Information Retrieval Technology: 6th Asia Information Retrieval Societies Conference, AIRS 2010, Taipei, Taiwan, December 1-3, 2010, Proceedings (Lecture Notes in Computer Science, 6458)

The Asia Information Retrieval Societies Conference (AIRS) 2010 was the sixth conference in the AIRS series, aiming to bring together international researchers and developers to exchange new ideas and the latest results in information - trieval. The scope of the conference encompassed the theory and practice of all aspects of information retrieval in text, audio, image, video, and multimedia data. AIRS 2010 continued the conference series that grew from the Information Retrieval with Asian Languages (IRAL) workshop series, started in 1996. It has become a mature venue for information retrieval work, ?nding support from the ACM Special Interest Group on Information Retrieval (SIGIR); the Association for Computational Linguistics and Chinese Language Processing (ACLCLP); ROCLING; and the Information Processing Society of Japan, Special Interest GrouponInformationFundamentals andAccess Technologies(IPSJSIG-IFAT). This year saw a sharp rise in the number of submissions over the previous year. A total of 120 papers were submitted, representing work by academics and practitioners not only from Asia, but also from Australia, Europe, North America, etc. The high quality of the work made it di?cult for the dedicated programcommitteetodecidewhichpaperstofeatureattheconference.Through adouble-blindreviewingprocess,26submissions(21%)wereacceptedasfulloral papers and 31 (25%) were accepted as short posters. The success of this conferencewas only possible with the support of allof the authorswho submitted papers for review, the programcommittee members who constructively assessedthe submissions, and the registered conference delegates. We thank them for their support of this conference, and for their long-term support of this Asian-centric venue for IR research and development Cover......Page 1 Lecture Notes in Computer Science 6458......Page 2 Information RetrievalTechnology......Page 3 ISBN-13 9783642171864......Page 4 Preface......Page 6 Organization......Page 8 Table of Contents......Page 14 Introduction......Page 20 Related Work......Page 21 General Model......Page 22 BM25 Kernel......Page 23 LMIR Kernel......Page 24 KL Kernel......Page 25 Relation to Conventional Models......Page 26 Efficient Implementation......Page 27 Experiments on the Web Search Dataset......Page 28 Experiments on the OHSUMED and the AP Datasets......Page 29 Conclusion......Page 30 References......Page 31 Introduction......Page 32 Seed Dataset (Entry-Point)......Page 35 Video and User Properties......Page 36 Linguistic Analysis of User Comments......Page 37 Social Network Analysis......Page 38 Discussion......Page 42 References......Page 43 Introduction......Page 44 Related Work......Page 45 Methodology......Page 46 Term Type Prediction......Page 47 Retrieval Model......Page 48 Data Collection......Page 49 Term Type Prediction Evaluation......Page 50 Retrieval Model Evaluation......Page 52 An Example......Page 53 References......Page 54 Introduction......Page 56 The Generative Framework......Page 58 Inference......Page 59 Computational Efficiency......Page 60 Query Refinement Using the Secondary Representation......Page 61 Retrieving Query-Relevant Facets......Page 63 Related Work......Page 65 References......Page 66 Introduction......Page 68 Background and Related Work......Page 69 Our Novel Similarity Measure......Page 70 The Clustering Criterion Functions......Page 71 Optimization Algorithm......Page 73 Experimental Setup and Evaluation Metrics......Page 74 Results......Page 75 Conclusions and Future Work......Page 78 References......Page 79 Introduction......Page 80 Train the Initial Naïve Bayes Classifier......Page 81 Ratio-Adjusted EM Steps......Page 83 Dataset......Page 84 Overall Performance......Page 85 Study on the Effectiveness of R-Step and Sensitivity of......Page 86 Study on Sensitivity of $\hat\gamma$......Page 87 Related Work......Page 88 References......Page 90 Introduction......Page 92 Related Work......Page 93 Precision as Effective Time Ratio......Page 94 Effective Time Ratio for Search Engines with Snippets......Page 95 Theoretical Analysis......Page 96 Experiment......Page 97 Data Collecting......Page 98 