http://hlt.suda.edu.cn/index.php?action=history&feed=atom&title=CCL2021_data_readme 2026-05-18T07:54:59Z 本wiki的该页面的版本历史 MediaWiki 1.35.2 http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3757&oldid=prev 2021-04-02T07:50:20Z

<p><span dir="auto"><span class="autocomment">文件格式说明</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:50的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l13" >第13行：</td> <td colspan="2" class="diff-lineno">第13行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== 文件格式说明 ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== 文件格式说明 ==</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>所有文件均采用utf-8编码</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>所有文件均采用utf-8编码</div></td></tr> <tr><td colspan="2"> </td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;"></ins></div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Train/Dev/Unlabeled数据文件均为“CoNLL”格式，在CoNLL格式中，每个词语占一行，每行10列，无值列用下划线 '_' 代替，列的分隔符为制表符'\t'，行的分隔符为换行符'\n'；句子与句子之间用空行'\n'分隔。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Train/Dev/Unlabeled数据文件均为“CoNLL”格式，在CoNLL格式中，每个词语占一行，每行10列，无值列用下划线 '_' 代替，列的分隔符为制表符'\t'，行的分隔符为换行符'\n'；句子与句子之间用空行'\n'分隔。</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>下面介绍一下我们本次评测使用到的列：</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>下面介绍一下我们本次评测使用到的列：</div></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3756&oldid=prev 2021-04-02T07:50:13Z

<p><span dir="auto"><span class="autocomment">文件格式说明</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:50的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l12" >第12行：</td> <td colspan="2" class="diff-lineno">第12行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== 文件格式说明 ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== 文件格式说明 ==</div></td></tr> <tr><td colspan="2"> </td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;">所有文件均采用utf-8编码</ins></div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Train/Dev/Unlabeled数据文件均为“CoNLL”格式，在CoNLL格式中，每个词语占一行，每行10列，无值列用下划线 '_' 代替，列的分隔符为制表符'\t'，行的分隔符为换行符'\n'；句子与句子之间用空行'\n'分隔。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>Train/Dev/Unlabeled数据文件均为“CoNLL”格式，在CoNLL格式中，每个词语占一行，每行10列，无值列用下划线 '_' 代替，列的分隔符为制表符'\t'，行的分隔符为换行符'\n'；句子与句子之间用空行'\n'分隔。</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>下面介绍一下我们本次评测使用到的列：</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>下面介绍一下我们本次评测使用到的列：</div></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3755&oldid=prev 2021-04-02T07:49:54Z

<p><span dir="auto"><span class="autocomment">Word Embedding预训练</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:49的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l23" >第23行：</td> <td colspan="2" class="diff-lineno">第23行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Word Embedding预训练 ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Word Embedding预训练 ==</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*是通过word2vec在Chinese Gigaword V3（约1100万句、自动分词）和所有领域Train/Unlabeled上训练10词迭代得到。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*是通过word2vec在Chinese Gigaword V3（约1100万句、自动分词）和所有领域Train/Unlabeled上训练10词迭代得到。</div></td></tr> <tr><td class='diff-marker'>−</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div><del style="font-weight: bold; text-decoration: none;">*Word Embedding文件均采用UTF-8编码。</del></div></td><td colspan="2"> </td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* 使用hugging face的脚本（transformers==3.3.1）在所有领域Train/Unlabeled上训练fine-tune bert 3轮。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* 使用hugging face的脚本（transformers==3.3.1）在所有领域Train/Unlabeled上训练fine-tune bert 3轮。</div></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3754&oldid=prev 2021-04-02T07:49:38Z

<p><span dir="auto"><span class="autocomment">fine-tune-bert</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:49的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l26" >第26行：</td> <td colspan="2" class="diff-lineno">第26行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td></tr> <tr><td class='diff-marker'>−</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>* <del class="diffchange diffchange-inline">我们基于 Peng et al</del>. <del class="diffchange diffchange-inline">(2019)中描述的预训练语言模型微调的方法，将 Google 发布的 </del>bert<del class="diffchange diffchange-inline">-base-chinese 在所有无标注数据（6 个领域的Train，5 个领域的 Unlabeled 数据）进行多轮微调（3 轮），即继续使用语言模型损失函数继续训练。</del></div></td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>* <ins class="diffchange diffchange-inline">使用hugging face的脚本（transformers==3</ins>.<ins class="diffchange diffchange-inline">3.1）在所有领域Train/Unlabeled上训练fine-tune </ins>bert <ins class="diffchange diffchange-inline">3轮。</ins></div></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3753&oldid=prev 2021-04-02T07:48:39Z

