Line Eikvil 原著 （1999.7） 陈鸿标 译 (2003.3)
信息抽取（Information Extraction: IE）是把文本里包含的信息进行结构化处理，变成表格一样的组织形式。输入信息抽取系统的是原始文本，输出的是固定格式的信息点。信息点从各种各样的文档中被抽取出来，然后以统一的形式集成在一起。这就是信息抽取的主要任务………
第一章导论
第二章简要介绍信息抽取技术
第三章介绍网页分装器(wrapper)的开发
第四章介绍已经开发出来的网站信息抽取系统
第五章介绍信息抽取技术的应用范围以及首批已经进入商业运作的商用系统
Silviu Cucerzan ，David Yarowsky
一种独立于语言的命名实体识别方法。
王建会对自动摘要算法改进方面所做的研究工作
这是介绍信息抽取(Information Extraction)的一篇报告,包括MUC、Web抽取(Web Extraction)等。
本文介绍了FASTUS系统，这是一个从自然语言文本中进行信息抽取的系统， 抽取来的信息输入数据库或者用作其它用途。
MUC-7信息抽取任务的定义
本文简要介绍了MUL-7/MET-2的任务
本文介绍了IE(Information Extration)技术(18页)。
摘要：信息抽取研究旨在为人们提供更有力的信息获取工具，以应对信息爆炸带来的严重挑战。与信息检索不同，信息抽取直接从自然语言文本中抽取事实信息。过去十多年来，信息抽取逐步发展成为自然语言处理领域的一个重要分支，其独特的发展轨迹——通过系统化、大规模地定量评测推动研究向前发展，以及某些成功启示，如部分分析技术的有效性、快速NLP系统开发的必要性，都极大地推动了自然语言处理研究的发展，促进了NLP研究与应用的紧密结合。回顾信息抽取研究的历史，总结信息抽取研究的现状，将有助于这方面研究工作向前发展。
Jian Sun, Ming Zhou, Jianfeng Gao
(Accepted by special issue "Word Formation and Chinese Language processing" of the International Journal of Computational Linguistics and Chinese Language Processing) Abstract: We address in this paper the problem of Chinese named entity (NE) identification using class-based language models (LM). This study is concentrated on three kinds of NEs that are most commonly used, namely, personal name (PER), location name (LOC) and organization name (ORG). Our main contributions are three-fold: (1) In our research, Chinese word segmentation and NE identification have been integrated into a unified framework. It consists of several sub-models, each of which in turn may include other sub-models, leads to the overall model a hierarchical architecture. The class-based hierarchical LM not only effectively captures the features of named entities, but also handles the data sparseness problem. (2) Modeling for NE abbreviation is put forward. Our modeling-based method for NE abbreviation has significant advantages over rule-based ones. (3) In addition, we employ a two-level architecture for ORG model, so that the nested entities in organization names can be identified. When decoding, two-step strategy is adopted: identifying PER and LOC; and identifying ORG. The evaluation on a large, wide-coverage open-test data has empirically demonstrated that the class-based hierarchical language modeling, which integrates segmentation and NE identification, unifies the abbreviation modeling into one framework, has achieved competitive results of Chinese NE identification.
Scott Miller, Michael Crystal, Heidi Fox, Lance Ramshaw, Richard Schwartz,
这是BBN的MUC7参评系统SIFT系统的说明，我把它翻译了一下，基本意思很明了，但部分细节我可能还没有把握准确，如果有问题，请给我来信说明。
Jian Sun, Jianfeng Gao, Lei Zhang, Ming Zhou, and Changning Huang
This is the slides for the 19th International Conference on Computational Linguistics
Jian Sun, Jianfeng Gao, Lei Zhang, Ming Zhou, and Changning Huang
We consider here the problem of Chinese named entity (NE) identification using statistical language model(LM). In this research, word segmentation and NE identification have been integrated into a unified framework that consists of several class-based language models. We also adopt a hierarchical structure for one of the LMs so that the nested entities in organization names can be identified. The evaluation on a large test set shows consistent improvements. Our experiments further demonstrate the improvement after seamlessly integrating with linguistic heuristic information, cache-based model and NE abbreviation identification.
Elaine Marsh, Dennis Perzanowski
reviews MUC-7 and introduces the result and progress during this conference
Charles L. Wayne
Topic Detection and Tracking (TDT) refers to automatic techniques for locating topically related material in streams of data such as newswire and broadcast news. DARPA-sponsored research has made enormous progress during the past three years, and the tasks have been made progressively more difficult and realistic. Well-designed corpora and objective performance evaluations have enabled this success.
