1. Relational Clustering for Entity Resolution Queries Indrajit Bhattacharya, Louis Licamele and Lise Getoor University of Maryland, College Park

2.  Discover the domain entities  Map each reference to an entity The Entity Resolution Problem Abdulla Ansari WeiWei WangChih Chen Wenyi WangLiyuan Li P1: “A mouse immunity model”, W.Wang, C.Chen, A.Ansari P2: “A better mouse immunity model”, W.Wang, A.Ansari P3: “Measuring protein-bound fluxetine”, L.Li, C.Chen, W.Wang P4: “Autoimmunity in biliary cirrhosis”, W.W.Wang, A.Ansari Chien-Te Chen

3. Query-time ER: Motivation  Most publicly available databases do not have resolved entities oPubMed, CiteSeer have many unresolved authors  Millions of queries everyday require resolved entities directly or indirectly o“I am looking for all papers by Stuart Russell”  How do we address this problem? 1.Leave the burden on the user to do the resolution 2.Ask owners to ‘clean’ their databases 3.Develop techniques for query-time resolution

4. Entity Resolution Queries  Disambiguation Query oAmong all papers with ‘W Wang’ as author, find those written by WeiWei Wang  Resolution Query oDo disambiguation oAlso retrieve papers by WeiWei Wang with a different author name, e.g. ‘W W Wang’ etc P1: “A mouse immunity model”, W.Wang, C.Chen, A.Ansari P2: “A better mouse immunity model”, W.Wang, A.Ansari P4: “Autoimmunity in biliary cirrhosis”, W.W.Wang, A.Ansari P1: “A mouse immunity model”, W.Wang, C.Chen, A.Ansari P2: “A better mouse immunity model”, W.Wang, A.Ansari P3: “Measuring protein-bound fluxetine”, L.Li, C.Chen, W.Wang

5. Query-time ER using Relations 1.Simple approach for resolving queries oUse attributes oQuick but not accurate 2.Use best techniques available oCollective resolution using relationships oHow can localize collective resolution?  Two-phase collective resolution for query oExtract minimal set of relevant records oCollective resolution on extracted records

6. Cut-based Evaluation of Relational Clustering Vertices embedded in attribute space Additional (hyper)edges represent relationships Good separation of attributes Many cluster-cluster relationships  C1-C3, C1-C4, C2-C4 Worse in terms of attributes Fewer cluster-cluster relationships  C1-C3, C2-C4 C1 C2 C4 C3 C1 C2 C4 C3

7. A Cut-based Objective Function weight for attributes weight for relations similarity of attributes 1 iff relational edge exists between c i and c j compatibility of c i and c j  Greedy clustering algorithm: merge cluster pair with max reduction in objective function Common cluster neighborhood Jaccard works better than intersection Similarity of attributes Jaro, Levenstein; TF-IDF

8. W Wang P4: W W Wang P1: W Wang P2: W Wang P3: W Wang P4: A Ansari P2: A Ansari P1: A Ansari P1: C Chen P3: C Chen P3: L Li P: A Ansari P: C Chen P: L Li Extracting Relevant Records Start with query name or record Alternate between 1.Name expansion: For any relevant record, include other records with that name 2.Hyper-edge Expansion: For any relevant record, include other related records Terminate at some depth k Name expansion Name expansion Hyper-edge expansion Query Level 0Level 1 Level 2

9. Adaptive Expansion for a Query  Too many records with unconstrained expansion oAdaptively select records based on ‘ambiguity’ o‘Chen’ is more ambiguous than ‘Ansari’  Adaptive Name Expansion oExpand the more ambiguous records  They need extra evidence  Adaptive Hyper-edge expansion oAdd fewer ambiguous records  They lead to imprecision

10. Unsupervised Estimation of Ambiguity  Probability of multiple entities sharing an attribute value  Estimate ambiguity of one single valued attribute (A1=a) using another (A2) oCount number of different values of A2 observed for records having A1=a oe.g. #different first initials for last-name ‘Smith’  Estimate improves with more independent attributes

11. Evaluation Datasets  arXiv High Energy Physics o29,555 publications, 58,515 refs to 9,200 authors oQueries: All ambiguous names (75 in total)  True authors per name: 2 to 11 (avg. is 2.4)  Elsevier BioBase o156,156 publications, 831,991 author refs oKeywords, topic classifications, language, country and affiliation of corresponding author, etc oQueries: 100 most frequent names  True authors per name: 1 to 100 (avg. is 32)

12. Growth Rate of Relevant Records and Query Processing Time Number of relevant references grows rapidly with expansion depth RC-ER is fast but not good enough for query-time resolution

13. Query-time ER Results Unconstrained expansion oCollective resolution more accurate oAccuracy improves beyond depth 1 A : pair-wise attributes similarity ; A+N: also neighbors’ attributes ; * : transitive closure AX-2 : adaptive expansion at depths 2 and beyond AX-1 : adaptive expansion even at depth 1 Adaptive expansion oMinimal loss in accuracy oDramatic reduction in query processing time

14. Conclusions  Query-centric entity resolution  Cut-based evaluation of relational clustering  Adaptive selection of relevant references for a query  Resolution at query-time with minimal loss in accuracy Future Directions  Spectral algorithm for relational clustering  Stronger coupling between extraction and resolution  Localized resolution for incoming records

15. References  "Query-Time Entity Resolution", Indrajit Bhattacharya, Louis Licamele and Lise Getoor, ACM SIGKDD, 2006  "A Latent Dirichlet Model for Unsupervised Entity Resolution", Indrajit Bhattacharya and Lise Getoor, SIAM Data Mining, 2006  "Entity Resolution in Graphs", Indrajit Bhattacharya and Lise Getoor, Chapter in Mining Graph Data, Lawrence B. Holder and Diane J. Cook, Editors, Wiley, 2006 (to appear).  "Relational Clustering for Multi-type Entity Resolution", Indrajit Bhattacharya and Lise Getoor, SIGKDD Workshop on Multi Relational Data Mining (MRDM), 2005