Metrics......Page 99 Basic Results......Page 100 Conclusion......Page 101 References......Page 102 Introduction......Page 104 Related Work......Page 105 Data......Page 106 Query Level Analysis......Page 107 Session Level Analysis......Page 108 Query and Non-click Behavior......Page 109 Non-click and Post-query Actions in Session......Page 111 Non-click and Users’ Click Preference......Page 113 Conclusions and Future Work......Page 114 References......Page 115 Introduction......Page 116 Retrieval Experimentation......Page 117 Score Estimation......Page 118 Experimental Investigation......Page 121 Conclusion......Page 126 References......Page 127 Introduction......Page 129 Meta-search Engine Construction......Page 130 Multi-search Architecture......Page 131 Details of the Proposed Methods in Multi-Search......Page 132 Query Translation and Returned Search Results Merging......Page 133 Experimental Results......Page 135 Evaluation of the Study......Page 136 References......Page 138 Introduction......Page 140 Overview......Page 141 Two-Layer Graph Model......Page 142 Ranking Homogeneous Objects......Page 143 Co-ranking Heterogeneous Objects......Page 145 Dataset and Evaluation Metrics......Page 146 Experimental Results......Page 147 References......Page 148 Introduction......Page 150 Related Work......Page 151 Generation of Transformation Rules......Page 152 Automatically Accepting Evidences......Page 153 Temporal Evaluation......Page 155 Baseline......Page 157 Conclusion and Future Work......Page 158 References......Page 159 Introduction......Page 160 Related Research......Page 161 The Approach......Page 162 The Proposed Architecture......Page 163 Document Annotation......Page 165 Semantic Search and Processing......Page 167 Conclusion......Page 168 References......Page 169 Introduction......Page 170 Rocchio's Relevance Feedback Method......Page 171 The DFR Probabilistic Framework......Page 172 Quality-Biased PRF......Page 173 Test Collections and Evaluation......Page 174 Performance of Basic Retrieval Models......Page 175 Comparison of the PRF Methods......Page 176 Conclusions and Future Work......Page 178 References......Page 179 Introduction......Page 181 Subtopic Aware Paradigm for Diversity......Page 183 Integration Approach......Page 184 Empirical Study......Page 186 Experimental Results......Page 188 References......Page 190 Introduction......Page 192 Related Work......Page 193 Experiment......Page 194 Methods of Analysis......Page 195 Overview of Results on Search Units......Page 196 Results of Actions in Each Search Unit......Page 197 Results of Eye Gaze Points for SERP......Page 198 Results of View Rank and Click Rank......Page 199 Discussion and Conclusion......Page 200 References......Page 201 Introduction......Page 202 Background and Related Work......Page 203 Retrieval Model......Page 204 Experimental Setup......Page 206 Results and Discussion......Page 207 References......Page 210 Introduction......Page 212 Background and Related Research......Page 213 The Approach......Page 214 Extraction of Image Description, URL and Surrounding Text......Page 215 Syntactic Analysis......Page 216 Evaluation......Page 218 Discussion and Conclusion......Page 220 References......Page 221 Introduction......Page 222 Cost-Sensitive Listwise Approach......Page 223 Conditions of Order Preservation for Cost-Sensitive Listwise Approach......Page 224 Generalization for Order Preserved Cost-Sensitive Listwise Approach......Page 225 A Case: Order Preserved Cost-Sensitive ListMLE Approach......Page 226 Experiment Results......Page 227 Conclusion......Page 228 References......Page 229 Introduction......Page 230 Related Work......Page 231 Motivation......Page 232 Query Expansion Methods......Page 233 Maximum Relevance and Minimum Redundancy Criterion for Query Expansion (mRMR-QE)......Page 234 Collections......Page 235 mRMR-QE Evaluation......Page 236 Conclusions......Page 237 References......Page 238 Introduction......Page 240 Frameworks of Conditional Markov Models......Page 241 Hybrid Deterministic and Nondeterministic Inference Algorithm......Page 242 Automatic Chunk Relation Construction......Page 243 Speed-Up Local Classifiers......Page 244 Experiments......Page 245 Overall Results......Page 246 References......Page 248 Introduction......Page 250 Related Work......Page 251 Problem