FY2010 - Oak Ridge National Laboratory

More documents

Recommendations

Info

Director’s R&D Fund— Ultrascale Computing and Data Science “Computational Studies of Nanomaterials for Energy Storage Applications,” YESS, Saclay, France (2009). “Why Do We Need Large-scale Simulations in Nanoscience Anyway?” colloquium at LSU, Baton Rouge LA (2009). “Carbon nanostructures for energy storage and nanoelectronics applications,” colloquium at NCSU, Raleigh, NC (2009). “Capacitor Models for Various Regimes, Carbons, and Electrolytes,” the 2009 Advanced Automotive Battery and Capacitor Conference (AABC-09), Long Beach, CA (2009). “Theoretical Model of Nanoporous Carbon Supercapacitors,” the 64 th American Chemical Society Southwest Regional Meeting (SWRM), Little Rock, AR (2008). “A Universal Model of Nanoporous Carbon Supercapacitors,” the International Conference on the Theory and Applications of Computational Chemistry, Shanghai, China (2008). 05274 Inferring and Predicting the Social Dynamics of Groups via Psycho-Textual and Communications Flow Analysis Jack Schryver, Edmon Begoli, Yu Jiao, and Tracy Warren Project Description The goal of this project is to develop a new capability in social network analysis using electronic communications data that can be analyzed with text analysis techniques. We extend traditional social networking analysis by including deep-dive analysis of message contents to obtain inferences of group processes. We consider group formation, recruitment, coalition, threat, conflict, and schism. Important indicators such as in-group bias and out-group antipathy are rarely coded explicitly in text. Instead, they are embedded in connotative/affective meanings. Semantic analysis has generally focused on denotative meaning, creating a huge potential for knowledge discovery. Most related research is in the field of sentiment analysis–detection of positive/negative orientations toward a predetermined subject. Our work departs from previous work in three main directions. First, we perform a more fine-grained analysis, narrowing the focus from document to sentence level and from pure sentiment to 22 affective states. Second, we integrate common sense affective knowledge. Third, we link affect with entities identified in documents for deeper understanding of affective meaning. This capability will be an indispensible aid for improving our nation’s ability to protect itself from terrorism, and has the potential to reshape the way information about individual interactions is stored and analyzed. Mission Relevance This research extends ORNL's Knowledge Discovery mission objective. In addition, this research is directly aligned with the mission of the Intelligence Advanced Research Projects Activity (IARPA), the Intelligence Community (IC), and the Department of Homeland Security (DHS). These agencies need innovative and reliable tools to help them analyze the deep contents of massive quantities of electronic messaging data (e-mail, text chat, blogs, and transcribed talk). This research will help establish a capability that is essential for long-term human factors analysis and intelligence analysis (DHS), socialnetwork analysis (IC), and computational and psycholinguistics (IARPA). With internationally recognized experience and expertise in knowledge discovery and data mining, high-performance computing, human factors, social networking technology, and geospatial sciences, ORNL is strategically positioned to develop this capability. 78
Director’s R&D Fund— Ultrascale Computing and Data Science Results and Accomplishments Several technical advances resulted from this research; however, the overall accomplishment was the introduction of a new fine-grained computer-based method for attributing affective meaning to single documents. Our algorithm extends affective meaning from the basic extraction of affect terms to include identification of affective relationships between entities found in texts. This objective is achieved in three sequential steps. First, we developed a technique for building a large affect lexicon from initial seed word lists and common sense knowledge embedded in open source resources like WordNet and ConceptNet. The well-known PageRank algorithm was adapted to regulate the flow of affective meaning from seed lists to other terms in the English lexicon. Second, affect terms from the affect lexicon are identified in the context of a document. Third, affective relationships among entities are discovered in the text by regulating the spreading activation of affective meaning from affect terms to entities. For example, “John loves Mary” describes a “liking” relationship between John and Mary. Finally, we incorporated these algorithms into a software prototype called TEAMSTER. The Java prototype has a graphical user interface that illustrates in different panels the results of the various steps of the extraction of affective meaning in a document. The prototype also includes an automated search interface for multiple document retrieval. When a single document is loaded, it highlights affect terms in a scrolling window; affect intensities are plotted as a word series. A pruned graph of the document is displayed, which users can further prune by selecting affects and entities. Finally, the extracted affect-entity relationship diagram is presented. TEAMSTER integrates open source code to perform entity extraction, tokenization, word stemming, part-of-speech tagging, and other natural language programming tasks. The