Fisher Spanish - Transcripts

<h3>Introduction</h3> <p>Fisher Spanish - Transcripts was developed by LDC and contains full orthographic transcripts of the telephone speech in <a href="http://catalog.ldc.upenn.edu/LDC2010S01" rel="nofollow">Fisher Spanish - Speech (LDC2010S01)</a>. Transcripts cover roughly 163 hours of telephone speech from 136 native Caribbean Spanish and non-Caribbean Spanish speakers. </p><p>The Fisher telephone conversation collection protocol was created at LDC to address a critical need of developers trying to build robust automatic speech recognition (ASR) systems. Previous collection protocols, such as CALLFRIEND and Switchboard-II and the resulting corpora, have been adapted for ASR research but were in fact developed for language and speaker identification respectively. Although the CALLHOME protocol and corpora were developed to support ASR technology, they feature small numbers of speakers making telephone calls of relatively long duration with narrow vocabulary across the collection. CALLHOME conversations are challengingly natural and intimate. Under the Fisher protocol, a very large number of participants each make a few calls of short duration speaking to other participants, whom they typically do not know, about assigned topics. This maximizes inter-speaker variation and vocabulary breadth although it also increases formality. </p><p>Previous protocols such as CALLHOME, CALLFRIEND and Switchboard relied upon participant activity to drive the collection. Fisher is unique in being platform driven rather than participant driven. Participants who wish to initiate a call may do so; however the collection platform initiates the majority of calls. Participants need only answer their phones at the times they specified when registering for the study. </p><p>To encourage a broad range of vocabulary, Fisher participants are asked to speak on an assigned topic which is selected at random from a list, which changes every 24 hours and which is assigned to all subjects paired on that day. Some topics are inherited or refined from previous Switchboard studies while others were developed specifically for the Fisher protocol. </p><p>In collecting data for this corpus, attempts were made to provide a representative distribution of subjects across a variety of demographic categories including: gender, age, dialect region, and education level. </p><p>This corpus joins other Fisher corpora: Arabic CTS Levantine Fisher Training Data Set 3 (<a href="http://catalog.ldc.upenn.edu/LDC2005S07" rel="nofollow">LDC2005S07</a>, <a href="http://catalog.ldc.upenn.edu/LDC2005T03" rel="nofollow">LDC2005T03</a>), Fisher English Training Part 2 (<a href="http://catalog.ldc.upenn.edu/LDC2005S13" rel="nofollow">LDC2005S13</a>, <a href="http://catalog.ldc.upenn.edu/LDC2005T19" rel="nofollow">LDC2005T19</a>), Fisher English Training Speech Part 1 (<a