International Workshop on Intelligent Information Access

Building blocks for semantic search engines: Ranking and compact indexing in entity-relation graphs

author: Soumen Chakrabarti, Indian Institute of Technology

Description

We see an evolutionary path to supporting semantic search over text facilitated by 1. extractors and annotators for ever-growing collections of entity and relation types and 2. search systems that exploit a smooth continuum between structured entities and relations on one hand and uninterpreted text on the other. The extractors and annotators will be imperfect and incomplete.

Categories

Top: Computer Science: Search Engines
Top: Computer Science: Semantic Web: Annotation

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Building Blocks for Semantic Search Engines: Ranking and Compact Indexing for Entity-Relation Graphs
0:26	Ranking and Indexing for Semantic Search
0:37	Working Notion of Semantic Search
2:38	Type-Annotated Corpus and Query e.g.
5:25	The Query Class We Address
7:25	Contribution 1: What is “NEAR”?
8:49	Type-Annotated Corpus and Query e.g. (a)
9:12	Contribution 1: What is “NEAR”? (a)
9:26	Contribution 2: Indexing Annotations
10:39	Part 1: Scoring and Ranking Nodes in Graphs
10:46	Two Flavors of Ranking Problems
13:10	Learning to Score Token Spans
16:03	Learning the Shape of the Decay Function
16:37	Learning to Score Token Spans (a)
17:07	Learning the Shape of the Decay Function (a)
19:05	Ranking Feature Vectors
21:31	Benign Loss Functions for Scoring
23:45	Learning Decay Function - Results
27:16	Searching Personal Information Networks pt 1
29:11	Searching Personal Information Networks pt 2
30:49	Ranking Nodes in ER Graphs
32:39	Edge Conductance
34:23	Constrained Design of Conductance
37:25	Breaking the p=Cp Recurrence
38:28	Setting up the Optimization
39:16	The Effect of a Limited Horizon
40:05	Appropriateness of Loss Approximation
41:46	Learning Rate and Robustness
43:22	Discovering Hidden Edge Weights
45:29	Part 1 Summary
46:16	Part 2: Indexing for Proximity Search
46:18	Part 2: Workload-Driven Indexing
47:53	Pre-Generalize (and Post-Filter)
49:20	(Pre-Generalize and) Post-Filter
49:45	Estimates Needed by Optimizer
50:50	Index Space Estimate Given R
51:35	Processing Time Bloat for One Query
52:37	Query Time Bloat - Results
52:58	Expected Bloat Over Many Queries
53:51	Smoothing Low-Probability Atypes
54:34	The R Selection Algorithm
56:20	Smoothing Low-Probability Atypes (a)
56:26	The R Selection Algorithm (a)
56:34	Optimized Space-Time Tradeoff
57:26	Optimized Index Sizes
58:03	Part 2 Summary
58:24	The Big Picture
59:39	Part 2 Summary (a)
60:10	Conclusion

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.