yangheng95 / PyABSA

Sentiment Analysis, Text Classification, Text Augmentation, Text Adversarial defense, etc.;
https://pyabsa.readthedocs.io
MIT License
923 stars 159 forks source link

多观点的抽出是否支持? #27

Closed LangDaoAI closed 3 years ago

LangDaoAI commented 3 years ago

确认一下多观点的抽出是否支持, 或者提供一些参考,以及一些ideas。

yangheng95 commented 3 years ago

数据我来提供,加入观点后标记规则是否要做修改 Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 2:59:11 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27) 12号后我再用新的模型验证一下,这几天出差,另外观点抽出可能模型架构要改动的,有空还请看一下,或者说当前模型架构是否支持这样的观点抽出? Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: Tang Yong @.> Sent: Friday, June 4, 2021 7:18:13 PM To: yangheng95/pyabsa @.>; yangheng95/pyabsa @.> Cc: Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27>) 好的感谢! Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Friday, June 4, 2021 7:17:27 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27>) 合并的意思是把四个分散的数据集全部放到一个里面,只是单纯追加吗,可否详细说一下,谢谢 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Friday, June 4, 2021 6:38:12 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27>) 本地替换了方法做了测试,如下图,切词没有问题, 但是BIOE标注(红框:颜色,aspect只抽取了颜)好像有点问题,请看一下, [image]https://user-images.githubusercontent.com/83859417/120736527-fbda2380-c51e-11eb-94b0-c30ecea42979.png 有空格和没有空格都是一样的。 Sentence with predicted labels: 尤(O) 其(O) 是(O) 照(O) 的(O) 大(O) 尺(O) 寸(O) 照(O) 片(O) 时(O) 效(B-ASP) 果(I-ASP) 也(O) 是(O) 非(O) 常(O) 不(O) 错(O) 的(O) {'aspect': '效 果', 'position': '11,12', 'sentiment': 'Good'} Sentence with predicted labels: 照(O) 大(O) 尺(O) 寸(O) 的(O) 照(O) 片(O) 的(O) 时(O) 候(O) 手(O) 机(O) 反(O) 映(O) 速(B-ASP) 度(I-ASP) 太(O) 慢(O) {'aspect': '速 度', 'position': '14,15', 'sentiment': 'Bad'} Sentence with predicted labels: 关(O) 键(O) 的(O) 时(O) 候(O) 需(O) 要(O) 表(O) 现(O) 持(O) 续(O) 影(O) 像(O) 的(O) 短(O) 片(O) 功(B-ASP) 能(I-ASP) 还(O) 是(O) 很(O) 有(O) 用(O) 的(O) {'aspect': '功 能', 'position': '16,17', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 这(O) 款(O) 手(O) 机(O) 的(O) 大(B-ASP) 小(I-ASP) 真(O) 的(O) 很(O) 薄(O) ,(O) 但(O) 是(O) 颜(B-ASP) 色(I-ASP) 不(O) 太(O) 好(O) 看(O) ,(O) 总(O) 体(O) 上(O) 我(O) 很(O) 满(O) 意(O) 啦(O) 。(O) {'aspect': '大 小', 'position': '5,6', 'sentiment': 'Bad'} {'aspect': '颜 色', 'position': '14,15', 'sentiment': 'Bad'} 这是综合4个中文数据集训练的模型的结果 颜色问题不知道是如何解决的,还是说模型做了更新还是数据量做了增强,很想知道, 另外, 新的代码构建的模型空了也帮忙上传一下, 我来做更新测试。 主要原因是将4个中文数据集合并了,数据量增多,代码已经更新,正在上传模型 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27<#27> (comment)<#27 (comment)<#27 (comment)>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWNZFDTWARXWMBAVCNTTRCUJJANCNFSM46AW7SCQ. 就是合并所有的中文训练集和测试集作为一整个训练集和测试集,现在可以从google drive下载中文模型,读取模型之后会打印训练时的参数,其中的datasets_files就是采用的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)<#27 (comment)>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWPQDEJONX3FEGBOKIDTRCY4PANCNFSM46AW7SCQ. 改模型倒不是最大的问题, 问题是没有标注好opinion的数据集 ― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMACZNRK5NF6JXYGXDTRMMD7ANCNFSM46AW7SCQ.