href="http://catalog.ldc.upenn.edu/LDC2004S13" rel="nofollow">LDC2004S13</a>, <a href="http://catalog.ldc.upenn.edu/LDC2004T19" rel="nofollow">LDC2004T19</a>), and Fisher Levantine Arabic Conversational Telephone Speech (<a href="http://catalog.ldc.upenn.edu/LDC2007S02" rel="nofollow">LDC2007S02</a>, <a href="http://catalog.ldc.upenn.edu/LDC2007T04" rel="nofollow">LDC2007T04</a>) </p><h3>Data</h3> <p>The transcript files are in plain-text, tab-delimited format (tdf) with UTF-8 character encoding. They were created with the LDC-developed transcription tool <a href="http://www.ldc.upenn.edu/tools/XTrans/" rel="nofollow">"XTrans"</a>, which allowed for improved handling of multi-channel audio and overlapping speakers. XTrans is available from LDC. </p><p>Transcribers followed LDC's <a href="http://www.ldc.upenn.edu/Catalog/docs/LDC2010T04/trans_guide_nqrt_span.doc" rel="nofollow">Transcription Guidelines (NQTR)</a>, which are included with the documentation for this release. </p><p> The first line of each transcript file provides the column headings; the next two lines are "comments" that can be ignored (these are used by XTrans; they are distinguished from non-comment lines by having an initial semicolon ";"). Actual transcript data, with time stamps, channel number, transcript text and additional information, begins at line 4 of each transcript file. </p><p><a href="http://catalog.ldc.upenn.edu/LDC2010S01" rel="nofollow">Fisher Spanish - Speech (LDC2010S01)</a> provides the digital audio used as the basis for the transcriptions in this corpus, in the form of 2-channel mu-law sample data with 8000 samples per second (as captured from the public telephone network), for 819 telephone conversations of 10 to 12 minutes in duration. The audio files are in NIST SPHERE format (1024-byte ASCII file headers). </p><p>Native speakers of Caribbean Spanish and non-Caribbean Spanish were recruited from within the continental United States and Puerto Rico. The following tables provide an overview of the demographics of the participants. The <a href="http://www.ldc.upenn.edu/Catalog/docs/LDC2010T04/fsp06_subj.tbl" rel="nofollow">Subjects Table</a> file, provided in the documentation, may be used to answer questions about specific combinations of participant characteristics (including level of participation). </p><table><tr> <td> <table> <tr> <td>Participants</td> <td>Country Raised</td> </tr> <tr> <td>47</td> <td>U.S.A.</td> </tr> <tr> <td>20</td> <td>Argentina</td> </tr> <tr> <td>14</td> <td>Mexico</td> </tr> <tr> <td>11</td> <td>Colombia</td> </tr> <tr> <td>7</td> <td>Chile</td> </tr> <tr> <td>6</td> <td>Puerto Rico</td> </tr> <tr> <td>5</td> <td>Spain</td> </tr> <tr> <td>5</td> <td>Peru</td> </tr> <tr> <td>3</td> <td>Venezuela</td> </tr> <tr> <td>3</td> <td>Canada</td> </tr> <tr> <td>3</td> <td>Panama</td> </tr> <tr> <td>3</td> <td>Guatemala</td> </tr> <tr> <td>2</td> <td>Paraguay</td> </tr> <tr> <td>1</td> <td>Cuba</td> </tr> <tr> <td>1</td> <td>Honduras</td> </tr> <tr> <td>1</td> <td>Uruguay</td> </tr> <tr> <td>1</td> <td>Bolivia</td> </tr> <tr> <td>1</td> <td>Dominican