{"id":32687,"date":"2022-02-25T23:20:26","date_gmt":"2022-02-25T15:20:26","guid":{"rendered":"https:\/\/swarma.org\/?p=32687"},"modified":"2022-02-25T23:20:26","modified_gmt":"2022-02-25T15:20:26","slug":"%e5%a6%82%e4%bd%95%e6%8a%8a%e5%9b%a0%e6%9e%9c%e6%8e%a8%e6%96%ad%e5%92%8cml%e5%8f%8anlp%e7%bb%93%e5%90%88%ef%bc%9f-%e5%91%a8%e6%97%a5%e7%9b%b4%e6%92%ad%c2%b7%e5%9b%a0%e6%9e%9c%e7%a7%91%e5%ad%a6","status":"publish","type":"post","link":"https:\/\/swarma.org\/?p=32687","title":{"rendered":"\u5982\u4f55\u628a\u56e0\u679c\u63a8\u65ad\u548cML\u53caNLP\u7ed3\u5408\uff1f| \u5468\u65e5\u76f4\u64ad\u00b7\u56e0\u679c\u79d1\u5b66\u8bfb\u4e66\u4f1a"},"content":{"rendered":"
\n
\n
\n

<\/p>\n

\n
\n
\n
\n
\n
\n

<\/section>\n<\/section>\n
\n
\n

\u5bfc\u8bed<\/strong><\/span><\/p>\n<\/section>\n<\/section>\n

\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n
Judea Pearl\u8ba4\u4e3a\uff0c\u56e0\u679c\u63a8\u7406\u548c\u673a\u5668\u5b66\u4e60\u878d\u5408\u6784\u5efa\u5177\u6709\u81ea\u52a8\u5316\u63a8\u7406\u7684Causal AI\u7cfb\u7edf\uff0c\u662f\u5b9e\u73b0\u5f3a\u4eba\u5de5\u667a\u80fd\u7684\u5173\u952e\u6b65\u9aa4\u3002\u8fd1\u5e74\u6765\uff0c\u5b66\u8005\u9010\u6e10\u610f\u8bc6\u5230\u73b0\u6709\u7684\u6a21\u578b\u5176\u5b9e\u5e76\u6ca1\u6709\u5b66\u5230\u4eba\u7c7b\u505a\u540c\u7c7b\u5224\u65ad\u65f6\u7684\u56e0\u679c\u6027\u7684\u63a8\u65ad\u8fc7\u7a0b\u3002\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08NLP\uff09\u4f5c\u4e3a\u673a\u5668\u5b66\u4e60\u4e00\u5927\u91cd\u8981\u4efb\u52a1\u5982\u4f55\u4e0e\u56e0\u679c\u63a8\u7406\u7ed3\u5408\u4e5f\u9010\u6e10\u8fdb\u5165\u5b66\u8005\u7684\u89c6\u91ce\u3002\u672c\u5468\u8bfb\u4e66\u4f1a\u5c06\u9080\u8bf7\u9a6c\u666e\u6240\u548cETH\u7684\u8054\u5408\u535a\u58eb\u91d1\u81f4\u9759\u4ecb\u7ecd\u56e0\u679c\u63a8\u7406\u4e0e\u673a\u5668\u5b66\u4e60\u7ed3\u5408\u7684\u6f5c\u529b\u548c\u7814\u7a76\u65b0\u52a8\u6001\uff0c\u5e76\u63a2\u8ba8\u56e0\u679c\u6a21\u578b\u5177\u4f53\u5728NLP\u4e0a\u7684\u5e94\u7528\u573a\u666f\u548c\u672a\u6765\u65b9\u5411\u3002<\/span><\/strong><\/section>\n


<\/p>\n

\u7531\u667a\u6e90\u793e\u533a\u3001\u96c6\u667a\u4ff1\u4e50\u90e8\u8054\u5408\u4e3e\u529e\u7684<\/strong>\u56e0\u679c\u79d1\u5b66\u4e0eCausal AI\u8bfb\u4e66\u4f1a\u7b2c\u4e09\u5b63<\/strong><\/a>\uff0c<\/strong>\u5176\u76ee\u6807\u662f\uff1a\u5c06\u4e3b\u8981\u7ed3\u5408\u56e0\u679c\u63a8\u65ad\u7684\u6f5c\u5728\u7ed3\u679c\uff08Potential Outcomes\uff09\u6846\u67b6\u5c55\u5f00\u8ba8\u8bba\uff0c\u5e76\u5c1d\u8bd5\u5728\u4e0d\u540c\u9886\u57df\uff0c\u5305\u62ec\u4f46\u4e0d\u9650\u4e8e\u533b\u5b66\u3001\u7ecf\u6d4e\u5b66\u3001\u673a\u5668\u5b66\u4e60\u7b49\u9886\u57df\uff0c\u5c1d\u8bd5\u5efa\u7acb\u4e0eSCM\u7684\u5bf9\u5e94\u5173\u7cfb\u3002\u8bfb\u4e66\u81ea2021\u5e7410\u670824\u65e5\u5f00\u59cb\uff0c\u6bcf\u5468\u65e5\u4e0a\u5348 09:00-11:00\u4e3e\u529e\uff0c\u6301\u7eed\u65f6\u95f4\u9884\u8ba1 2-3 \u4e2a\u6708<\/strong>\u3002<\/strong><\/p>\n

