Andrew B. Williams

Towards applying text mining and natural language processing for biomedical ontology acquisition

The use of text mining and natural language processing can extend into the realm of knowledge acquisition and management for biomedical applications. In this paper, we describe how we implemented natural language processing and text mining techniques on the transcribed verbal descriptions from retinal experts of biomedical disease features. The feature-attribute pairs generated were then incorporated within a user interface for a collaborative ontology development tool. This tool, IDOCS, is being used in the biomedical domain to help retinal specialists reach a consensus on a common ontology for describing age-related macular degeneration (AMD). We compare the use of traditional text mining and natural language processing techniques with that of a retinal specialists analysis and discuss how we might integrate these techniques for future biomedical ontology and user interface development.

Sequential auctions for heterogeneous task allocation in multiagent routing domains

Many realistic problem domains are composed of heterogeneous tasks distributed in a physical environment. A team of mobile agents has to autonomously allocate these tasks, navigate to them and finally execute them. Recently auctions have been used for task allocation among homogeneous agents. Less studied is the case of allocation where both the tasks and the agents are heterogeneous in nature. In this paper, we investigate the market-based allocation of heterogeneous tasks to heterogeneous agents in domains where the distribution of the task heterogeneity is known a priori. We present a model of task heterogeneity, and define a metric that allows us to assess the fitness of a team for a particular task domain. We then present a sequential, round-based, auction setup for allocating heterogeneous tasks to heterogeneous teams and empirically investigate the performance of three different allocation strategies.

Introducing an experimental cognitive robotics curriculum at historically black colleges and universities

A successful collaboration between Spelman College and Carnegie Mellon University led to an NSF-funded Broadening Participation in Computing project to set up robotics education laboratories and introduce undergraduate instruction in cognitive robotics at three other Historically Black Colleges and Universities (HBCUs). We give a brief overview of cognitive robotics and the Tekkotsu software architecture, and describe our experiences teaching computer science students with no previous robotics exposure to program sophisticated mobile robots.

Effects of SMILE Emotional Model on Humanoid Robot User Interaction

Naturalistic conversation and emotions, while difficult to approximate in robots, facilitate interactions with non-expert users and serve to make robots more relatable and predictable. This paper describes the implementation and evaluation of two major improvements upon an existing interface, the SMILE app for the MU-L8 humanoid robot. The original version of the app is compared to a version in which popups and extraneous user touches are removed, and they are both compared to a third version in which the robots emotions decay with time. These versions are tested in terms of ease of use, user engagement, and naturalness of interaction. User feedback and observer ratings are collected for 15 participants, and their results are described. These improvements contribute advances in the field of smartphone humanoid robotics interfaces toward a more ideal emotional and conversational model.

Towards a social and mobile humanoid exercise coach

In the near future, humanoid robots may be available to act as personal health coaches that can socially interact and exercise with people to increase their physical activity and improve their nutritional habits. Although there has been work to demonstrate the long-term effects of using a robot to motivate and record exercise and nutrition data, we are developing a social and mobile humanoid health coach that will explain and perform the physical exercises along with the human in an effort to increase their physical activity. In this paper, we describe a pilot study to compare the effects on young adults of coaching delivered by a social and mobile humanoid robot health coach versus a human health coach. While data analysis bore out no significant statistical effect of coach type on daily activity level, the results demonstrated encouraging trends and suggest further research with a larger sample size.

A Case Study on the Cybersecurity of Social Robots

During the last few years, we have observed a rapid increase in the number of robots that are manufactured and marketed for our homes. These social robots promise to become not just a personal assistant, but a companion that knows its owner»s tastes and habits. As the advances in artificial intelligence move the dream of home robots closer, exploitation techniques that could specifically target these robots need to be thoroughly examined. In this paper, we present our observations from performing an initial vulnerability analysis to a commercial social robot. Our results indicate that both manufacturers and application developers need to take cybersecurity into account when considering the use of their robots.

Culturally Responsive Social Robotics Instruction for Middle School Girls

To attract more female students to study human-robot interaction (HRI) and robotics in college, new approaches for engaging middle and high school students need to be developed. This paper presents a unique approach to teaching middle school girls primary human-robot interaction concepts and humanoid robot programming using culturally responsive pedagogy. We describe what culturally responsive social robotics (CRSR) instruction is, how we incorporated CRSR and human-robot interaction instruction into a middle school girls social robotics program, and how we evaluated its effectiveness using mixed methods evaluation.

Smart Topic Detection for Robot Conversation

In order for humanoid robots to have believable conversations with humans, the robots will need a reliable method for detecting the topics shared in the interaction to formulate a relevant response. This paper presents a novel application of intelligent indexing and ontology analysis for use in conversational topic detection for human-robot interaction. We evaluate a method for training on a corpus of transcribed phone conversations and using a concept association matrix to determine the strongest common-sense linkages to words in the conversation. This model is placed within the conversation and emotion interface of our humanoid robot, MU-L8, and tested with users. Evaluation is performed both computationally, with the corpus, and perceptually with users, and the promising results are presented in this paper.

Experience Report: A Component-Based Data Management and Knowledge Discovery Framework for Aviation Studies

Organizations are beginning to apply data mining and knowledge discovery techniques to their corporate data sets, thereby enabling the identification of trends and the discovery of inductive knowledge. Since traditional transaction databases are not optimized for analytical processing, they must be transformed. This article proposes the use of modular components to decrease the overall amount of human processing and intervention necessary for the transformation process. Our approach configures components to extract data-sets using a set of “extraction hints.†Our framework incorporates decentralized, generic components that are reusable across domains and databases. Finally, we detail an implementation of our component-based framework for an aviation data set.