Republic</td> </tr> <tr> <td>1</td> <td>Switzerland</td> </tr> <tr> <td>1</td> <td>Ecuador</td> </tr> </table> </td> <td> <table> <tr> <td>Conversation Sides</td> <td>Participants</td> </tr> <tr> <td>1</td> <td>6</td> </tr> <tr> <td>2</td> <td>5</td> </tr> <tr> <td>3</td> <td>4</td> </tr> <tr> <td>4</td> <td>3</td> </tr> <tr> <td>5</td> <td>3</td> </tr> <tr> <td>6</td> <td>2</td> </tr> <tr> <td>7</td> <td>2</td> </tr> <tr> <td>8</td> <td>1</td> </tr> <tr> <td>9</td> <td>1</td> </tr> <tr> <td>10</td> <td>13</td> </tr> <tr> <td>11</td> <td>10</td> </tr> <tr> <td>12</td> <td>9</td> </tr> <tr> <td>13</td> <td>8</td> </tr> <tr> <td>14</td> <td>8</td> </tr> <tr> <td>15</td> <td>7</td> </tr> <tr> <td>16</td> <td>7</td> </tr> <tr> <td>17</td> <td>7</td> </tr> <tr> <td>18</td> <td>7</td> </tr> <tr> <td>19</td> <td>7</td> </tr> <tr> <td>20</td> <td>6</td> </tr> <tr> <td>21</td> <td>5</td> </tr> <tr> <td>22</td> <td>5</td> </tr> <tr> <td>23</td> <td>5</td> </tr> <tr> <td>24</td> <td>5</td> </tr> </table> </td> <td> <table> <tr> <td>Years Education</td> <td>Participants</td> </tr> <tr> <td>2</td> <td>1</td> </tr> <tr> <td>4</td> <td>2</td> </tr> <tr> <td>5</td> <td>2</td> </tr> <tr> <td>6</td> <td>1</td> </tr> <tr> <td>11</td> <td>1</td> </tr> <tr> <td>12</td> <td>15</td> </tr> <tr> <td>13</td> <td>7</td> </tr> <tr> <td>14</td> <td>16</td> </tr> <tr> <td>15</td> <td>12</td> </tr> <tr> <td>16</td> <td>25</td> </tr> <tr> <td>17</td> <td>10</td> </tr> <tr> <td>18</td> <td>16</td> </tr> <tr> <td>19</td> <td>3</td> </tr> <tr> <td>20</td> <td>9</td> </tr> <tr> <td>21</td> <td>2</td> </tr> <tr> <td>22</td> <td>6</td> </tr> <tr> <td>23</td> <td>4</td> </tr> <tr> <td>24</td> <td>1</td> </tr> <tr> <td>25</td> <td>2</td> </tr> <tr> <td>28</td> <td>1</td> </tr> </table> </td> </tr></table><table><tr> <td> <table> <tr> <td>Participants</td> <td>Dialect</td> </tr> <tr> <td>91</td> <td>Non-Caribbean</td> </tr> <tr> <td>45</td> <td>Caribbean</td> </tr> </table> </td> <td> <table> <tr> <td>Participants</td> <td>Age Group</td> </tr> <tr> <td>23</td> <td>Young</td> </tr> <tr> <td>106</td> <td>Middle</td> </tr> <tr> <td>7</td> <td>Old</td> </tr> </table> </td> <td> <table> <tr> <td>Participants</td> <td>Sex</td> </tr> <tr> <td>84</td> <td>Female</td> </tr> <tr> <td>52</td> <td>Male</td> </tr> </table> </td> </tr></table><h3>Samples</h3> <p>Please examine this <a href="desc/addenda/LDC2010T04.jpg" rel="nofollow">sample</a> to see an example of the data in this corpus. </p><h3>Updates</h3> <p>12/22/10 - The online documentation for this data now contains a new version of the "fsp_call.tbl", in which an extra column has been included in each row to identify the topic number used in each call. A copy of the topic list in English is also available under fsp06_topics_in_english.txt.</p> </br>