<\/strong><\/strong><\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
<\/section>\n<\/section>\n

<\/section>\n
<\/section>\n

<\/section>\n

<\/section>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

\u80cc\u666f<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/p>\n

<\/h1>\n

\u56e0\u679c\u63a8\u65ad\u548c\u673a\u5668\u5b66\u4e60 (ML) \u8fd9\u4e24\u4e2a\u9886\u57df\u5728\u957f\u671f\u4ee5\u6765\u5404\u81ea\u53d1\u5c55\u548c\u58ee\u5927\u3002\u7136\u800c\uff0c\u968f\u7740ML\u7684\u53d1\u5c55\u8d8a\u6765\u8d8a\u5feb\uff0c\u9010\u6e10\u9047\u5230\u4e86\u7528\u5355\u7eaf\u7684\u7edf\u8ba1\u53bb\u6355\u6349\u76f8\u5173\u6027\u7684\u4e0d\u8db3\u4e4b\u5904\uff0c\u8d8a\u6765\u8d8a\u591a\u7684\u975eiid\u60c5\u51b5 (\u6bd4\u5982domain shift\uff0cadversarial attack\u7b49)\uff0c\u663e\u793a\u4e86ML\u6a21\u578b\u5176\u5b9e\u5e76\u6ca1\u6709\u5b66\u5230\u4eba\u7c7b\u505a\u540c\u7c7b\u5224\u65ad\u65f6\u7684\u56e0\u679c\u6027\u7684\u63a8\u65ad\u8fc7\u7a0b\u3002<\/span><\/p>\n


<\/span><\/p>\n

\u628a\u56e0\u679c\u63a8\u65ad\u7ed3\u5408\u8fdbML\uff0c\u662f\u628aML\u63a8\u52a8\u5230\u80fd\u89e3\u51b3\u975eiid\u95ee\u9898\u7684\u4e00\u4e2a\u91cd\u8981\u65b9\u5411\u3002\u540c\u65f6\uff0c\u8bb2\u8005\u4f1a\u7740\u91cd\u4ecb\u7ecd\u56e0\u679c\u65b9\u9762\u7684insights\u80fd\u591f\u600e\u4e48\u5e94\u7528\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406 (NLP) \u91cc\uff0c\u5305\u62ec\u5bf9NLP\u5404\u7c7b\u4efb\u52a1\u7684\u56e0\u679c\u6027\u5206\u6790\uff0c\u4ee5\u53ca\u5982\u4f55\u5e94\u7528causal effect estimation\u6765\u89e3\u51b3\u5177\u4f53NLP\u7684\u95ee\u9898\uff0c\u548c\u524d\u666f\u5c55\u671b\u3002<\/span><\/p>\n


<\/p>\n

<\/h1>\n

<\/section>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

\u5185\u5bb9\u7b80\u4ecb<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/p>\n

\u672c\u6b21\u6f14\u8bb2\u5c06\u57fa\u4e8eBernhard Schoelkopf\u548cYoshua Bengio\u5bf9\u56e0\u679c\u63a8\u65ad\u548cML\u7ed3\u5408\u7684\u84dd\u56fe \u201cTowards Causal Representation Learning\u201d \u548c \u201cCausality for Machine Learning\u201d \u51fa\u53d1\uff0c\u4ecb\u7ecd\u56e0\u679c\u63a8\u65ad\u548c\u673a\u5668\u5b66\u4e60\u7ed3\u5408\u7684\u6f5c\u529b\uff0c\u548c\u5177\u4f53\u7684\u7814\u7a76\u65b0\u52a8\u6001\u3002\u8bb2\u5ea7\u9664\u4e86\u4f1a\u4ecb\u7ecd\u57fa\u7840\u6027\u77e5\u8bc6\u4e4b\u5916\uff0c\u8fd8\u4f1a\u63a2\u8ba8\u56e0\u679c\u6a21\u578b\u5177\u4f53\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406 (NLP) \u4e0a\u7684\u5e94\u7528\u573a\u666f\u548c\u672a\u6765\u65b9\u5411\u3002<\/span><\/p>\n