Formulation......Page 252 FolkRank......Page 253 FolkDiffusion......Page 254 Comparing with Other Methods......Page 256 Conclusion and Future Work......Page 258 References......Page 259 Introduction......Page 260 Feature Extraction......Page 261 Problem Formulation......Page 262 Entropy-and-Relevance-Based Summarization......Page 263 Regression-Based Summarization......Page 264 Experimental Setting......Page 265 Performance Evaluation......Page 266 Related Work......Page 267 Conclusion......Page 268 References......Page 269 Introduction......Page 270 Decision Tree Induction......Page 271 Support Vector Machines......Page 272 Spammer Behavioral Patterns......Page 273 Email Corpus......Page 275 Evaluation Metrics......Page 276 Experimental Results and Discussion......Page 277 Conclusion and Future Work......Page 278 References......Page 279 Related Work......Page 280 Our Approach......Page 281 Tagging with CRF model......Page 283 Experimental Strategy and Results......Page 284 Conclusions......Page 287 References......Page 288 Introduction......Page 289 Related Work......Page 290 Preliminaries......Page 291 Evaluation of the Semantic Relation Classification Performance......Page 293 Robustness with Respect to the Inter-view Correlation Measure......Page 296 References......Page 297 Introduction......Page 299 Related Research......Page 300 N-Gram Conflation and Co-occurrence Analysis for Language-Independent and Corpus-Based Stemming......Page 302 Experiments......Page 304 Conclusion and Future Work......Page 306 References......Page 307 Introduction......Page 309 The Stemming Inversion Problem......Page 310 Procedure......Page 313 Results......Page 314 References......Page 317 Introduction......Page 319 Pattern Matching......Page 320 Supervised IE......Page 321 More Queries and Fewer Answers......Page 322 Statistical Re-ranking......Page 323 Data and Scoring Metric......Page 324 System/Human Comparison......Page 325 Impact of Statistical Re-ranking......Page 326 Conclusion......Page 327 References......Page 328 Introduction......Page 329 Related Work......Page 330 Method......Page 331 Relation Extraction between Concepts by Web Search......Page 332 Experimental Environment......Page 334 Experimental Results......Page 335 Extracted Relations......Page 336 Conclusion......Page 337 References......Page 338 Introduction......Page 339 Problem Definition......Page 341 Score Function for Opinion-Oriented Chinese Sentence Compression......Page 343 Experiment Setup......Page 344 Experiment Results......Page 345 Conclusion and Future Work......Page 346 References......Page 347 Introduction......Page 349 Related Work......Page 350 Definition of Conditional Random Field......Page 351 Relations in Vietnamese......Page 352 Relation Extraction......Page 354 Experiments and Discussion......Page 355 Conclusions......Page 357 References......Page 358 Introduction......Page 359 Sparse L$_2$-Regularized SVMs Optimization......Page 360 Speed-Up Local Classifiers......Page 362 Search the Optimal Feature Set......Page 363 Results......Page 364 References......Page 367 Introduction......Page 369 Related Work......Page 370 The Compared Chinese Question Classifiers......Page 371 Test-Assisted Rule-Based Question Classifier......Page 372 The SVM-Based Question Classifier......Page 373 The Datasets......Page 374 Discussion......Page 375 References......Page 377 Introduction......Page 379 Related Work......Page 380 Introduction of the Model and Q-Function......Page 381 The EM Algorithm......Page 382 Experimental Setting......Page 384 Results......Page 385 References......Page 387 Introduction......Page 389 Related Work......Page 390 The KG-DRank Algorithm......Page 391 Baseline......Page 393 Comparison to Baseline......Page 394 Conclusion......Page 396 References......Page 397 Introduction......Page 398 An Interactive CLIR Interface......Page 399 Experiment to Collect Human Assessments......Page 400 Semantic Class and Relevance......Page 402 Familiarity and Relevance......Page 404 Conclusive Remarks......Page 406 References......Page 407 Introduction......Page 408 Methodology......Page 410 Evaluation Protocols......Page 411 Our Approach for Personalizing Mobile Search Using a Spatio-Temporal User Profile......Page 412 Evaluation Framework Application......Page 414 Measuring Results Consistency over the Two Evaluation