使用IOB形式将opinion标记在数据集第4列(polarity之后),如果您使用开源数据集标注,请不要直接修改原始数据集,将重新标注的数据集重新命名上传。

LangDaoAI commented 3 years ago

晚上我标记一个例子,确认后再大规模标记

Get Outlook for Androidhttps://aka.ms/AAb9ysg


From: YangHeng @.> Sent: Sunday, June 6, 2021 4:47:36 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27)

数据我来提供,加入观点后标记规则是否要做修改 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 2:59:11 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27) 12号后我再用新的模型验证一下,这几天出差,另外观点抽出可能模型架构要改动的,有空还请看一下,或者说当前模型架构是否支持这样的观点抽出? Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: Tang Yong @.> Sent: Friday, June 4, 2021 7:18:13 PM To: yangheng95/pyabsa @.>; yangheng95/pyabsa @.> Cc: Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27>) 好的感谢! Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Friday, June 4, 2021 7:17:27 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27>) 合并的意思是把四个分散的数据集全部放到一个里面,只是单纯追加吗,可否详细说一下,谢谢 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Friday, June 4, 2021 6:38:12 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 本地替换了方法做了测试,如下图,切词没有问题, 但是BIOE标注(红框:颜色,aspect只抽取了颜)好像有点问题,请看一下, [image]https://user-images.githubusercontent.com/83859417/120736527-fbda2380-c51e-11eb-94b0-c30ecea42979.png 有空格和没有空格都是一样的。 Sentence with predicted labels: 尤(O) 其(O) 是(O) 照(O) 的(O) 大(O) 尺(O) 寸(O) 照(O) 片(O) 时(O) 效(B-ASP) 果(I-ASP) 也(O) 是(O) 非(O) 常(O) 不(O) 错(O) 的(O) {'aspect': '效 果', 'position': '11,12', 'sentiment': 'Good'} Sentence with predicted labels: 照(O) 大(O) 尺(O) 寸(O) 的(O) 照(O) 片(O) 的(O) 时(O) 候(O) 手(O) 机(O) 反(O) 映(O) 速(B-ASP) 度(I-ASP) 太(O) 慢(O) {'aspect': '速 度', 'position': '14,15', 'sentiment': 'Bad'} Sentence with predicted labels: 关(O) 键(O) 的(O) 时(O) 候(O) 需(O) 要(O) 表(O) 现(O) 持(O) 续(O) 影(O) 像(O) 的(O) 短(O) 片(O) 功(B-ASP) 能(I-ASP) 还(O) 是(O) 很(O) 有(O) 用(O) 的(O) {'aspect': '功 能', 'position': '16,17', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 这(O) 款(O) 手(O) 机(O) 的(O) 大(B-ASP) 小(I-ASP) 真(O) 的(O) 很(O) 薄(O) ,(O) 但(O) 是(O) 颜(B-ASP) 色(I-ASP) 不(O) 太(O) 好(O) 看(O) ,(O) 总(O) 体(O) 上(O) 我(O) 很(O) 满(O) 意(O) 啦(O) 。(O) {'aspect': '大 小', 'position': '5,6', 'sentiment': 'Bad'} {'aspect': '颜 色', 'position': '14,15', 'sentiment': 'Bad'} 这是综合4个中文数据集训练的模型的结果 颜色问题不知道是如何解决的,还是说模型做了更新还是数据量做了增强,很想知道, 另外, 新的代码构建的模型空了也帮忙上传一下, 我来做更新测试。 主要原因是将4个中文数据集合并了,数据量增多,代码已经更新,正在上传模型 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-854606848>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWNZFDTWARXWMBAVCNTTRCUJJANCNFSM46AW7SCQ. 就是合并所有的中文训练集和测试集作为一整个训练集和测试集,现在可以从google drive下载中文模型,读取模型之后会打印训练时的参数,其中的datasets_files就是采用的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-854627966>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWPQDEJONX3FEGBOKIDTRCY4PANCNFSM46AW7SCQ. 改模型倒不是最大的问题, 问题是没有标注好opinion的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-855350091>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMACZNRK5NF6JXYGXDTRMMD7ANCNFSM46AW7SCQ.