affect propagation algorithm was validated using a high-affect 50-document subset of the MPQA corpus. These documents were manually annotated for affects, entities and affective relationships. Eleven free parameters in the affect propagation algorithm were optimized using simulated annealing. The optimized affect propagation algorithm was found to perform 74% better than comparable random guess models. The results demonstrate that the affect propagation algorithm is a good start toward computer-mediated extraction of affective/connotative meaning from text, but further progress is still needed before it can approach the accuracy of human judgment. Additional studies were performed to identify influential social actors in a social network graph using sentiment analysis techniques. The PageRank algorithm was used to propagate social actor influence through the social network with limited success. Accomplishments in program development are ongoing. We have held discussions with DHS Human Factors Division, Heather McCallum-Bayliss, Incisive Analysis Program Manager at IARPA; Rebecca Goolsby, Office of Naval Research; and Alenka Brown-VanHoozer, DOD-CIO/ASD-NII. We have also provided two briefings to representatives of the intelligence community. As these discussions continue, our objective is to identify relevant BAAs scheduled for release in FY 2011, or prepare unsolicited proposals for submission. We are working with local program managers who have contacts in the intelligence community toward this objective. Information Shared Schryver, J. C., E. Begoli, A. C. Jose, and C. Griffin. 2010. Inferring Group Processes from Computer- Mediated Affective Text Analysis. ORNL/TM-2010/277, Oak Ridge National Laboratory, Oak Ridge, TN. 79
Page 1 and 2:
ORNL/PPA-2011/1 Laboratory Directed
Page 3:
ORNL/PPA-2011/1 Oak Ridge National
Page 6 and 7:
NEUTRON SCIENCES ..................
Page 8 and 9:
NATIONAL SECURITY SCIENCE AND TECHN
Page 10 and 11:
05887 Controlling the Catalytic Pro
Page 13 and 14:
Introduction INTRODUCTION The Labor
Page 15 and 16:
Introduction projects for next-gene
Page 17 and 18:
Introduction ― Develop new mode
Page 19 and 20:
Introduction To select the best and
Page 21 and 22:
Introduction Fig. 2. Distribution o
Page 23:
SUMMARIES OF PROJECTS SUPPORTED THR
Page 26 and 27:
Director’s R&D Fund— Science fo
Page 28 and 29:
Page 30 and 31:
Page 32 and 33:
Page 34 and 35:
Page 36 and 37:
Page 38 and 39:
Page 40 and 41: Director’s R&D Fund— Science fo
Page 53 and 54: Director’s R&D Fund— Neutron Sc
Page 65 and 66: 05306 Structure and Structure Evolu
Page 67 and 68: 05404 Asynchronous In Situ Neutron
Page 77 and 78: Director’s R&D Fund— Ultrascale
Page 89: Director’s R&D Fund— Ultrascale
Page 105 and 106: Director’s R&D Fund— Systems Bi
Page 117: Director’s R&D Fund— Systems Bi
Page 120 and 121: Director’s R&D Fund— Advanced E
Page 135 and 136: Director’s R&D Fund— Emerging S
Page 137 and 138: Director’s R&D Fund— Emerging S
Page 139: Director’s R&D Fund— Emerging S
Page 142 and 143:
Director’s R&D Fund— Understand
Page 144 and 145:
Page 146 and 147:
Page 148 and 149:
Page 150 and 151:
Page 153 and 154:
Director’s R&D Fund— National S
Page 155 and 156:
Page 157 and 158:
Page 159 and 160:
Page 161 and 162:
Page 163:
05573 Rapid Radiochemistry Applicat
Page 166 and 167:
Director’s R&D Fund— Energy Sto
Page 168 and 169:
Page 170 and 171:
Page 172 and 173:
Page 174 and 175:
Page 176 and 177:
Director’s R&D Fund— General Re
Page 178 and 179:
Director’s R&D Fund— General de
Page 180 and 181:
Director’s R&D Fund— General su
Page 182 and 183:
Director’s R&D Fund— General 05
Page 184 and 185:
Director’s R&D Fund— General 20
Page 187 and 188:
Seed Money Fund— Biosciences Divi
Page 189 and 190:
Page 191 and 192:
Page 193:
Page 196 and 197:
Seed Money Fund— Center for Nanop
Page 199 and 200:
Seed Money Fund— Chemical Science
Page 201 and 202:
Page 203 and 204:
Page 205:
Page 208 and 209:
Seed Money Fund— Computational Sc
Page 210 and 211:
Seed Money Fund— Computational Sc
Page 213 and 214:
Seed Money Fund— Computer Science
Page 215 and 216:
Seed Money Fund— Energy and Trans
Page 217 and 218:
Page 219:
Page 222 and 223:
Seed Money Fund— Environmental Sc
Page 224 and 225:
Seed Money Fund— Environmental Sc
Page 227 and 228:
Seed Money Fund— Fusion Energy Di
Page 229 and 230:
Seed Money Fund— Materials Scienc
Page 231 and 232:
Page 233 and 234:
Page 235 and 236:
Page 237 and 238:
Page 239 and 240:
Page 241 and 242:
Seed Money Fund— Measurement Scie
Page 243 and 244:
Page 245 and 246:
Page 247 and 248:
05858 Fabrication of Ultrathin Grap
Page 249 and 250:
Page 251:
Page 254 and 255:
Seed Money Fund— Global Nuclear S
Page 256 and 257:
Seed Money Fund— Neutron Scatteri
Page 259 and 260:
Seed Money Fund— Reactor and Nucl
Page 261 and 262:
Page 263 and 264:
Page 265:
Page 268 and 269:
Seed Money Fund— Physics Division
Page 270 and 271:
Seed Money Fund— Physics Division
Page 272 and 273:
Seed Money Fund— Research Acceler
Page 275 and 276:
Laboratory-Wide Fellowships— Wein
Page 277 and 278:
Page 279 and 280:
Page 281:
Page 284 and 285:
Laboratory-Wide Fellowships— Wign
Page 286 and 287:
Laboratory-Wide Fellowships— Wign
Page 288 and 289:
Index of Project Contributors Coope
Page 290 and 291:
Index of Project Contributors Mille
Page 292 and 293:
Index of Project Contributors Yang,
Page 294:
Index of Project Numbers 05501 ....
show all

FY2010 - Oak Ridge National Laboratory

Create successful ePaper yourself

Delete template?

Save as template?