骆卫华的综述报告
Cymfony公司的IE系统，主要面向QA，包括已实现的NE系统和将要实现的CE和GE的原型。
作者Stephen Soderland为华盛顿州立大学计算机科学系教授。本文的被引用次数高达50多次。论文以信息抽取系统WHISK系统为例，描述了如何以机器学习的方式，利用小规模样本训练系统自动学习目标文本的抽取模式，从而实现自动化信息抽取的一种技术。这种技术不但极具启发意义而且很有实用价值。
本文出自北京大学计算机科学与技术系，综述了信息抽取的一些基本概念。
作者分析了Lixto抽取系统的架构，介绍了一种半自动化的Wrapper生成技术与自动化Web信息抽取技术。
作者将目前的几种Web数据抽取工具按照六种分类：Wrapper开发语言，可感知HTML的工具，基于NLP的工具，Wrapper归纳工具，基于建模的工具，基于语义的工具依次介绍了各Web数据抽取工具的工作原理与特点，并且比较了它们的一般输出质量。
本文前半段将会介绍有关本体的相关概念，后一部分将介绍本体在我们系统中的应用。为了配合信息提取，需要一些先验性的知识和统计信息。所以，我们构造了自己的针对BBS短文本的提取标注工具。为此构建了本体知识并以直观方式展现出来。结合本体推理机，我们的标注工具在标注的同时具备推理能力使得标注智能化，并能通过引用一个包装好的提取算法进行提取预览。
Ling Liu Calton Pu Wei Han
This paper describes the methodology and the software development of XWRAP an XMLenabled wrapper construction system for semiautomatic generation of wrapper programs By XMLenabled we mean that the metadata about information content that are implicit in the original web pages will be extracted and encoded explicitly as XML tags in the wrapped documents In addition the querybased content ltering process is performed against the XML documents The XWRAP wrapper generation framework has three distinct features First it explicitly separates tasks of building wrappers that are specic to a Web source from the tasks that are repetitive for any source and uses a component library to provide basic building blocks for wrapper programs Second it provides a userfriendly interface program to allow wrapper developers to generate their wrapper code with a few mouse clicks Third and most importantly we introduce and develop a twophase code generation framework The rstphase utilizes an interactive interface facility to encode the sourcespecic metadata knowledge identied by individual wrapper developers as declarative information extraction rules The second phase combines the information extrac tion rules generated at the rst phase with the XWRAP component library to construct an executable wrapper program for the given web source We report the initial experiments on performance of the XWRAP code generation system and the wrapper programs generated by XWRAP
Rayid Ghani, Rosie Jones, Dunja Mladeni′cy, Kamal Nigam, Se′an Slattery
Information extractors and classifiers operating on unrestricted, unstructured texts are an errorful source of large amounts of potentially useful information, especially when combined with a crawler which automatically augments the knowledge base from the world-wide web. At the same time, there is much structured information on the WorldWideWeb. Wrapping the web-sites which provide this kind of information provide us with a second source of information; possiblyless up-to-date, but reliable as facts. We give a case study of combining information from these two kinds of sources in the context of learning facts about companies. We provide results of association rules, propositional and relational learning, which demonstrate that data-mining can help us improve our extractors, and that using information from two kinds of sources improves the reliability of data-mined rules.
Alberto H. F. Laender Berthier A. RibeiroNeto
Altigran S. da Silva Juliana S. Teixeira
In the last few years, several works in the literature have addressed the problem of data extraction from Web pages. The importance of this problem derives from the fact that, once extracted, the data can be handled in a way similar to instances of a traditional database. The approaches proposed in the literature to address the problem of Web data extraction use techniques borrowed from areas such as natural language processing, languages and grammars, machine learning, information retrieval,...
Information is ubiquitous, and we are ooded with more than we can process. Somehow, we must rely less on visual processing, point-and-click navigation, and manual decision making and more on computer sifting and organization of information and auto-mated negotiation and decision making. A resolution of these problems requires software with semantic understanding|a grand challenge of our time. More particularly, we must solve problems of automated interoperability, integration, and knowledge sharing, and we must build information agents and process agents that we can trust to give us the information we want and need and to negotiate on our behalf in harmony with our beliefs and goals. This paper pro ers the use of information-extraction ontologies as an approach that may lead
to semantic understanding. Keywords: Semantics, information extraction, high-precision classi cation, schema mapping, data inte-gration, Semantic Web, agent communication, ontology, ontology generation.
The Chinese message structure is composed of several Chinese fragments which may be characters words or phrases. Every message structure carries certain information. We have developed a HowNet-based extractor that can extract Chinese message structures from a real text and serves as an interactive tool for building large-scale bank of Chinese message structures. The system utilizes the HowNet Knowledge System as its basic resources. It is an integrated system of rule-based analyzer, statistics based on the examples and the analogy given by HowNet-based concept similarity calculator.
Keyword: Chinese message structure; Knowledge Database Mark-up Language (KDML); parsing; chunk;
Recently many systems have been built that auto matically interact with Internet information resources However these resources are usually formatted for use by peopleeg the relevant content is embedded in HTML pages Wrappers are often used to extract a resources content but handcoding wrappers is tedious and errorprone We advocate wrapper induction a technique for automatically constructing wrappers We have identi ed several wrapper classes that can be learned quickly most sites require only a handful of examples consuming a few CPU seconds of processing yet which are useful for handling numerous Internet resources of surveyed sites can be handled by our techniques
In this paper, we present the W4F toolkit for the generation of wrappers for Web sources. W4F consists of a retrieval language to identify Web sources, a declarative extraction language (the HTML Extraction Language) to express robust extraction rules and a map-ping interface to export the extracted information into some userde ned data-structures. To assist the user and make the creation of wrappers rapid and easy, the toolkit o ers some wysiwyg support via some wizards. Together, they permit the fast and semi-automatic generation of ready-to-go wrappers provided as Java classes. W4F has been successfully used to generate wrappers for database systems and software agents, making the content of Web sources easily accessible to any kind of application.
(LP)2 is a covering algorithm for adaptive Information Extraction from text (IE). It induces symbolic rules that insert SGML tags into texts by learning from examples found in a userdefined tagged corpus. Training is performed in two steps: initially a set of tagging rules is learned; then additional rules are induced to correct mistakes and imprecision in tagging. Induction is performed by bottom-up generalization of examples in the training corpus. Shallow knowledge about Natural Language Processing (NLP) is used in the generalization process. The algorithm has a considerable success story. From a scientific point of view, experiments report excellent results with respect to the current state of the art on two publicly available corpora. From an application point of view, a successful industrial IE tool has been based on (LP)2. Real world applications have been developed and licenses have been released to external companies for building other applications. This paper presents (LP)2, experimental results and applications, and discusses the role of shallow NLP in rule induction.