Protocols......Page 415 Conclusion......Page 416 References......Page 417 Introduction......Page 418 Data Set......Page 419 Rule-Based Classifier......Page 420 Machine Learning Classifier......Page 421 Research Questions and Methodology......Page 423 Combination of Query Intents......Page 424 Patterns of Query Intents......Page 425 Query Intents and Query Re-formulation......Page 426 References......Page 427 Introduction......Page 429 Related Work......Page 430 Query Recommendation with Good User Experience......Page 431 User Behavior Features......Page 432 Training Data Annotation......Page 434 Evaluation of Recommendation’s Relevance and Search Performance......Page 435 Popularity of Recommended Queries......Page 436 Conclusion and Future Work......Page 437 References......Page 438 Introduction......Page 439 The Model......Page 441 Unsupervised Learning and the Rules......Page 442 Interpolated Kneser-Ney Smoothing......Page 443 Adapted Gibbs Sampling......Page 444 Experiment Setup......Page 446 Supervised Learning......Page 447 Unsupervised Learning......Page 448 Discussion and Conclusion......Page 449 References......Page 450 Introduction......Page 451 Cross-Document IE Annotation......Page 452 Motivation of Using IE for Summarization......Page 453 Relations/Events Can Push Up Relevant Sentences......Page 454 Event Coreference Can Remove Redundancy......Page 455 IE-Based Re-ranking and Redundancy Removal......Page 456 TAC Responsiveness Scores......Page 458 Discussion......Page 459 Conclusion......Page 460 References......Page 461 Introduction......Page 462 The Transition Stage......Page 463 The Transmission Stage......Page 465 Baseline Systems......Page 467 Our Approach......Page 468 Classification......Page 469 Conclusions......Page 470 References......Page 471 Introduction......Page 473 Related Work......Page 474 Domain-Topic Model......Page 475 Distributed Gibbs Sampling of DTM......Page 476 Ranking Terms Using DTM......Page 477 Keyword Extraction......Page 478 Content-Based Tag Recommendation......Page 481 References......Page 483 Introduction......Page 485 Graph Models for Polarity Lexicon Induction......Page 486 Analysis of the Two Kinds of Models......Page 487 Graph Models for Polarity Lexicon Induction......Page 488 Polarity Lexicon Induction with Morphological Features......Page 489 Integrating Graph Models and Morphological Features......Page 490 Experiments with Graph Models......Page 491 Experiments with Models of Morphological Features......Page 492 Experiments on Integration......Page 493 Discussion......Page 494 Conclusion and Future Work......Page 495 References......Page 496 Introduction......Page 497 Supplementary Data Assisted Ranking......Page 499 Supplementary Learning to Rank......Page 500 Boosting-Based Algorithms......Page 501 RankBoost-Heter......Page 502 Supplementary Ranking on Homogeneous Data......Page 504 Discussions......Page 506 References......Page 508 Introduction......Page 509 Temporal TextTiling Model......Page 511 Temporal TextTiling......Page 512 Context Similarity......Page 513 Named Entity Influence......Page 514 Temporal Proximity......Page 515 Evaluation Metrics......Page 516 Performance and Discussion......Page 517 Conclusion......Page 519 References......Page 520 Introduction......Page 521 Machine Learning and Sequence Labeling Tasks......Page 523 Impact of Varying the Model Size......Page 525 Experimental Results......Page 526 Improving Data Caching......Page 529 References......Page 532 Introduction......Page 533 Risk Minimization Framework......Page 535 Computing the Optimal Scores......Page 536 Considering True Relevance Feedback......Page 539 Experimental Setup......Page 540 The Effectiveness of the Iterative Optimization Algorithm......Page 541 Effect of Considering Relevance Feedback Information......Page 542 Related Work......Page 543 Conclusions......Page 544 References......Page 545 Introduction......Page 546 Smoothness of QL as the Document Weight......Page 548 Improving Lower Weights......Page 550 Evaluation Configuration......Page 552 Evaluation on Weight Allocation Methods......Page 554 Discussion......Page 555 Conclusions and Future Work......Page 556 References......Page 557 Introduction......Page 558 Related Work......Page 560 Variable Dependency Model......Page 561 Parameter Estimation......Page 563 Test