Analysis and Retrieval of Printed and Electronic Documents for Recycle of Information

信息再利用是指将网页和文档中包含的复合信息分解为素材，再生产有用信息的过程。在这项研究中，我们已经调查了从印刷和电子文档的信息回收，以及“重用”的信息作为回收的前一步。本文的主要研究成果如下：1.文档图像局部检索及其在问答系统中的应用：作为一种印刷文档的再利用方法，本文提出了一种基于关键字二维密度分布的文档图像局部检索方法。在多种英文和日文文档上的实验结果表明了该方法的有效性。在此基础上实现了问答系统的一些基本功能。问题回答是在文档上定位可能的答案以响应用自然语言编写的问题的过程，因此 关于我们 is a kind类of recycling回收information信息.在英文文档上的实验结果表明，该方法能够对大约一半的问题定位正确答案。2.印刷文档中电子数据的嵌入和恢复：作为印刷文档信息重用的另一种方法，我们实现了在印刷文档中嵌入文本信息的方法。4 KB的数据被成功地嵌入和恢复在B5页面上，同时允许20%的阅读错误。3.电子文档的段落检索及其在问答中的应用：我们还提出了一种基于关键字密度分布的电子文档部分检索方法。它被应用到问答，以及和证明，该方法是能够定位正确的答案在顶部的一半的问题。4.从电子文档中的信息提取：作为一种方式，从电子文档中回收的信息，已提出了一种方法，列表中包含的信息文件。我们已经将此方法应用到7000篇日本新闻文章中提取个人档案信息，并显示了其有效性。此外，我们评估了从网页和图像中提取信息的基本方法。少

Recycle of information is the process of reproducing useful information based on materials decomposed from compound information included in Web pages and documents. In this research we have investigated the recycle of information from both printed and electronic documents as well as "reuse'" of information as a previous step of recycling. The results of this research are summarized as follows.1.Retrieval of parts of document images and its application to question answering : As a method of reuse of printed documents, we have proposed a method of retrieving parts of document images based on two-dimensional density distributions of keywords. Experimental results on various English and Japanese documents show the effectiveness of the proposed method. Some basic functions of question answering have also been implemented based on this retrieval method. Question answering is the process of locating possible answers on documents in response to questions written in natural languages, and thus … More is a kind of recycling information. Experimental results on English documents show that the method is capable of locating correct answers at the first rank for about half of questions.2.Embedding and recovery of electronic data on printed documents : As another approach of reusing information on printed documents we have implemented a method of embedding text information on printed documents when they are printed. 4KB of data are successfully embedded and recovered on B5 pages while permitting 20% of reading errors.3.Passage retrieval of electronic documents and its application to question answering : We have also proposed a method of retrieving parts of electronic documents based on the density distributions of keywords. It is applied to question answering as well and proven that the method is capable of locating correct answers at the top for half of questions.4.Information extraction from electronic documents : As a way of recycling information from electronic documents, a method of tabulating information included in documents has been proposed. We have applied this method to 7000 Japanese news articles for extracting personal profile information and shown its effectiveness. In addition, we evaluated basic methods for information extraction both from web pages and with images. Less