«Strengthening the Nation through Diversity, Innovation & Leadership in STEM San Antonio,Texas · October 3-6, 2013 Get Connected! Connect with the ...»
Our research sets out to explore ideal methods to persuade a person through computer-generated dialogue. In order for a dialogue system to construct a successful persuasion strategy, we must first determine which sequence of persuasion strategies work. Based on previous literature on the effectiveness of persuasion in computer-mediated communication, we developed a pattern of persuasive strategies that include emotional positive, emotional negative, rational positive, and rational negative strategies. To determine which sequence of strategies is most effective, we are currently conducting a laboratory experiment to determine how interactive dialogue systems can be most persuasive.
We present a Wizard of Oz study where participants interact with a human wizard instead of the dialogue system via a custom-developed web-based chat interface. The wizard will attempt to persuade participants to make a healthy choice and take a free, virtual tai chi class. We will measure the success of our system using different criteria: we will compare their answers to pre- and post-test questions gauging their interest in learning tai chi, we will check whether they print a flyer to receive more information on learning tai chi, and whether they later attend a tai chi class. We will also compare our results to persuasion in a face-to-face setting. The results of our experiment can help dialogue system designers implement ideal strategies for persuasive systems.
AUV MODEL AND SIMULATION USING JAVA REAL-TIME TO DETECT MINES AND OBSTACLESChaz Johnson, Samori Price.
California State University Dominguez Hills, Inglewood, CA.
Automated guided vehicles (AGV) are very intelligent robotic-enhanced vehicles, tethered to either land or sea, where pilots control their movement and actions. AGVs are equipped with manipulator arms for grabbing, moving, or placing items in their perspective place. High-definition video and still cameras on the vehicles record images of the land/sea life, precious metals, war mines, geology, and experiments. AGVs can go where humans cannot or should not tread for both investigative and offensive means. The robot is portable and is controlled using either RF (radio frequency) or a tethered line that can send visual or position data feedback to the user. The robot’s versatility is remarkable, allowing attachments for additional video monitoring equipment, chemical sensors, distracting devices, and weapons/ destructive devices. The vehicles carry a variety of sampling equipment and sensors for collecting information about the land, ocean, and seafloor. AGVs are very effective on land and in the sea; they provide great information during special military operations, search and rescue missions, and recreational use. They must be able to navigate on sand, water, or mud without any human interaction. We will be modeling and simulating a land/sea detection vehicle using a remote controller. By using a remote controller, we will implement a simulation model of a solar powered/remote controlled AGV. The AGV must be able to react to the situation immediately, so a human operated remote controller will help enhance that requirement.
RESDEC: A MOBILE RESISTOR DECODERMatias Akman, Charles Norona.
Miami Dade College, Miami, FL.
Resistor decoder (ResDec) is an application created on the mobile Android operating system. Its main function is to decode the value of a resistor in an automatic and user-friendly manner. Resistors are one of the most commonly used electronic components and their values are measured in ohms (Ω). These are typically represented by colorcoded bands in sets of 4 bands. ResDec’s objective is to decode a resistor’s value via an image captured using a mobile device. The intended users for this application are people with visual disabilities, such as color blindness, although it could be used for anyone’s benefit. This application will help these users to solve a recurrent problem that hasn’t had a solution so far. ResDec’s current implementation allows the user to take a picture of the desired component and, as a result, the image will be reduced only to the resistor, cutting out most of the unnecessary
165 UNDERGRADUATE POSTER ABSTRACTS
background. To achieve this, image processing algorithms and tools like the Haar-Training and a cascade classifier are used. This obtained region of interest (ROI) will be used to quantize the colors of the image. The objective of the quantization is to reduce the amount of colors present to just 13 possibilities, that are used on common resistors.
When completed, the image will be composed of only those colors, facilitating the localization of the bands. After this process is successfully implemented the next milestones to address are the localization of the bands and the translation to a numeric value.
SOFTWARE FOR MULTITOUCH TABLE TO ASSIST IN ARCHITECTURE DESIGN REVIEWSOrlando Nieves, Rafael Arce.
University of Puerto Rico, Rio Piedras Campus, San Juan, PR.
Multitouch technology is a new way to interact with computers. Since its invention, it has been used in touchpads, computer monitors, and cell phones. One of the many applications for multitouch tables is in education, where students can work together to solve problems, and professors can correct students’ assignments in an interactive manner. Our project consists of a software application for a multitouch table where architecture professors can electronically correct design projects while the students make recordings of the professors’ corrections. The software will emulate the effect of the translucent tracing paper used by architecture professors during design reviews by using layers similar to those used in graphics editing programs. In addition, it will provide various styles and colors of pens, lines, and geometric shapes to assist the professor in his/her corrections. The application works on top of a software stack that includes Community Core Vision to convert images captured by the table’s infrared camera into multi-touch gestures; TUIO, an open framework that defines a common protocol and API for tangible multi-touch surfaces; and PyMT, a Python library for developing multi-touch enabled media rich applications. Currently, we have implemented the functionalities of loading and manipulating multiple images simulating the architecture students’ designs and drawing on top of them, simulating the professor’s corrections. Future work includes implementing the functionality of multiple layers to simulate what is physically done using trace paper over the students’ blueprints, saving the professor’s marks, and being able to load back the work from a file.
TEMPT: A BLUEPRINT FOR A MORE INTELLIGENT DIALOG SYSTEMCharla Earls, Abdul Mohammed.
Northeastern Illinois University, Chicago, IL.
