We present a novel approach to the automatic acquisition of taxonomic relations. The main difference to earlier approaches is that we do not only consider one single source of evidence, i.e. a specific algorithm or approach, but examine the possibility of learning taxonomic relations by considering various and heterogeneous forms of evidence. In particular, we derive these different evidences by using well-known NLP techniques and resources and combine them via two simple strategies. Our approach shows very promising results compared to other results from the literature. The main aim of the work presented in this paper is (i) to gain insight into the behaviour of different approaches to learn taxonomic relations, (ii) to provide a first step towards combining these different approaches, and (iii) to establish a baseline for further research.

Without the proliferation of formal semantic annotations, the Semantic Web is certainly doomed to failure. In earlier work we presented a new paradigm to avoid this: the ’Self Annotating Web’, in which globally available knowledge is used to annotate resources such as web pages. In particular, we presented a concrete method instantiating this paradigm, called PANKOW (Pattern-based ANnotation through Knowledge On the Web). In PANKOW, a named entity to be annotated is put into several linguistic patterns that convey competing semantic meanings. The patterns that are matched most often on the Web indicate the meaning of the named entity — leading to automatic or semi-automatic annotation. In this paper we present C-PANKOW (Context-driven PANKOW), which alleviates several shortcomings of PANKOW. First, by downloading abstracts and processing them off-line, we avoid the generation of large number of linguistic patterns and correspondingly large number of Google queries. Second, by linguistically analyzing and normalizing the downloaded abstracts, we increase the coverage of our pattern matching mechanism and overcome several limitations of the earlier pattern generation process. Third, we use the annotation context in order to distinguish the significance of a pattern match for the given annotation task. Our experiments show that C-PANKOW inherits all the advantages of PANKOW (no training required etc.), but in addition it is far more efficient and effective.

The World-Wide Web consists of a huge number of unstructured documents, but it also contains structured data in the form of HTML tables. We extracted 14.1 billion HTML tables from Google’s general-purpose web crawl, and used statistical classification techniques to find the estimated 154M that contain high-quality relational data. Because each relational table has its own “schema ” of labeled and typed columns, each such table can be considered a small structured database. The resulting corpus of databases is larger than any other corpus we are aware of, by at least five orders of magnitude. We describe the WebTables system to explore two fundamental questions about this collection of databases. First, what are effective techniques for searching for structured data at search-engine scales? Second, what additional power

Enabling applications that exploit heterogeneous data in the Semantic Web will require us to harness a broad variety of semantics. Considering the role of semantics in a number of research areas in computer science, we organize semantics in three forms: implicit, formal and powerful, and explore their roles in enabling some of the key capabilities related to the Semantic Web. The central message of this paper is that building the Semantic Web purely on description logics will artificially limit it potential, and that we will need to both exploit well known techniques that support implicit semantics, and

Views over probabilistic data contain correlations between tuples, and the current approach is to capture these correlations using explicit lineage. In this paper we propose an alternative approach to materializing probabilistic views, by giving conditions under which a view can be represented by a block-independent disjoint (BID) table. Not all views can be represented as BID tables and so we propose a novel partial representation that can represent all views but may not define a unique probability distribution. We then give conditions on when a query’s value on a partial representation will be uniquely defined. We apply our theory to two applications: query processing using views and information exchange using views. In query processing on probabilistic data, we can ignore the lineage and use materialized views to more efficiently answer queries. By contrast, if the view has explicit lineage, the query evaluation must reprocess the lineage to compute the query resulting in dramatically slower execution. The second application is information exchange when we do not wish to disclose the entire lineage, which otherwise may result in shipping the entire database. The paper contains several theoretical results that completely solve the problem of deciding whether a conjunctive view can be represented as a BID and whether a query on a partial representation is uniquely determined. We validate our approach experimentally showing that representable views exist in real and synthetic workloads and show over three magnitudes of improvement in query processing versus a lineage based approach.

Saarbrücken/Germany suchanek aO mpii.mpg.de The World Wide Web provides a nearly endless source of knowledge, which is mostly given in natural language. A first step towards exploiting this data automatically could be to extract pairs of a given semantic relation from text documents – for example all pairs of a person and her birthdate. One strategy for this task is to find text patterns that express the semantic relation, to generalize these patterns, and to apply them to a corpus to find new pairs. In this paper, we show that this approach profits significantly when deep linguistic structures are used instead of surface text patterns. We demonstrate how linguistic structures can be represented for machine learning, and we provide a theoretical analysis of the pattern matching approach. We show the practical relevance of our approach by extensive experiments with our prototype system Leila.