<p><span dir="auto"><span class="autocomment">Word Embedding预训练</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:48的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l22" >第22行：</td> <td colspan="2" class="diff-lineno">第22行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Word Embedding预训练 ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== Word Embedding预训练 ==</div></td></tr> <tr><td class='diff-marker'>−</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>*<del class="diffchange diffchange-inline">Word Embedding是通过word2vec在Chinese </del>Gigaword <del class="diffchange diffchange-inline">V3（约1100万句、自动分词）和目标领域Train</del>/<del class="diffchange diffchange-inline">Unlabeled上训练10次迭代得到。</del></div></td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>*<ins class="diffchange diffchange-inline">是通过word2vec在Chinese </ins>Gigaword <ins class="diffchange diffchange-inline">V3（约1100万句、自动分词）和所有领域Train</ins>/<ins class="diffchange diffchange-inline">Unlabeled上训练10词迭代得到。</ins></div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*Word Embedding文件均采用UTF-8编码。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*Word Embedding文件均采用UTF-8编码。</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* 我们基于 Peng et al. (2019)中描述的预训练语言模型微调的方法，将 Google 发布的 bert-base-chinese 在所有无标注数据（6 个领域的Train，5 个领域的 Unlabeled 数据）进行多轮微调（3 轮），即继续使用语言模型损失函数继续训练。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* 我们基于 Peng et al. (2019)中描述的预训练语言模型微调的方法，将 Google 发布的 bert-base-chinese 在所有无标注数据（6 个领域的Train，5 个领域的 Unlabeled 数据）进行多轮微调（3 轮），即继续使用语言模型损失函数继续训练。</div></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3750&oldid=prev 2021-04-02T07:38:44Z

<p><span dir="auto"><span class="autocomment">fine-tune-bert</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:38的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l26" >第26行：</td> <td colspan="2" class="diff-lineno">第26行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== fine-tune-bert ==</div></td></tr> <tr><td colspan="2"> </td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;">* 我们基于 Peng et al. (2019)中描述的预训练语言模型微调的方法，将 Google 发布的 bert-base-chinese 在所有无标注数据（6 个领域的Train，5 个领域的 Unlabeled 数据）进行多轮微调（3 轮），即继续使用语言模型损失函数继续训练。</ins></div></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3749&oldid=prev 2021-04-02T07:36:01Z

<p><span dir="auto"><span class="autocomment">Word Embedding预训练</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:36的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l24" >第24行：</td> <td colspan="2" class="diff-lineno">第24行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*Word Embedding是通过word2vec在Chinese Gigaword V3（约1100万句、自动分词）和目标领域Train/Unlabeled上训练10次迭代得到。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*Word Embedding是通过word2vec在Chinese Gigaword V3（约1100万句、自动分词）和目标领域Train/Unlabeled上训练10次迭代得到。</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*Word Embedding文件均采用UTF-8编码。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>*Word Embedding文件均采用UTF-8编码。</div></td></tr> <tr><td colspan="2"> </td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;"></ins></div></td></tr> <tr><td colspan="2"> </td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;">== fine-tune-bert ==</ins></div></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3748&oldid=prev 2021-04-02T07:32:05Z