使用IOB形式将opinion标记在数据集第4列(polarity之后),如果您使用开源数据集标注,请不要直接修改原始数据集,将重新标注的数据集重新命名上传。

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/yangheng95/PyABSA/issues/27#issuecomment-855363169, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMGE7XZ5WWWWQRKFJDTRMY2RANCNFSM46AW7SCQ.

yangheng95 commented 3 years ago

晚上我标记一个例子,确认后再大规模标记 Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 4:47:36 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27) 数据我来提供,加入观点后标记规则是否要做修改 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 2:59:11 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27<#27>) 12号后我再用新的模型验证一下,这几天出差,另外观点抽出可能模型架构要改动的,有空还请看一下,或者说当前模型架构是否支持这样的观点抽出? Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: Tang Yong @.> Sent: Friday, June 4, 2021 7:18:13 PM To: yangheng95/pyabsa @.>; yangheng95/pyabsa @.> Cc: Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27>) 好的感谢! Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Friday, June 4, 2021 7:17:27 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27>) 合并的意思是把四个分散的数据集全部放到一个里面,只是单纯追加吗,可否详细说一下,谢谢 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Friday, June 4, 2021 6:38:12 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 本地替换了方法做了测试,如下图,切词没有问题, 但是BIOE标注(红框:颜色,aspect只抽取了颜)好像有点问题,请看一下, [image]https://user-images.githubusercontent.com/83859417/120736527-fbda2380-c51e-11eb-94b0-c30ecea42979.png 有空格和没有空格都是一样的。 Sentence with predicted labels: 尤(O) 其(O) 是(O) 照(O) 的(O) 大(O) 尺(O) 寸(O) 照(O) 片(O) 时(O) 效(B-ASP) 果(I-ASP) 也(O) 是(O) 非(O) 常(O) 不(O) 错(O) 的(O) {'aspect': '效 果', 'position': '11,12', 'sentiment': 'Good'} Sentence with predicted labels: 照(O) 大(O) 尺(O) 寸(O) 的(O) 照(O) 片(O) 的(O) 时(O) 候(O) 手(O) 机(O) 反(O) 映(O) 速(B-ASP) 度(I-ASP) 太(O) 慢(O) {'aspect': '速 度', 'position': '14,15', 'sentiment': 'Bad'} Sentence with predicted labels: 关(O) 键(O) 的(O) 时(O) 候(O) 需(O) 要(O) 表(O) 现(O) 持(O) 续(O) 影(O) 像(O) 的(O) 短(O) 片(O) 功(B-ASP) 能(I-ASP) 还(O) 是(O) 很(O) 有(O) 用(O) 的(O) {'aspect': '功 能', 'position': '16,17', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 这(O) 款(O) 手(O) 机(O) 的(O) 大(B-ASP) 小(I-ASP) 真(O) 的(O) 很(O) 薄(O) ,(O) 但(O) 是(O) 颜(B-ASP) 色(I-ASP) 不(O) 太(O) 好(O) 看(O) ,(O) 总(O) 体(O) 上(O) 我(O) 很(O) 满(O) 意(O) 啦(O) 。(O) {'aspect': '大 小', 'position': '5,6', 'sentiment': 'Bad'} {'aspect': '颜 色', 'position': '14,15', 'sentiment': 'Bad'} 这是综合4个中文数据集训练的模型的结果 颜色问题不知道是如何解决的,还是说模型做了更新还是数据量做了增强,很想知道, 另外, 新的代码构建的模型空了也帮忙上传一下, 我来做更新测试。 主要原因是将4个中文数据集合并了,数据量增多,代码已经更新,正在上传模型 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27<#27> (comment)<#27 (comment)<#27 (comment)>>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWNZFDTWARXWMBAVCNTTRCUJJANCNFSM46AW7SCQ. 就是合并所有的中文训练集和测试集作为一整个训练集和测试集,现在可以从google drive下载中文模型,读取模型之后会打印训练时的参数,其中的datasets_files就是采用的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27<#27> (comment)<#27 (comment)<#27 (comment)>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWPQDEJONX3FEGBOKIDTRCY4PANCNFSM46AW7SCQ. 改模型倒不是最大的问题, 问题是没有标注好opinion的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)<#27 (comment)>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMACZNRK5NF6JXYGXDTRMMD7ANCNFSM46AW7SCQ. 使用IOB形式将opinion标记在数据集第4列(polarity之后),如果您使用开源数据集标注,请不要直接修改原始数据集,将重新标注的数据集重新命名上传。 ― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMGE7XZ5WWWWQRKFJDTRMY2RANCNFSM46AW7SCQ.