Portions © 2005-2006, 2010 Trustees of the University of Pennsylvania

<h3>介绍</h3> <p>Fisher Spanish - Transcripts 由语言数据联盟（Linguistic Data Consortium, LDC）开发，包含了<a href="http://catalog.ldc.upenn.edu/LDC2010S01" rel="nofollow">Fisher Spanish - Speech (LDC2010S01)</a>中电话语音的全正字法转录文本。该转录数据集涵盖了来自136名加勒比西班牙语母语者与非加勒比西班牙语母语者的约163小时电话语音数据。</p><p>Fisher电话会话采集协议由LDC制定，旨在满足研发鲁棒自动语音识别（Automatic Speech Recognition, ASR）系统的开发者们的关键需求。此前的采集协议如CALLFRIEND、Switchboard-II及其对应的语料库，虽已适配ASR研究，但最初分别是为语言识别与说话人识别开发的。尽管CALLHOME协议及语料库是为支持ASR技术研发而制定的，但其存在说话人数量少、通话时长较长、采集词汇范围狭窄的问题。CALLHOME的对话极具自然性与私密性，难度较高。</p><p>在Fisher协议框架下，数量众多的参与者各自进行数次短时通话，通话对象通常为互不相识的陌生人，且围绕指定话题展开交流。该模式最大化了说话人间的差异与词汇覆盖广度，但同时也提升了对话的正式程度。</p><p>此前如CALLHOME、CALLFRIEND与Switchboard等协议均依赖参与者主动发起活动以推进采集工作，而Fisher的独特之处在于其采用平台驱动模式，而非参与者驱动模式。有意发起通话的参与者可自行操作，但采集平台将发起绝大多数通话。参与者仅需在注册研究时指定的时段接听电话即可。</p><p>为促进词汇范围的多样性，Fisher的参与者需围绕从每日更新的话题列表中随机抽取的指定话题进行交流，当日配对的所有受试者将分配到同一话题。部分话题继承自此前的Switchboard研究并加以优化，其余话题则专为Fisher协议开发。</p><p>在为本语料库采集数据时，课题组力求在性别、年龄、方言区域与教育水平等多种人口统计类别中实现受试者的代表性分布。</p><p>本语料库隶属于Fisher系列语料库，包括：阿拉伯语CTS黎凡特Fisher训练数据集3（<a href="http://catalog.ldc.upenn.edu/LDC2005S07" rel="nofollow">LDC2005S07</a>、<a href="http://catalog.ldc.upenn.edu/LDC2005T03" rel="nofollow">LDC2005T03</a>）、Fisher英语训练部分2（<a href="http://catalog.ldc.upenn.edu/LDC2005S13" rel="nofollow">LDC2005S13</a>、<a href="http://catalog.ldc.upenn.edu/LDC2005T19" rel="nofollow">LDC2005T19</a>）、Fisher英语训练语音部分1（<a href="http://catalog.ldc.upenn.edu/LDC2004S13" rel="nofollow">LDC2004S13</a>、<a href="http://catalog.ldc.upenn.edu/LDC2004T19" rel="nofollow">LDC2004T19</a>）以及Fisher黎凡特阿拉伯语会话电话语音（<a href="http://catalog.ldc.upenn.edu/LDC2007S02" rel="nofollow">LDC2007S02</a>、<a href="http://catalog.ldc.upenn.edu/LDC2007T04" rel="nofollow">LDC2007T04</a>）</p><h3>数据</h3> <p>转录文件采用UTF-8字符编码的纯文本制表符分隔格式（tdf），由LDC开发的<a href="http://www.ldc.upenn.edu/tools/XTrans/" rel="nofollow">XTrans转录工具</a>生成，该工具可优化处理多声道音频与重叠说话人语音。XTrans可从LDC获取。</p><p>转录员遵循LDC的<a href="http://www.ldc.upenn.edu/Catalog/docs/LDC2010T04/trans_guide_nqrt_span.doc" rel="nofollow">转录指南（NQTR）</a>，该指南随本次发布的文档一同提供。</p><p>每个转录文件的首行为列标题，接下来的两行为可忽略的“注释”（由XTrans使用，以首分号";"与非注释行区分）。实际转录数据（含时间戳、声道编号、转录文本及其他附加信息）始于每个文件的第4行。</p><p><a href="http://catalog.ldc.upenn.edu/LDC2010S01" rel="nofollow">Fisher Spanish - Speech (LDC2010S01)</a>提供了作为本语料库转录基础的数字音频数据，采用819段时长10至12分钟的电话会话的双声道μ-law采样数据，采样率为8000样本/秒（取自公共电话网络）。音频文件采用NIST SPHERE格式（含1024字节ASCII文件头）。</p><p>加勒比西班牙语与非加勒比西班牙语母语者均从美国本土与波多黎各招募。下述表格概述了参与者的人口统计特征。文档中提供的<a href="http://www.ldc.upenn.edu/Catalog/docs/LDC2010T04/fsp06_subj.tbl" rel="nofollow">受试者表格</a>可用于查询特定参与者特征（包括参与程度）的组合信息。