately translate and learn African languages, this project will support economic growth for both the U.S. and African countries and support health and welfare by facilitating communication with African-speaking Americans and recent immigrants. African languages are very diverse with more than 2000 distinct languages across the continent. They are difficult for non-native speakers to learn and for translation apps to correctly interpret, primarily due to the tonal and guttural sounds and slight pronunciation differences that make similar sounding words have completely different meanings. The proposed AI-enabled database is first-of-its kind. The project will establish the data processing, model training, and database evaluation steps necessary to produce AI-enabled databases. The goal is to train a database to decipher these tonal shifts and ensure that the correct meaning is conveyed, beginning with a large dataset of correctly spoken audio and visual examples of words and phrases. The primary objective of this project is to develop the entry-level, consistent, Machine Learning (ML) core functionalities and multimodal interactions in a database that can be utilized in the creation of other tonally based language ML/AI databases.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.