这个issue我先关闭了,后面沟通就gitter上吧。数据的标注可以参照https://github.com/lixin4ever/E2E-TBSA

LangDaoAI commented 3 years ago

好的,晚上没抽出时间,gitter沟通是啥意思

Get Outlook for Androidhttps://aka.ms/AAb9ysg


From: YangHeng @.> Sent: Sunday, June 6, 2021 11:41:29 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27)

晚上我标记一个例子,确认后再大规模标记 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 4:47:36 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27) 数据我来提供,加入观点后标记规则是否要做修改 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 2:59:11 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27>) 12号后我再用新的模型验证一下,这几天出差,另外观点抽出可能模型架构要改动的,有空还请看一下,或者说当前模型架构是否支持这样的观点抽出? Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: Tang Yong @.> Sent: Friday, June 4, 2021 7:18:13 PM To: yangheng95/pyabsa @.>; yangheng95/pyabsa @.> Cc: Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 好的感谢! Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Friday, June 4, 2021 7:17:27 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 合并的意思是把四个分散的数据集全部放到一个里面,只是单纯追加吗,可否详细说一下,谢谢 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Friday, June 4, 2021 6:38:12 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 本地替换了方法做了测试,如下图,切词没有问题, 但是BIOE标注(红框:颜色,aspect只抽取了颜)好像有点问题,请看一下, [image]https://user-images.githubusercontent.com/83859417/120736527-fbda2380-c51e-11eb-94b0-c30ecea42979.png 有空格和没有空格都是一样的。 Sentence with predicted labels: 尤(O) 其(O) 是(O) 照(O) 的(O) 大(O) 尺(O) 寸(O) 照(O) 片(O) 时(O) 效(B-ASP) 果(I-ASP) 也(O) 是(O) 非(O) 常(O) 不(O) 错(O) 的(O) {'aspect': '效 果', 'position': '11,12', 'sentiment': 'Good'} Sentence with predicted labels: 照(O) 大(O) 尺(O) 寸(O) 的(O) 照(O) 片(O) 的(O) 时(O) 候(O) 手(O) 机(O) 反(O) 映(O) 速(B-ASP) 度(I-ASP) 太(O) 慢(O) {'aspect': '速 度', 'position': '14,15', 'sentiment': 'Bad'} Sentence with predicted labels: 关(O) 键(O) 的(O) 时(O) 候(O) 需(O) 要(O) 表(O) 现(O) 持(O) 续(O) 影(O) 像(O) 的(O) 短(O) 片(O) 功(B-ASP) 能(I-ASP) 还(O) 是(O) 很(O) 有(O) 用(O) 的(O) {'aspect': '功 能', 'position': '16,17', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 这(O) 款(O) 手(O) 机(O) 的(O) 大(B-ASP) 小(I-ASP) 真(O) 的(O) 很(O) 薄(O) ,(O) 但(O) 是(O) 颜(B-ASP) 色(I-ASP) 不(O) 太(O) 好(O) 看(O) ,(O) 总(O) 体(O) 上(O) 我(O) 很(O) 满(O) 意(O) 啦(O) 。(O) {'aspect': '大 小', 'position': '5,6', 'sentiment': 'Bad'} {'aspect': '颜 色', 'position': '14,15', 'sentiment': 'Bad'} 这是综合4个中文数据集训练的模型的结果 颜色问题不知道是如何解决的,还是说模型做了更新还是数据量做了增强,很想知道, 另外, 新的代码构建的模型空了也帮忙上传一下, 我来做更新测试。 主要原因是将4个中文数据集合并了,数据量增多,代码已经更新,正在上传模型 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-854606848>>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWNZFDTWARXWMBAVCNTTRCUJJANCNFSM46AW7SCQ. 就是合并所有的中文训练集和测试集作为一整个训练集和测试集,现在可以从google drive下载中文模型,读取模型之后会打印训练时的参数,其中的datasets_files就是采用的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-854627966>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWPQDEJONX3FEGBOKIDTRCY4PANCNFSM46AW7SCQ. 改模型倒不是最大的问题, 问题是没有标注好opinion的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-855350091>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMACZNRK5NF6JXYGXDTRMMD7ANCNFSM46AW7SCQ. 使用IOB形式将opinion标记在数据集第4列(polarity之后),如果您使用开源数据集标注,请不要直接修改原始数据集,将重新标注的数据集重新命名上传。 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-855363169>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMGE7XZ5WWWWQRKFJDTRMY2RANCNFSM46AW7SCQ.