</p><table><tr> <td> <table> <tr> <td>参与者数量</td> <td>成长国家</td> </tr> <tr> <td>47</td> <td>美国</td> </tr> <tr> <td>20</td> <td>阿根廷</td> </tr> <tr> <td>14</td> <td>墨西哥</td> </tr> <tr> <td>11</td> <td>哥伦比亚</td> </tr> <tr> <td>7</td> <td>智利</td> </tr> <tr> <td>6</td> <td>波多黎各</td> </tr> <tr> <td>5</td> <td>西班牙</td> </tr> <tr> <td>5</td> <td>秘鲁</td> </tr> <tr> <td>3</td> <td>委内瑞拉</td> </tr> <tr> <td>3</td> <td>加拿大</td> </tr> <tr> <td>3</td> <td>巴拿马</td> </tr> <tr> <td>3</td> <td>危地马拉</td> </tr> <tr> <td>2</td> <td>巴拉圭</td> </tr> <tr> <td>1</td> <td>古巴</td> </tr> <tr> <td>1</td> <td>洪都拉斯</td> </tr> <tr> <td>1</td> <td>乌拉圭</td> </tr> <tr> <td>1</td> <td>玻利维亚</td> </tr> <tr> <td>1</td> <td>多米尼加共和国</td> </tr> <tr> <td>1</td> <td>瑞士</td> </tr> <tr> <td>1</td> <td>厄瓜多尔</td> </tr> </table> </td> <td> <table> <tr> <td>通话侧数</td> <td>参与者数量</td> </tr> <tr> <td>1</td> <td>6</td> </tr> <tr> <td>2</td> <td>5</td> </tr> <tr> <td>3</td> <td>4</td> </tr> <tr> <td>4</td> <td>3</td> </tr> <tr> <td>5</td> <td>3</td> </tr> <tr> <td>6</td> <td>2</td> </tr> <tr> <td>7</td> <td>2</td> </tr> <tr> <td>8</td> <td>1</td> </tr> <tr> <td>9</td> <td>1</td> </tr> <tr> <td>10</td> <td>13</td> </tr> <tr> <td>11</td> <td>10</td> </tr> <tr> <td>12</td> <td>9</td> </tr> <tr> <td>13</td> <td>8</td> </tr> <tr> <td>14</td> <td>8</td> </tr> <tr> <td>15</td> <td>7</td> </tr> <tr> <td>16</td> <td>7</td> </tr> <tr> <td>17</td> <td>7</td> </tr> <tr> <td>18</td> <td>7</td> </tr> <tr> <td>19</td> <td>7</td> </tr> <tr> <td>20</td> <td>6</td> </tr> <tr> <td>21</td> <td>5</td> </tr> <tr> <td>22</td> <td>5</td> </tr> <tr> <td>23</td> <td>5</td> </tr> <tr> <td>24</td> <td>5</td> </tr> </table> </td> <td> <table> <tr> <td>受教育年限</td> <td>参与者数量</td> </tr> <tr> <td>2</td> <td>1</td> </tr> <tr> <td>4</td> <td>2</td> </tr> <tr> <td>5</td> <td>2</td> </tr> <tr> <td>6</td> <td>1</td> </tr> <tr> <td>11</td> <td>1</td> </tr> <tr> <td>12</td> <td>15</td> </tr> <tr> <td>13</td> <td>7</td> </tr> <tr> <td>14</td> <td>16</td> </tr> <tr> <td>15</td> <td>12</td> </tr> <tr> <td>16</td> <td>25</td> </tr> <tr> <td>17</td> <td>10</td> </tr> <tr> <td>18</td> <td>16</td> </tr> <tr> <td>19</td> <td>3</td> </tr> <tr> <td>20</td> <td>9</td> </tr> <tr> <td>21</td> <td>2</td> </tr> <tr> <td>22</td> <td>6</td> </tr> <tr> <td>23</td> <td>4</td> </tr> <tr> <td>24</td> <td>1</td> </tr> <tr> <td>25</td> <td>2</td> </tr> <tr> <td>28</td> <td>1</td> </tr> </table> </td> </tr></table><table><tr> <td> <table> <tr> <td>参与者数量</td> <td>方言类别</td> </tr> <tr> <td>91</td> <td>非加勒比西班牙语</td> </tr> <tr> <td>45</td> <td>加勒比西班牙语</td> </tr> </table> </td> <td> <table> <tr> <td>参与者数量</td> <td>年龄组</td> </tr> <tr> <td>23</td> <td>青年组</td> </tr> <tr> <td>106</td> <td>中年组</td> </tr> <tr> <td>7</td> <td>老年组</td> </tr> </table> </td> <td> <table> <tr> <td>参与者数量</td> <td>性别</td> </tr> <tr> <td>84</td> <td>女性</td> </tr> <tr> <td>52</td> <td>男性</td> </tr> </table> </td> </tr></table><h3>样本</h3> <p>请查看<a href="desc/addenda/LDC2010T04.jpg" rel="nofollow">此样本</a>以了解本语料库数据的示例格式。</p><h3>更新记录</h3> <p>2010年12月22日：本数据的在线文档已更新"fsp_call.tbl"版本，新增一列用于标识每次通话使用的话题编号。英文话题列表副本可通过fsp06_topics_in_english.txt获取。</p><br/> <p>部分内容 © 2005-2006、2010 宾夕法尼亚大学托管委员会</p>