Abstract:

We develop technology to improve the present and future lives of millions of children throughout the world who, while equipped with basic literacy skills, do not have access to the tools they need to develop fully functional literacy. There is a very deep need for children who are attending school to have functional access to the almost limitless amount of useful information available on the Internet. But the ability to make use of search engine technology for effective learning can not easily be acquired. For most children, even after they have learned how to post a simple query to a search engine like Google, Bing, or Baidu, the set of links that appear is, for many of them, all but useless. While the documents returned may match the content requested in the query, they are seldom suitable for young users with their developing literacy and limited scope of experience. The task is made harder by the abundance of information that is unreliable or unsuitable for young readers. In recent years, the problem has been amplified by the humanitarian crisis of millions of refugees and underage children whose only opportunity to continue their education at refugee camps lies in their cell phones or the cell phones of their parents.

We are addressing this challenge by creating a digital environment with a friendly “Virtual Librarian”, a novel web-based AI tool that allows school-aged children to submit a query and returns open or other freely available educational resources that are appropriate for their reading comprehension capability. The Virtual Librarian is empowered by a fully functional web search infrastructure. Using Machine Learning and text analysis models, it creates a personalized model for each user's knowledge of a topic and makes recommendations that reflect the user's evolving ability to process more advanced levels as their familiarity with various topics increases. This paper includes a discussion of how we address the cold start problem and experimental results from pedagogical pilot studies with refugee children in three locations worldwide. Concerence participants will be able to access a live demo.

Keywords:

AI for education, personalized learning, e-learning, NLP, Machine Learning, virtual librarian.