这个issue我先关闭了,后面沟通就gitter上吧。数据的标注可以参照https://github.com/lixin4ever/E2E-TBSA

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/yangheng95/PyABSA/issues/27#issuecomment-855418679, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMIHCZ2RZ3DAN7NHHTTROJKTANCNFSM46AW7SCQ.

yangheng95 commented 3 years ago

好的,晚上没抽出时间,gitter沟通是啥意思 Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 11:41:29 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27) 晚上我标记一个例子,确认后再大规模标记 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 4:47:36 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27<#27>) 数据我来提供,加入观点后标记规则是否要做修改 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 2:59:11 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27>) 12号后我再用新的模型验证一下,这几天出差,另外观点抽出可能模型架构要改动的,有空还请看一下,或者说当前模型架构是否支持这样的观点抽出? Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: Tang Yong @.> Sent: Friday, June 4, 2021 7:18:13 PM To: yangheng95/pyabsa @.>; yangheng95/pyabsa @.> Cc: Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 好的感谢! Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Friday, June 4, 2021 7:17:27 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 合并的意思是把四个分散的数据集全部放到一个里面,只是单纯追加吗,可否详细说一下,谢谢 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Friday, June 4, 2021 6:38:12 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 本地替换了方法做了测试,如下图,切词没有问题, 但是BIOE标注(红框:颜色,aspect只抽取了颜)好像有点问题,请看一下, [image]https://user-images.githubusercontent.com/83859417/120736527-fbda2380-c51e-11eb-94b0-c30ecea42979.png 有空格和没有空格都是一样的。 Sentence with predicted labels: 尤(O) 其(O) 是(O) 照(O) 的(O) 大(O) 尺(O) 寸(O) 照(O) 片(O) 时(O) 效(B-ASP) 果(I-ASP) 也(O) 是(O) 非(O) 常(O) 不(O) 错(O) 的(O) {'aspect': '效 果', 'position': '11,12', 'sentiment': 'Good'} Sentence with predicted labels: 照(O) 大(O) 尺(O) 寸(O) 的(O) 照(O) 片(O) 的(O) 时(O) 候(O) 手(O) 机(O) 反(O) 映(O) 速(B-ASP) 度(I-ASP) 太(O) 慢(O) {'aspect': '速 度', 'position': '14,15', 'sentiment': 'Bad'} Sentence with predicted labels: 关(O) 键(O) 的(O) 时(O) 候(O) 需(O) 要(O) 表(O) 现(O) 持(O) 续(O) 影(O) 像(O) 的(O) 短(O) 片(O) 功(B-ASP) 能(I-ASP) 还(O) 是(O) 很(O) 有(O) 用(O) 的(O) {'aspect': '功 能', 'position': '16,17', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 这(O) 款(O) 手(O) 机(O) 的(O) 大(B-ASP) 小(I-ASP) 真(O) 的(O) 很(O) 薄(O) ,(O) 但(O) 是(O) 颜(B-ASP) 色(I-ASP) 不(O) 太(O) 好(O) 看(O) ,(O) 总(O) 体(O) 上(O) 我(O) 很(O) 满(O) 意(O) 啦(O) 。(O) {'aspect': '大 小', 'position': '5,6', 'sentiment': 'Bad'} {'aspect': '颜 色', 'position': '14,15', 'sentiment': 'Bad'} 这是综合4个中文数据集训练的模型的结果 颜色问题不知道是如何解决的,还是说模型做了更新还是数据量做了增强,很想知道, 另外, 新的代码构建的模型空了也帮忙上传一下, 我来做更新测试。 主要原因是将4个中文数据集合并了,数据量增多,代码已经更新,正在上传模型 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27<#27> (comment)<#27 (comment)<#27 (comment)>>>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWNZFDTWARXWMBAVCNTTRCUJJANCNFSM46AW7SCQ. 就是合并所有的中文训练集和测试集作为一整个训练集和测试集,现在可以从google drive下载中文模型,读取模型之后会打印训练时的参数,其中的datasets_files就是采用的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27<#27><#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27<#27> (comment)<#27 (comment)<#27 (comment)>>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWPQDEJONX3FEGBOKIDTRCY4PANCNFSM46AW7SCQ. 改模型倒不是最大的问题, 问题是没有标注好opinion的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27<#27> (comment)<#27 (comment)<#27 (comment)>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMACZNRK5NF6JXYGXDTRMMD7ANCNFSM46AW7SCQ. 使用IOB形式将opinion标记在数据集第4列(polarity之后),如果您使用开源数据集标注,请不要直接修改原始数据集,将重新标注的数据集重新命名上传。 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)<#27 (comment)>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMGE7XZ5WWWWQRKFJDTRMY2RANCNFSM46AW7SCQ. 这个issue我先关闭了,后面沟通就gitter上吧。数据的标注可以参照https://github.com/lixin4ever/E2E-TBSA ― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMIHCZ2RZ3DAN7NHHTTROJKTANCNFSM46AW7SCQ.