The World-Wide Web consists of a huge number of unstructured hypertext documents, but it also contains structured data in the form of HTML tables. Many of these tables contain both relational-style data and a small “schema ” of labeled and typed columns, making each such table a small structured database. The WebTables project is an effort to extract and make use of the huge number of these structured tables on the Web. A clean collection of relational-style tables could be useful for improving web search, schema design, and many other applications. This paper describes the first stage of the WebTables project. First, we give an in-depth study of the Web’s HTML table corpus. For example, we extracted 14.1 billion HTML tables from a several-billion-page portion of Google’s generalpurpose web crawl, and estimate that 154 million of these tables contain high-quality relational-style data. We also describe the crawl’s distribution of table sizes and data types. Second, we describe a system for performing relation recovery. The Web mixes relational and non-relational tables indiscriminately (often on the same page), so there is no simple way to distinguish the 1.1 % of good relations from the remainder, nor to recover column label and type information. Our mix of hand-written detectors and statistical classifiers takes a raw Web crawl as input, and generates a collection of databases that is five orders of magnitude larger than any other collection we are aware of. Relation recovery achieves precision and recall that are comparable to other domain-independent information extraction systems. 1.

There is growing number of applications that require access to both structured and unstructured data. Such collections of data have been referred to as dataspaces, and Dataspace Support Platforms (DSSPs) were proposed to offer several services over dataspaces, including search and query, source discovery and categorization, indexing and some forms of recovery. One of the key services of a DSSP is to provide seamless querying on the structured and unstructured data. Querying each kind of data in isolation has been the main subject of study for the fields of databases and information retrieval. Recently the database community has studied the problem of answering keyword queries on structured data such as relational data or XML data. The only combination that has not been fully explored is answering structured queries on unstructured data. This paper explores an approach in which we carefully construct a keyword query from a given structured query, and submit the query to the underlying engine (e.g., a web-search engine) for querying unstructured data. We take the first step towards extracting keywords from structured queries even without domain knowledge and propose several directions we can explore to improve keyword extraction when domain knowledge exists. The experimental results show that our algorithm works fairly well for a large number of datasets from various domains. 1.

The Internet has changed the way people look for information. Users now expect the answers to their questions to be available through a simple web search. Web search engines are increasingly efficient at identifying the best sources for any given keyword query, and are often able to identify the answer within the sources. Unfortunately, many web sources are not trustworthy, because of erroneous, misleading, biased, or outdated information. In many cases, users are not satisfied with —or do not trust — the results from any single source and prefer checking several sources for corroborating evidence. In this paper, we propose methods to aggregate query results from different sources in order to save users the hassle of individually checking query-related web sites to corroborate answers. To return the best aggregated answers to the users, our techniques consider the number, importance, and similarity of the web sources reporting each answer, as well as the importance of the answer within the source. We present an experimental evaluation of our technique on real web queries, comparing the corroborated answers returned with real user clicks. 1.

Text is ubiquitous and, not surprisingly, many important applications rely on textual data for a variety of tasks. As a notable example, information extraction applications derive structured relations from unstructured text; as another example, focused crawlers explore the Web to locate pages about specific topics. Execution plans for text-centric tasks follow two general paradigms for processing a text database: either we can scan, or “crawl, ” the text database or, alternatively, we can exploit search engine indexes and retrieve the documents of interest via carefully crafted queries constructed in task-specific ways. The choice between crawl- and query-based execution plans can have a substantial impact on both execution time and output “completeness ” (e.g., in terms of recall). Nevertheless, this choice is typically ad hoc and based on heuristics or plain intuition. In this article, we present fundamental building blocks to make the choice of execution plans for text-centric tasks in an informed, cost-based way. Towards this goal, we show how to analyze query- and crawl-based plans in terms of both execution time and output completeness. We adapt results from random-graph theory and statistics to develop a rigorous cost model for the execution plans. Our cost model reflects the fact that the performance of the plans depends on fundamental task-specific properties of the underlying text databases. We identify these properties and present