Collections......Page 564 Experimental Results......Page 565 Analysis and Discussion......Page 568 References......Page 569 Introduction......Page 571 Proposed Method......Page 573 Clues Extraction......Page 574 Query Generation and Ranking......Page 575 Filtering......Page 576 Experiment Setup......Page 577 Experiment Results and Discussion......Page 578 Discussion......Page 580 References......Page 581 Introduction......Page 583 Query Expansion......Page 584 Query-Click Graph......Page 585 Term-Relationship Graph......Page 586 Naïve Method......Page 587 Pruning......Page 588 Query Expansion......Page 589 Design of Experiments......Page 590 Results......Page 591 References......Page 593 Introduction......Page 595 Related Work......Page 596 Bilingual Snippets Collection......Page 597 Candidates Extraction......Page 598 Frequency Distance Model......Page 599 Transliteration Model......Page 600 Snippets Collection Experiment......Page 601 OOV Translation Selection Experiments......Page 602 CLIR Experiments......Page 603 References......Page 605 Introduction......Page 607 Modification of the Score Based on the Boolean Query......Page 608 Differences between WWW Document Retrieval and QA Retrieval......Page 610 Query Construction Using Synonyms and Variation Lists......Page 611 Discussion of the Experimental Results......Page 614 Conclusion......Page 616 References......Page 617 Introduction......Page 618 Related Work......Page 619 Co_Tags Semantic Similarity......Page 620 T_SimRank Semantic Similarity......Page 621 T_PageRank Popularity......Page 622 Datasets and Evaluation Measure......Page 624 Experiment Result......Page 625 Conclusions......Page 627 References......Page 628 Introduction......Page 629 Image Re-ranking Based on Quality......Page 631 Image Features......Page 632 Datasets......Page 636 Ranking Performance......Page 637 Feature Analysis......Page 639 References......Page 641 Author Index......Page 644 The Asia Information Retrieval Societies Conference (AIRS) 2010 was the sixth conference in the AIRS series, aiming to bring together international researchers and developers to exchange new ideas and the latest results in information - trieval. The scope of the conference encompassed the theory and practice of all aspects of information retrieval in text, audio, image, video, and multimedia data. AIRS 2010 continued the conference series that grew from the Information Retrieval with Asian Languages (IRAL) workshop series, started in 1996. It has become a mature venue for information retrieval work,?nding support from the ACM Special Interest Group on Information Retrieval (SIGIR); the Association for Computational Linguistics and Chinese Language Processing (ACLCLP); ROCLING; and the Information Processing Society of Japan, Special Interest GrouponInformationFundamentals andAccess Technologies(IPSJSIG-IFAT). This year saw a sharp rise in the number of submissions over the previous year. A total of 120 papers were submitted, representing work by academics and practitioners not only from Asia, but also from Australia, Europe, North America, etc. The high quality of the work made it di?cult for the dedicated programcommitteetodecidewhichpaperstofeatureattheconference. Through adouble-blindreviewingprocess,26submissions(21%)wereacceptedasfulloral papers and 31 (25%) were accepted as short posters. The success of this conferencewas only possible with the support of allof the authorswho submitted papers for review, the programcommittee members who constructively assessedthe submissions, and the registered conference delegates. We thank them for their support of this conference, and for their long-term support of this Asian-centric venue for IR research and development Annotation This book constitutes the refereed proceedings of the 6th Asia Information Retrieval Symposium, AIRS 2010, held in Taipei, Taiwan, in December 2010. The 26 revised full papers and 31 revised poster papers presented were carefully reviewed and selected from 120 submissions. All current aspects of information retrieval - in theory and practice - are addressed; the papers are organized in topical sections on information retrieval models, machine learning for information retrieval, user studies and evaluation, natural language processing for information retrieval, Web and question answering, and multimedia