我建了个Gitter。issue用来反馈bug和问题比较好,不适合讨论功能更新

LangDaoAI commented 3 years ago

好的

Get Outlook for Androidhttps://aka.ms/AAb9ysg


From: YangHeng @.> Sent: Monday, June 7, 2021 11:27:10 AM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27)

好的,晚上没抽出时间,gitter沟通是啥意思 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 11:41:29 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27) 晚上我标记一个例子,确认后再大规模标记 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 4:47:36 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27>) 数据我来提供,加入观点后标记规则是否要做修改 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Sunday, June 6, 2021 2:59:11 PM To: yangheng95/PyABSA @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/PyABSA] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 12号后我再用新的模型验证一下,这几天出差,另外观点抽出可能模型架构要改动的,有空还请看一下,或者说当前模型架构是否支持这样的观点抽出? Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: Tang Yong @.> Sent: Friday, June 4, 2021 7:18:13 PM To: yangheng95/pyabsa @.>; yangheng95/pyabsa @.> Cc: Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 好的感谢! Get Outlook for Androidhttps://aka.ms/AAb9ysg ____ From: YangHeng @.> Sent: Friday, June 4, 2021 7:17:27 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 合并的意思是把四个分散的数据集全部放到一个里面,只是单纯追加吗,可否详细说一下,谢谢 Get Outlook for Androidhttps://aka.ms/AAb9ysg … ____ From: YangHeng @.> Sent: Friday, June 4, 2021 6:38:12 PM To: yangheng95/pyabsa @.> Cc: LangDaoAI @.>; Author @.> Subject: Re: [yangheng95/pyabsa] 多观点的抽出是否支持? (#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27>) 本地替换了方法做了测试,如下图,切词没有问题, 但是BIOE标注(红框:颜色,aspect只抽取了颜)好像有点问题,请看一下, [image]https://user-images.githubusercontent.com/83859417/120736527-fbda2380-c51e-11eb-94b0-c30ecea42979.png 有空格和没有空格都是一样的。 Sentence with predicted labels: 尤(O) 其(O) 是(O) 照(O) 的(O) 大(O) 尺(O) 寸(O) 照(O) 片(O) 时(O) 效(B-ASP) 果(I-ASP) 也(O) 是(O) 非(O) 常(O) 不(O) 错(O) 的(O) {'aspect': '效 果', 'position': '11,12', 'sentiment': 'Good'} Sentence with predicted labels: 照(O) 大(O) 尺(O) 寸(O) 的(O) 照(O) 片(O) 的(O) 时(O) 候(O) 手(O) 机(O) 反(O) 映(O) 速(B-ASP) 度(I-ASP) 太(O) 慢(O) {'aspect': '速 度', 'position': '14,15', 'sentiment': 'Bad'} Sentence with predicted labels: 关(O) 键(O) 的(O) 时(O) 候(O) 需(O) 要(O) 表(O) 现(O) 持(O) 续(O) 影(O) 像(O) 的(O) 短(O) 片(O) 功(B-ASP) 能(I-ASP) 还(O) 是(O) 很(O) 有(O) 用(O) 的(O) {'aspect': '功 能', 'position': '16,17', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 相(O) 比(O) 较(O) 原(O) 系(O) 列(O) 锐(B-ASP) 度(I-ASP) 高(O) 了(O) 不(O) 少(O) 这(O) 一(O) 点(O) 好(O) 与(O) 不(O) 好(O) 大(O) 家(O) 有(O) 争(O) 议(O) {'aspect': '锐 度', 'position': '6,7', 'sentiment': 'Good'} Sentence with predicted labels: 这(O) 款(O) 手(O) 机(O) 的(O) 大(B-ASP) 小(I-ASP) 真(O) 的(O) 很(O) 薄(O) ,(O) 但(O) 是(O) 颜(B-ASP) 色(I-ASP) 不(O) 太(O) 好(O) 看(O) ,(O) 总(O) 体(O) 上(O) 我(O) 很(O) 满(O) 意(O) 啦(O) 。