Heiko Stoermer

Entity Name System: The Back-Bone of an Open and Scalable Web of Data

Recognizing that information from different sources refers to the same (real world) entity is a crucial challenge in instance-level information integration, as it is a pre-requisite for combining the information about one entity from different sources. The required entity matching is time consuming and thus imposes a crucial limit for large-scale, dynamic information integration. An increased re-use of entity identifiers (or names) across different information collections such as RDF repositories, databases and document collections, eases this situation.In the ideal case, entity matching can be reduced to the trivial problem of spotting the same entity identifier in different information collections. In this paper we propose the use of an entity name system (ENS) - as it is currently under development in the EU-funded project OKKAM - for systematically supporting the re-use of entity identifiers. The main purpose of the ENS is to provide unique and uniform names for entities for the use in information collections, so that the same name is used for an entity, even when it is referenced in different contexts. Of course the creation of an ENS that can efficiently deal with entities on the Web scale raises scalability issues of its own. This paper focuses on the role of an ENS in contributing to the scalability of ad-hoc and on demand information integration tasks.

An entity name system (ENS) for the semantic web

In this paper, we argue that implementing the grand vision of the Semantic Web would greatly benefit from a service which can enable the reuse of globally unique URIs across semantic datasets produced in a fully decentralized and open environment. Such a service, which we call Entity Name System (ENS), stores pre-existing URIs and makes them available for reuse mainly - but not only - in Semantic Web contents and applications. The ENS will make the integration of semantic datasets much easier and faster, and will foster the development of a whole family of applications which will exploit the data level integration through global URIs for implementing smart semantic-based solutions.

A Conceptual Model for a Web-Scale Entity Name System

The problem of identity and reference is receiving increasing attention in the (semantic) web community and is emerging as one of the key features which distinguish traditional knowledge representation from knowledge representation on the web with respect to data interlinking and knowledge integration on a large scale. As part of this debate, the OKKAM project proposed the creation of an Entity Name System which provides rigid identifiers, named OKKAMids, for any type of concrete and particular entities, and links OKKAMids to existing identifiers which have been created elsewhere for the same entity. The introduction of these identifiers raises some practical and conceptual concerns. In this paper we address them by extending two proposed ontologies (IRE and IRW) to accomodate the notion of OKKAMid, describe their formal properties, illustrate why they may play an important role in the construction of the Semantic Web and discuss how they can be integrated with other approaches for mapping URIs onto each others.

Towards a general entity representation model

In this paper, we argue for an infrastructure responsible for assigning and managing unique identifiers for entities in the semantic web, and we propose a conceptual model for the storage and management of these entities.

Feature-based entity matching: the FBEM model, implementation, evaluation

Entity matching or resolution is at the heart of many integration tasks in modern information systems. As with any core functionality, good quality of results is vital to ensure that upper-level tasks perform as desired. In this paper we introduce the FBEM algorithm and illustrate its usefulness for general-purpose use cases. We analyze its result quality with a range of experiments on heterogeneous data sources, and show that the approach provides good results for entities of different types, such as persons, organizations or publications, while posing minimal requirements to input data formats and requiring no training.

From web data to entities and back

We present the Entity Name System (ENS), an enabling infrastructure, which can host descriptions of named entities and provide unique identifiers, on large-scale. In this way, it opens new perspectives to realize entity-oriented, rather than keyword-oriented, Web information systems. We describe the architecture and the functionality of the ENS, along with tools, which all contribute to realize the Web of entities.

A context-based architecture for RDF knowledge bases: approach, implementation and preliminary results

In this paper we present a context-based architecture and implementation for supporting the construction and management of contextualized RDF knowledge bases. The goal of this work is to take explicitly into account any possible contextual dependency of a collection of RDF models, without losing sight of performance and scalability issues. We are illustrating motivations, as well as theoretical background, implementation details and test-results of our latest works.

A novel approach for entity linkage

The problem of Data Linkage in the SemanticWeb can be divided in two lines of action: schema and ontology matching/mapping, which allows us to draw conclusions about sets of individuals through concept relations, and entity-level linkage, where more information can be reached from distributed sources because of the fact that the information is about the same entity. While the area of schema and ontology matching is traditionally much addressed, it appears that today the Semantic Web looks very much like a collection of “information islands” that are very poorly integrated with each other, especially on the individual level; and when some of these islands are linked, this is often the result of a lot of hard and time-consuming manual work. The general problem we are working on is to provide a structured approach of how to improve the situation of data linkage at the level of individuals in the Web of Data. As a specific contribution, in this article we describe a novel algorithm for entity linkage - called Name Feature Match - based on a recent empirical investigation about how humans describe individuals (or entities). We show in a first experimental evaluation that such an approach, which takes into account the cognitive point of view of entity representation by humans, can provide an improvement over other relevant approaches.

Searching for individual entities: A query analysis

Searching for information about individual entities such as persons, locations, events, is an important activity in Internet search today, and is in its core a very semantic-oriented task. Several ways for accessing such information exist, but for locating entity-specific information, search engines are the most commonly used approach. In this context, keyword queries are the primary means of retrieving information about a specific entity. We believe that an important first step of performing such a task is to understand what type of entity the user is looking for. We call this process Entity Type Disambiguation. In this paper we present a Naive Bayesian Model for entity type disambiguation that explores our assumption that an entity type can be inferred from the attributes a user specifies in a search query. The model has been applied to queries provided by a large sample of participants in an experiment performing an entity search task. The beneficial impact of this approach for the development of new search systems is discussed.

Entity Type Disambiguation in User Queries

Searching for information about individual entities such as persons, locations, events, is an important activity in Internet search today, and is in its core a very semantic-oriented task. Several ways for accessing such information exist, but for locating entity-specific information, search engines are the most commonly used approach. In this context, keyword queries are the primary means of retrieving information about a specific entity. We believe that an important first step of performing such a task is to understand what type of entity the user is looking for. We call this process Entity Type Disambiguation. In this paper, we present a Naive Bayesian Model for entity type disambiguation that explores our assumption that an entity type can be inferred from the attributes a user specifies in a search query. The model has been applied to queries provided by a large sample of participants in an experiment performing an entity search task. The beneficial impact of this approach for the development of new search systems is discussed.