MADCAT Phase 2 Training Set
收藏资源简介：
<h3>Introduction</h3><br> <p>MADCAT (Multilingual Automatic Document Classification Analysis and Translation) Phase 2 Training Set contains all training data created by the Linguistic Data Consortium to support Phase 2 of the DARPA MADCAT Program. The data in this release consists of handwritten Arabic documents, scanned at high resolution and annotated for the physical coordinates of each line and token. Digital transcripts and English translations of each document are also provided, with the various content and annotation layers integrated in a single MADCAT XML output.</p><br> <p>The goal of the MADCAT program is to automatically convert foreign text images into English transcripts. MADCAT Phase 2 data was collected from Arabic source documents in three genres: newswire, weblog and newsgroup text. Arabic speaking scribes copied documents by hand, following specific instructions on writing style (fast, normal, careful), writing implement (pen, pencil) and paper (lined, unlined). Prior to assignment, source documents were processed to optimize their appearance for the handwriting task, which resulted in some original source documents being broken into multiple pages for handwriting. Each resulting handwritten page was assigned to up to five independent scribes, using different writing conditions.</p><br> <p>The handwritten, transcribed documents were checked for quality and completeness, then each page was scanned at a high resolution (600 dpi, greyscale) to create a digital version of the handwritten document. The scanned images were then annotated to indicate the physical coordinates of each line and token. Explicit reading order was also labeled, along with any errors produced by the scribes when copying the text.</p><br> <p>The final step was to produce a unified data format that takes multiple data streams and generates a single MADCAT XML output file with all required information. The resulting madcat.xml file has these distinct components: (1) a text layer that consists of the source text, tokenization and sentence segmentation, (2) an image layer that consist of bounding boxes, (3) a scribe demographic layer that consists of scribe ID and partition (train/test) and (4) a document metadata layer.</p><br> <p>LDC has also released:</p><br> <ul><br> <li>MADCAT Phase 1 Training Set (<a href="http://catalog.ldc.upenn.edu/LDC2012T15/">LDC2012T15</a>)</li><br> <li>MADCAT Phase 3 Training Set (<a href="http://catalog.ldc.upenn.edu/LDC2013T15/">LDC2013T15</a>)</li><br> <li>MADCAT Chinese Pilot Training Set (<a href="http://catalog.ldc.upenn.edu/LDC2014T13/">LDC2014T13</a>)</li><br> </ul><br> <h3>Data</h3><br> <p>This release includes 27,814 annotation files in both GEDI XML and MADCAT XML formats (gedi.xml and madcat.xml) along with their corresponding scanned image files in TIFF format. The annotation results in GEDI XML output files include ground truth annotations and source transcripts.</p><br> <p>Files are named as follows:</p><br> <ul><br> <li>galeID_page#_scribeID.