Extracting Relations from Italian Wikipedia Using Self-Training

2010

Information-extraction (IE) systems seek to distill semantic relations from naturallanguage text, but most systems use supervised learning of relation-specific examples and are thus limited by the availability of training data. Open IE systems such as TextRunner, on the other hand, aim to handle the unbounded number of relations found on the Web. But how well can these open systems perform? This paper presents WOE, an open IE system which improves dramatically on TextRunner's precision and recall. The key to WOE's performance is a novel form of self-supervised learning for open extractors-using heuristic matches between Wikipedia infobox attribute values and corresponding sentences to construct training data. Like TextRunner, WOE's extractor eschews lexicalized features and handles an unbounded set of semantic relations. WOE can operate in two modes: when restricted to POS tag features, it runs as quickly as TextRunner, but when set to use dependency-parse features its precision and recall rise even higher.

Acl, 2006

Most information extraction systems either use hand written extraction patterns or use a machine learning algorithm that is trained on a manually annotated corpus. Both of these approaches require massive human effort and hence prevent information extraction from becoming more widely applicable. In this paper we present URES (Unsupervised Relation Extraction System), which extracts relations from the Web in a totally unsupervised way. It takes as input the descriptions of the target relations, which include the names of the predicates, the types of their attributes, and several seed instances of the relations. Then the system downloads from the Web a large collection of pages that are likely to contain instances of the target relations. From those pages, utilizing the known seed instances, the system learns the relation patterns, which are then used for extraction. We present several experiments in which we learn patterns and extract instances of a set of several common IE relations, comparing several pattern learning and filtering setups. We demonstrate that using simple noun phrase tagger is sufficient as a base for accurate patterns. However, having a named entity recognizer, which is able to recognize the types of the relation attributes significantly, enhances the extraction performance. We also compare our approach with KnowItAll's fixed generic patterns.

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 2020

Unsupervised relation extraction (URE) extracts relations between named entities from raw text without manually-labelled data and existing knowledge bases (KBs). URE methods can be categorised into generative and discriminative approaches, which rely either on hand-crafted features or surface form. However, we demonstrate that by using only named entities to induce relation types, we can outperform existing methods on two popular datasets. We conduct a comparison and evaluation of our findings with other URE techniques, to ascertain the important features in URE. We conclude that entity types provide a strong inductive bias for URE. 1

Acta Cybern., 2010

In this paper we present solutions for the crucial task of extracting structured information from massive free-text resources, such as Wikipedia, for the sake of semantic databases serving upcoming Semantic Web technologies. We demonstrate both a verb frame-based approach using deep natural language processing techniques with extraction patterns developed by human knowledge experts and machine learning methods using shallow linguistic processing. We also propose a method for learning verb frame-based extraction patterns automatically from labeled data. We show that labeled training data can be produced with only minimal human effort by utilizing existing semantic resources and the special characteristics of Wikipedia. Custom solutions for named entity recognition are also possible in this scenario. We present evaluation and comparison of the different approaches for several different relations.

Lecture Notes in Computer Science, 2010

Wikipedia is the largest encyclopedia on the web and has been widely used as a reliable source of information. Researchers have been extracting entities, relationships and attribute-value pairs from Wikipedia and using them in information retrieval tasks. In this paper we present a self-supervised approach for autonomously extract attributevalue pairs from Wikipedia articles. We apply our method to the Wikipedia automatic infobox generation problem and outperformed a method presented in the literature by 21.92% in precision, 26.86% in recall and 24.29% in F1.

With the advent of the Web and the explosion of available textual data, the field of domain ontology engineering has gained more and more importance. The last decade, several successful tools for automatically harvesting knowledge from web data have been developed, but the extraction of taxonomic and non taxonomic ontological relationships is still far from being fully solved. This paper describes a new approach which extracts ontological relations from Wikipedia. The non-taxonomic relations extraction process is performed by analyzing the titles which appear in each document of the studied corpus. This method is based on regular expressions which appear in titles and from which we can extract not only the two arguments of the relationships but also the labels which describe the relations. The resulting set of labels is used in order to retrieve new relations by analyzing the title hierarchy in each document. Other relations can be extracted from titles and subtitles containing only one term. An enrichment step is also applied by considering each term which appears as a relation argument of the extracted links in order to discover new concepts and new relations. The experiments have been performed on French Wikipedia articles related to the medical field. The precision and recall values are encouraging and seem to validate our approach.

This paper proposes the automatic acquisition of binary relational patterns (i.e. portions of text expressing a relation between two entities) from Wikipedia. There are a few advantages behind the use of Wikipedia: (i) relations are represented in the DBpedia ontology, which provides a repository of concepts to be used as semantic variables within patterns; (ii) most of the DBpedia relations appear in Wikipedia infoboxes, and are likely to be expressed in the corresponding pages, which increases the effectiveness of the extraction process; (iii) finally, as Wikipedia has pages in several languages, this opens the possibility for the acquisition of multilingual relational patterns. We show that a relatively simple methodology performs quite well on a set of selected DBpedia relations, for which a benchmark has been realized.

2007

The exponential growth and reliability of Wikipedia have made it a promising data source for intelligent systems. The first challenge of Wikipedia is to make the encyclopedia machine-processable. In this study, we address the problem of extracting relations among entities from Wikipedia's English articles, which in turn can serve for intelligent systems to satisfy users' information needs. Our proposed method first anchors the appearance of entities in Wikipedia articles using some heuristic rules that are supported by their encyclopedic style. Therefore, it uses neither the Named Entity Recognizer (NER) nor the Coreference Resolution tool, which are sources of errors for relation extraction. It then classifies the relationships among entity pairs using SVM with features extracted from the web structure and subtrees mined from the syntactic structure of text. The innovations behind our work are the following: a) our method makes use of Wikipedia characteristics for entity allocation and entity classification, which are essential for relation extraction; b) our algorithm extracts a core tree, which accurately reflects a relationship between a given entity pair, and subsequently identifies key features with respect to the relationship from the core tree. We demonstrate the effectiveness of our approach through evaluation of manually annotated data from actual Wikipedia articles.

Many errors produced by unsupervised and semi-supervised relation extraction (RE) systems occur because of wrong recogni- tion of entities that participate in the rela- tions. This is especially true for systems that do not use separate named-entity rec- ognition components, instead relying on general-purpose shallow parsing. Such sys- tems have greater applicability, because they are able to extract relations that contain attributes of unknown types. However, this generality comes with the cost in accuracy. In this paper we show how to use corpus statistics to validate and correct the arguments of extracted relation instances, improving the overall RE performance. We test the methods on SRES - a self-supervised Web relation extraction system. We also compare the performance of corpus-based methods to the performance of validation and correc- tion methods based on supervised NER components.

Proceedings of the 8th International Conference on Web Information Systems and Technologies

What is the relationship between terms? Document analysis tells us that "Crime" is close to "Victim" and not so close to "Banana". While for common terms like Sun and Light the nature of the relationship is clear, the measure becomes more fuzzy when dealing with more uncommonly used terms and concepts and partial information. Semantic relatedness is typically calculated from an encyclopedia like Wikipedia, but Wikipedia contains a lot of information that is not common knowledge. So, when a computer calculates that Belarus and Ukraine are closely related, what does it mean to me as a human? In this paper, we take a look at perceived relationship and qualify it in a human-readable way. The result is a search engine, designed to take two terms and explain how they relate to each other. We evaluate this through a user study which gauges how useful this extra information is to humans when making a judgment about relationships.