Artificial intelligence systems are able to imitate human behavior and are being used more and more as advances in capabilities of systems are discovered. This article will focus on the methodology and effectiveness of linguistic accommodation using artificial intelligence systems to persuade a person to take certain actions. Previous studies of face-to-face interaction have shown a person is more apt to take a suggestion from someone who speaks like them. We are testing the effectiveness of this accommodation in computer-human interactions. Our research uses an artificial intelligence dialog system to monitor utterances a user makes. These utterances are then filtered by a program and categorized into stylistic groups. Based on these groups, our system tailors its language to that of the user. The system tailors its language by filtering words from the communications on the subject’s Facebook and Twitter profiles. We receive these profiles voluntarily from subjects participating in our research. The language profiles allow us to categorize based on the aforementioned groups and tries to persuade the subject to adopt a healthier lifestyle. We hypothesize that a dialog system is more likely to persuade a human if linguistic accommodation is used.
We also hope to prove it is an effective tool of persuasion.
MOBILIZING A VIRTUAL WORLD THROUGH AN ANDROID DEVICERaymond Garcia Jr., Scott King.
Texas A&M-Corpus Christi, Corpus Christi, TX.
Virtual reality has been a fantasy for decades now. Although we have some forms of virtual reality, we have yet to move past the basic keyboard and mouse paradigm. There are systems available, like the CAVE System, which do allow the user to be mobile in a very simple virtual world but only in a small designated area. Our research is aimed at making a virtual world fully immersible with the help of a mobile device and virtual reality (VR) glasses. Gyroscopes on the VR glasses are used to control the orientation of the camera and mobile device sensors will be used to track physical position. Our initial application is a virtual representation of a building local to our campus on an android
166 UNDERGRADUATE POSTER ABSTRACTS
device allowing the user to explore the building by walking around using only what is shown on our VR glasses. This research can be used for training professionals in a hands-on manner to offer a valuable real experience. It could also be used in education. Students can be immersed into a virtual representation of what subject they are learning which Computer Science can greatly aid in their retention.
MACHINES EMULATING HUMANS: A PERSUASIVE DIALOGUE MANAGEMENT SYSTEMAbdul Rahman Mohammed, Timothy Nguyen, Lathaniel Mejias, Francisco Lacobelli.
Northeastern Illinois University, Chicago, IL.
Persuasion is omnipresent in human dialogue. For example, a friend trying to persuade another that he has found the best pub in town; or a doctor trying to persuade a patient to take her medicine. However, persuasion is hard for automatic dialogue systems. Our research aims to build a dialogue system that can emulate persuasion strategies present in human interaction. By examining the way in which humans use emotional and rational persuasion strategies in conjunction with positive and negative framing, we build a probabilistic model that mimics human interaction. In order to switch strategies, the system will be constantly monitoring features of the dialogue such as sentiment. In addition, there is evidence to suggest that mirroring the interlocutor’s linguistic styles results in a more pleasant conversation. Therefore, our dialogue system can produce utterances in a variety of linguistic styles to accommodate the users. We hypothesize that the dialogue system will be more successful at persuasion when it uses both the model of persuasion strategies as well as linguistic accommodation. We analyze the interactions using standard satisfaction metrics derived from log files and surveys. In this research, we quickly describe the study that leads to our model of persuasion strategies and we detail the components of the dialogue system. Finally, we present the results of a study in which the system tries to persuade users to make a healthy choice in their lives that they probably have not considered previously.
A VIRTUAL GEOCACHING BACKEND: DATA INTEGRATION WITH A CONTENT MANAGEMENT FRAMEWORKKarina Valtierra, Natalia Villanueva-Rosales.
University of Texas at El Paso, El Paso, TX.
The Virtual Geocaching project aims to promote the use of remote sensing satellite data, Visible Infrared Imaging Radiometer Suite (VIIRS), to create a science, technology, engineering, and mathematics (STEM) learning experience for students through a gaming environment. Virtual Geocaching mirrors the real-world treasure hunt by using satellite data to provide clues leading to a hidden virtual location. The focus of this project was on using a semantic-based approach with a content management system to facilitate data integration and management. As part of this project, we followed database-design best practices to redesign the database backend prototype of the game’s website to be embedded into Drupal, a widely used content management framework (CMF). During this project, we focused on combining our own database with the default Drupal database. By following database-design best practices, we developed a normalized database that integrates the legacy information from the first prototype, the information required to manage the new Virtual Geocaching version, and external data sources required for the geocaches. Even though there are still large amounts of data to be processed for the Virtual Geocaching project, the use of a content management framework, such as Drupal, has given an insight into how they can be beneficial for data integration and management practices. Future steps include the integration of external sources with heterogeneous formats through semantic web technologies.
DEVELOPING TOOLS AND TECHNIQUES FOR UNDERSTANDING THE BRAIN AND BODY AT WORKJenny Sanchez1, Kyung Su2, Kay Robbins2.
University of the Incarnate Word, San Antonio, TX, 2University of Texas at San Antonio, San Antonio, TX.
1 This research develops multilevel/ hierarchical models to account for subject differences in electroencephalograph (EEG) data. We evaluate the effectiveness of these models in several different task classification problems using R and MATLAB. Our models are based on previous recordings from Schalk, G, et al., and Goldberger, Al, EEG Motor Movement/ Imagery Dataset. Previous research has shown highly variable performance of classifiers across 109 subjects and 14 tasks. Multiple tasks include both body and eye movements. We expect our models to allow the capability of predicting which brain signals, from a certain category of individuals, are responsible for a particular task.
By modeling subject differences, we hope to improve the effectiveness of brain-computer interfaces (BCI) in practical settings supporting individuals with motor disabilities and overall physical limitations.
A SCALABLE DATA MANAGEMENT SYSTEM FOR ECOLOGICAL RESEARCHAlla Dove, Christine Laney, Natalia Villanueva-Rosales, Craig Tweedie.
University of Texas at El Paso, El Paso, TX.