{tif|gedi.xml|madcat.xml}</li><br> </ul><br> <h3>Samples</h3><br> <p>Please view the following samples:</p><br> <ul><br> <li><a href="desc/addenda/LDC2013T09.jpg" rel="nofollow">Raw Image</a></li><br> <li><a href="desc/addenda/LDC2013T09.gedi.xml.jpg" rel="nofollow">GEDI XML</a></li><br> <li><a href="desc/addenda/LDC2013T09.madcat.xml.jpg" rel="nofollow">MADCAT XML</a></li><br> </ul><br> <h3>Sponsorship</h3><br> <p>This work was supported in part by the Defense Advanced Research Projects Agency, MADCAT Program Grant No. HR0011-08-1-004 and GALE Program Grant No. HR0011-06-1-0003. The content of this publication does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred.</p><br> <h3>Updates</h3><br> <p>None at this time.</p></br> Portions © 2004, 2005, 2008 Agence France Presse, © 2007-2008 Al-Ahram, Al Hayat, Al Quds - Al Arabi, Asharq Al-Awsat, An Nahar, © 2004, 2007 Assabah, © 2004, 2005, 2007, 2008, 2010, 2013 Trustees of the University of Pennsylvania
<h3>简介</h3><br> <p>MADCAT（多语言自动文档分类分析与翻译，Multilingual Automatic Document Classification Analysis and Translation）第二阶段训练集包含由语言数据联盟（Linguistic Data Consortium，LDC）为支持美国国防高级研究计划局（Defense Advanced Research Projects Agency，DARPA）MADCAT项目第二阶段所创建的全部训练数据。本发布数据集包含高分辨率扫描的手写阿拉伯语文档，并对每一行与每个Token（词元）的物理坐标进行了标注。同时提供每份文档的数字转录文本与英文翻译，各类内容与标注层均整合至单一MADCAT XML输出文件中。</p><br> <p>MADCAT项目的目标是将外语文本图像自动转换为英文转录文本。MADCAT第二阶段数据来源于三种体裁的阿拉伯语源文档：新闻专线、博客与新闻组文本。阿拉伯语抄写员按照特定书写规范（快速、常规、精细）、书写工具（钢笔、铅笔）与纸张类型（带格、无格）手动抄写文档。在分配抄写任务前，源文档会经过处理以优化手写抄写任务的呈现效果，部分原始源文档因此被拆分为多页用于手写抄写。每份生成的手写页面会分配给至多5名独立抄写员，使用不同的书写条件。</p><br> <p>经质量与完整性检查后的手写转录文档，会以高分辨率（600dpi，灰度）扫描以生成手写文档的数字版本。随后对扫描图像进行标注，以指明每一行与每个词元的物理坐标。同时会标注明确的阅读顺序，以及抄写员在抄写文本时产生的各类错误。</p><br> <p>最后一步是生成统一的数据格式，整合多数据流并生成包含所有必要信息的单一MADCAT XML输出文件。生成的madcat.xml文件包含四个独立组成部分：(1) 文本层：包含源文本、词元化与分句结果；(2) 图像层：包含边界框信息；(3) 抄写员人口统计层：包含抄写员ID与数据集划分（训练集/测试集）；(4) 文档元数据层。</p><br> <p>语言数据联盟同时发布了：</p><br> <ul><br> <li>MADCAT第一阶段训练集（<a href="http://catalog.ldc.upenn.edu/LDC2012T15/">LDC2012T15</a>）</li><br> <li>MADCAT第三阶段训练集（<a href="http://catalog.ldc.upenn.edu/LDC2013T15/">LDC2013T15</a>）</li><br> <li>MADCAT中文试点训练集（<a href="http://catalog.ldc.upenn.edu/LDC2014T13/">LDC2014T13</a>）</li><br> </ul><br> <h3>数据说明</h3><br> <p>本发布数据集包含27,814份GEDI XML与MADCAT XML格式的标注文件（gedi.xml与madcat.xml），以及对应的TIFF格式扫描图像文件。GEDI XML输出文件中的标注结果包含基准真值标注与源转录文本。</p><br> <p>文件命名规则如下：</p><br> <ul><br> <li>galeID_page#_scribeID.{tif|gedi.xml|madcat.xml}</li><br> </ul><br> <h3>示例</h3><br> <p>请查看以下示例：</p><br> <ul><br> <li><a href="desc/addenda/LDC2013T09.jpg" rel="nofollow">原始图像</a></li><br> <li><a href="desc/addenda/LDC2013T09.gedi.xml.jpg" rel="nofollow">GEDI XML标注结果</a></li><br> <li><a href="desc/addenda/LDC2013T09.madcat.xml.jpg" rel="nofollow">MADCAT XML标注结果</a></li><br> </ul><br> <h3>资助说明</h3><br> <p>本研究部分受美国国防高级研究计划局MADCAT项目（拨款编号HR0011-08-1-004）与GALE项目（拨款编号HR0011-06-1-0003）资助。本出版物内容不一定反映政府的立场或政策，不应推断存在官方背书。</p><br> <h3>更新记录</h3><br> <p>暂无更新。</p><br> <p>部分内容 © 2004、2005、2008 法国新闻社（Agence France Presse），© 2007-2008 金字塔报（Al-Ahram）、生活报（Al Hayat）、耶路撒冷阿拉伯报（Al Quds - Al Arabi）、中东报（Asharq Al-Awsat）、每日新闻报（An Nahar），© 2004、2007 萨巴赫报（Assabah），© 2004、2005、2007、2008、2010、2013 宾夕法尼亚大学托管会（Trustees of the University of Pennsylvania）</p>
