Our aim is to develop a culturally aware robot capable of communicating with people from different ethnic and cultural backgrounds and performing competently in a multi-lingual, cross-cultural context. Our test bed is a female robot receptionist, named Hala, deployed at the reception area in Carnegie Mellon University in Qatar. Hala answers questions in Arabic and English about people, locations of offices, classrooms and other rooms in the building. She also provides information about the weather, Education City, and her personal life. Our first model, Hala 1.0, was a bilingual robot extending an American model whose personality and utterances conform to the American culture. Three years of interaction logs have shown that 89% of Hala 1.0's interactions were in English. We conjecture that this is due to the robot's poor ability to equally portray both Arabic and American cultures and to its limited Arabic content. In order for us to investigate cultural factors that bear on communication significantly, we developed Hala 2.0 which is also a bilingual robot designed to be an Arab-American robot with more Arabic features in appearance, expression and interaction. The robot's personality is constructed taking into account the socio-cultural context in which its interactions will take place. To achieve bilingualism we had to create symmetry between Arabic and English linguistic content. Since the robot's utterances were developed primarily in English we resorted to translating them into Arabic and adapting them to the constraints of our socio-cultural context. Since Arabic is a highly inflected language, we adopted the plural case in formulating the robot's replies so as to avoid gender bias. To improve query coverage, we added word synonyms, including context-related synonyms (exp:هل تحبين عملك؟/ هل يعجبك عملك؟ ) and different formulations for the same question (exp: do you sleep? / do you go to sleep? and هل تنامين؟/ أتنامين؟). Furthermore, based on three years of recorded query logs, we expanded the range of topics that the robot is knowledgeable about by adding 3000 question/answer sentences to increase the robot's capacity for engaging users. All content and utterances were developed to align with the robot's designed personal traits.


Article metrics loading...

Loading full text...

Full text loading...

This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error