Ming Jiang, assistant professor of Data Science in the Human Centered Computing Department at Luddy Indianapolis, is Principal Investigator and recipient of a two-year, $174.9K National Science Foundation (NSF) EAGER grant. EAGER (EArly-concept Grants for Exploratory Research) funding supports exploratory work in its early stages on untested, but potentially transformative, research ideas or approaches that may involve radically different approaches, apply new expertise, or engage novel disciplinary or interdisciplinary perspectives.
The long-term goal of Jiang’s project, EAGER: HCC: Mining the Potential of Language Technologies for Human-centered Cultural Competence, is to identify and validate principles that can inform the next generation of AI-driven language technologies (like Google Translate or AI chatbots) in supporting communication across different cultures, an aspect that current Natural Language Processing (NLP) techniques do not fully consider.
If successful, this research could provide a solid basis to innovate language technologies in such a way that they are better at avoiding misunderstandings that often happen when translating or interacting between languages. This means the technology would not only focus on correct words or grammar but also gradually understand the cultural context of the conversation, which is crucial for effective real-world communication.
“The overall objective of this proposal is to expand our understanding of NLP systems from a content-based paradigm to a culturally aware perspective, with an emphasis on the systems’ capabilities to seek common ground across geo-cultures,” Jiang said.
Making NLP Systems culturally aware
Natural Language Processing is the technology that allows computers to understand and process human language (like Siri or Alexa). Right now, most NLP systems tend to emphasize the “content”, ie, what is being said, but this research aims to make these systems aware of different cultures and find ways for people from various parts of the world to understand each other better, despite cultural differences.
Shifting the way we think about NLP could lead to more intelligent language systems that don’t just translate words but understand the deeper meaning behind them. For example, humor, politeness, and respect can be very different in different cultures, and an NLP system with cultural awareness could handle this much more effectively. This could improve communication in global businesses, diplomacy, and even everyday social media interactions.
Jiang’s project is motivated by the idea that if we start analyzing language technologies from a cultural point of view, it will help the people who create these models (like engineers and developers) prove that their systems are good at handling cultural differences.
If developers can prove that their models are “geo-culturally competent,” it could lead to more trust in these technologies. This means users from around the world might feel more comfortable using tools like translation apps or virtual assistants because they know the technology will consider their cultural context. It could also make these technologies more marketable and useful globally.
Future impact on downstream applications
Once language technologies become better at understanding different cultures, it will improve many other applications, or “downstream” systems, that depend on language technology. This includes things like translating languages (machine translation), answering questions (question-answering systems like chatbots), and finding information (search engines like Google).
This could lead to major improvements in a wide range of technologies that people use every day. For example, Google Translate might become more accurate, search engines could provide more relevant results based on cultural understanding, and customer service bots could give better answers by considering the user’s cultural background. This would make technology more accessible and effective for people worldwide.
Through this research, Jiang seeks to make language technology not just good at understanding words but good at understanding people from different places and cultures. The result could be smarter, more reliable tools that help us communicate better globally, improving everything from translations to online searches.
Executive Associate Dean of Luddy Indianapolis Davide Bolchini said, “One of the major limitations of current AI language systems stems from the fact they lack common ground, that shared experiential understanding so fundamental for effective communication, especially across cultures. Ming’s research is very timely, as it explores and empirically evaluates how to computationally embed important cross-cultural nuances in language models. The applications of this fundamental research for the design of future AI systems can be game changing.”
Media Contact
Joanne Lovrinic
jebehele@iu.edu
317-278-9208