<p><span dir="auto"><span class="autocomment">train_dev_unlabeled</span></span></p> <table class="diff diff-contentalign-left diff-editfont-monospace" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="zh-Hans-CN"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">←上一版本</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">2021年4月2日 (五) 07:32的版本</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l2" >第2行：</td> <td colspan="2" class="diff-lineno">第2行：</td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== train_dev_unlabeled==</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>== train_dev_unlabeled==</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* Unlabeled文件夹：</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* Unlabeled文件夹：</div></td></tr> <tr><td class='diff-marker'>−</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>**包括PB-Unlabeled.conll、PC-Unlabeled.conll、ZX-Unlabeled.<del class="diffchange diffchange-inline">conll三个文件；</del></div></td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>**包括PB-Unlabeled.conll、PC-Unlabeled.conll、ZX-Unlabeled.<ins class="diffchange diffchange-inline">conll、FIN-Unlabeled.conll、LEG-Unlabeled.conll五个文件；</ins></div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>**无标注数据仅提供分词和词性（第四列）信息；</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>**无标注数据仅提供分词和词性（第四列）信息；</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* Train文件夹：</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* Train文件夹：</div></td></tr> <tr><td class='diff-marker'>−</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>**包括BC-Train.conll、PB-Train.conll、PC-Train.conll、ZX-Train.<del class="diffchange diffchange-inline">conll四个文件；</del></div></td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>**包括BC-Train<ins class="diffchange diffchange-inline">-full</ins>.conll、PB-Train<ins class="diffchange diffchange-inline">-full</ins>.conll、PC-Train<ins class="diffchange diffchange-inline">-full</ins>.conll、ZX-Train<ins class="diffchange diffchange-inline">-full.conll、FIN-Train-full.conll、LEG-Train-full</ins>.<ins class="diffchange diffchange-inline">conll六个文件；</ins></div></td></tr> <tr><td class='diff-marker'>−</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>**<del class="diffchange diffchange-inline">文件中最后一列为依存弧的概率，如果概率为1，则对应弧为人工标注，否则为模型补全；</del></div></td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>**<ins class="diffchange diffchange-inline">文件中最后一列为依存弧的概率，如果概率为2，则对应弧为人工标注，否则为模型补全；</ins></div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* Dev文件夹：</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>* Dev文件夹：</div></td></tr> <tr><td class='diff-marker'>−</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>**包括BC-Dev.conll、PB-Dev.conll、PC-Dev.conll、ZX-Dev.<del class="diffchange diffchange-inline">conll四个文件；</del></div></td><td class='diff-marker'>+</td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div>**包括BC-Dev.conll、PB-Dev.conll、PC-Dev.conll、ZX-Dev.<ins class="diffchange diffchange-inline">conll、FIN-Dev.conll、LEG-Dev.conll六个文件；</ins></div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>**文件中第7列为“-1”且第8列为“none”,代表该弧未进行人工标注。</div></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>**文件中第7列为“-1”且第8列为“none”,代表该弧未进行人工标注。</div></td></tr> <tr><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td><td class='diff-marker'> </td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"></td></tr> </table>

Myzhou http://hlt.suda.edu.cn/index.php?title=CCL2021_data_readme&diff=3747&oldid=prev 2021-04-02T07:27:46Z

<p>创建页面，内容为“ == train_dev_unlabeled== * Unlabeled文件夹： **包括PB-Unlabeled.conll、PC-Unlabeled.conll、ZX-Unlabeled.conll三个文件； **无标注数据仅提供分...”</p> <p><b>新页面</b></p><div><br /> == train_dev_unlabeled==<br /> * Unlabeled文件夹：<br /> **包括PB-Unlabeled.conll、PC-Unlabeled.conll、ZX-Unlabeled.conll三个文件；<br /> **无标注数据仅提供分词和词性（第四列）信息；<br /> * Train文件夹：<br /> **包括BC-Train.conll、PB-Train.conll、PC-Train.conll、ZX-Train.conll四个文件；<br /> **文件中最后一列为依存弧的概率，如果概率为1，则对应弧为人工标注，否则为模型补全；<br /> * Dev文件夹：<br /> **包括BC-Dev.conll、PB-Dev.conll、PC-Dev.conll、ZX-Dev.conll四个文件；<br /> **文件中第7列为“-1”且第8列为“none”,代表该弧未进行人工标注。<br /> <br /> == 文件格式说明 ==<br /> Train/Dev/Unlabeled数据文件均为“CoNLL”格式，在CoNLL格式中，每个词语占一行，每行10列，无值列用下划线 '_' 代替，列的分隔符为制表符'\t'，行的分隔符为换行符'\n'；句子与句子之间用空行'\n'分隔。<br /> 下面介绍一下我们本次评测使用到的列：<br /> *1.当前词在句子中的序号，从1开始；<br /> *2.当前词语或标点；<br /> *4.当前词语的词性；<br /> *7.当前词语的核心词；<br /> *8.当前词语与核心词的依存关系；<br /> *10.依存弧的概率；<br /> <br /> == Word Embedding预训练 ==<br /> *Word Embedding是通过word2vec在Chinese Gigaword V3（约1100万句、自动分词）和目标领域Train/Unlabeled上训练10次迭代得到。<br /> *Word Embedding文件均采用UTF-8编码。</div>

Myzhou