(O) {'aspect': '大 小', 'position': '5,6', 'sentiment': 'Bad'} {'aspect': '颜 色', 'position': '14,15', 'sentiment': 'Bad'} 这是综合4个中文数据集训练的模型的结果 颜色问题不知道是如何解决的,还是说模型做了更新还是数据量做了增强,很想知道, 另外, 新的代码构建的模型空了也帮忙上传一下, 我来做更新测试。 主要原因是将4个中文数据集合并了,数据量增多,代码已经更新,正在上传模型 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-854606848>>>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWNZFDTWARXWMBAVCNTTRCUJJANCNFSM46AW7SCQ. 就是合并所有的中文训练集和测试集作为一整个训练集和测试集,现在可以从google drive下载中文模型,读取模型之后会打印训练时的参数,其中的datasets_files就是采用的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27><#27https://github.com/yangheng95/PyABSA/issues/27#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-854627966>>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWPQDEJONX3FEGBOKIDTRCY4PANCNFSM46AW7SCQ. 改模型倒不是最大的问题, 问题是没有标注好opinion的数据集 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27<#27https://github.com/yangheng95/PyABSA/issues/27> (comment)<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-855350091>>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMACZNRK5NF6JXYGXDTRMMD7ANCNFSM46AW7SCQ. 使用IOB形式将opinion标记在数据集第4列(polarity之后),如果您使用开源数据集标注,请不要直接修改原始数据集,将重新标注的数据集重新命名上传。 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27https://github.com/yangheng95/PyABSA/issues/27 (comment)<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-855363169>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMGE7XZ5WWWWQRKFJDTRMY2RANCNFSM46AW7SCQ. 这个issue我先关闭了,后面沟通就gitter上吧。数据的标注可以参照https://github.com/lixin4ever/E2E-TBSA D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#27 (comment)https://github.com/yangheng95/PyABSA/issues/27#issuecomment-855418679>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWMIHCZ2RZ3DAN7NHHTTROJKTANCNFSM46AW7SCQ.

我建了个Gitter讨论组,https://gitter.im/PyABSA/community?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge。issue用来反馈bug和问题比较好,不适合讨论功能更新https://gitter.im/PyABSA/community?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge%E3%80%82issue%E7%94%A8%E6%9D%A5%E5%8F%8D%E9%A6%88bug%E5%92%8C%E9%97%AE%E9%A2%98%E6%AF%94%E8%BE%83%E5%A5%BD%EF%BC%8C%E4%B8%8D%E9%80%82%E5%90%88%E8%AE%A8%E8%AE%BA%E5%8A%9F%E8%83%BD%E6%9B%B4%E6%96%B0

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/yangheng95/PyABSA/issues/27#issuecomment-855548647, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AT7ZPWOKGI62ELY7FTJUXHDTRQ4A5ANCNFSM46AW7SCQ.