<\/span><\/p>\n


<\/span><\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

\u4e3b\u8bb2\u4eba<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/p>\n

<\/p>\n

\u91d1\u81f4\u9759<\/span><\/strong>\uff0c\u9a6c\u666e\u6240\u548cETH\u7684\u8054\u5408\u535a\u58eb\uff0c\u4e3b\u5bfc\u5e08\u4e3a\u56e0\u679c\u63a8\u65ad\u9886\u519b\u5b66\u8005Bernhard Schoelkopf\u3002\u4e3b\u653bNLP + \u56e0\u679c\u63a8\u65ad\uff0c\u5df2\u53d1\u8868NLP\/AI\u8bba\u658721\u7bc7 (\u5305\u62ecACL, EMNLP, NAACL, AAAI, COLING, AISTATS\u7b49)\u3002\u4e3b\u8981\u5408\u4f5c\u8005\u548c\u5bfc\u5e08\u5305\u62ecProf Rada Mihalcea (\u5bc6\u6b47\u6839\u5927\u5b66), Prof Mrinmaya Sachan (ETH), Prof Ryan Cotterell (ETH)\u3002\u53e6\u6709\u4e3b\u529e\u4f17\u591a\u5173\u4e8eNLP+\u56e0\u679c\u63a8\u65ad\u7684\u4f1a\u8bae\u548c\u6d3b\u52a8\uff0c\u5305\u62ec1st conference on Causal Learning and Reasoning (CLeaR 2022)\uff0cRobustML workshop (ICLR 2021)\uff0cTutorial on CausalNLP (EMNLP 2022)\u3002\u66f4\u591a\u4fe1\u606f\u8be6\u89c1zhijing-jin.com<\/span><\/p>\n


<\/h1>\n

\u53c2\u8003\u6587\u732e<\/span><\/strong><\/span><\/h1>\n
    \n
  1. \n
    Towards Causal Representation Learning (Bernhard Sch\u00f6lkopf, Francesco Locatello, Stefan Bauer, Nan Rosemary Ke, Nal Kalchbrenner, Anirudh Goyal, Yoshua Bengio): https:\/\/arxiv.org\/abs\/2102.11107<\/span><\/section>\n<\/li>\n
  2. \n
    Causality for Machine Learning (Bernhard Sch\u00f6lkopf): https:\/\/arxiv.org\/abs\/1911.10500<\/span><\/section>\n<\/li>\n
  3. \n
    [EMNLP 2021 oral] Causal Direction of Data Collection Matters: Implications of Causal and Anticausal Learning for NLP (Zhijing Jin, Julius von K\u00fcgelgen, Jingwei Ni, Tejas Vaidhya, Ayush Kaushal, Mrinmaya Sachan, Bernhard Sch\u00f6lkopf): https:\/\/arxiv.org\/abs\/2110.03618<\/span><\/section>\n<\/li>\n
  4. \n
    [EMNLP 2021 Findings] Mining the Cause of Political Decision-Making from Social Media: A Case Study of COVID-19 Policies across the US States (Zhijing Jin, Zeyu Peng, Tejas Vaidhya, Bernhard Sch\u00f6lkopf, Rada Mihalcea): https:\/\/aclanthology.org\/2021.findings-emnlp.27\/<\/span><\/section>\n<\/li>\n
  5. \n
    Learning Logical Fallacies (2021) (Zhijing Jin, Abhinav Lalwani, Tejas Vaidhya, Xiaoyu Shen, Yiwen Ding, Zhiheng Lyu, Mrinmaya Sachan, Rada Mihalcea, and Bernhard Sch\u00f6lkopf): https:\/\/bit.ly\/logical-nlp<\/span><\/section>\n<\/li>\n
  6. \n
    Original or Translated? A Causal Analysis of the Impact of Translationese on Machine Translation Performance (2022) (Jingwei Ni, Zhijing Jin, Markus Freitag, Mrinmaya Sachan, Bernhard Sch\u00f6lkopf): https:\/\/drive.google.com\/file\/d\/143Q32czLplRt3uICBEzS7bWrOXaVnyN8\/<\/span><\/section>\n<\/li>\n
  7. \n
    Slangvolution: A Causal Analysis of Semantic Change and Frequency Dynamics in Slang (2021) (Daphna Keidar, Andreas Opedal, Zhijing Jin, Mrinmaya Sachan): https:\/\/drive.google.com\/file\/d\/1k6-7yluZV3Ndby8nqxYyubf_-n05LCv1\/<\/span><\/section>\n<\/li>\n<\/ol>\n

    <\/span><\/section>\n


    <\/span><\/p>\n

    \n
    \n
    \n
    \n
    \n

    <\/section>\n

    <\/section>\n
    \n
    <\/svg><\/section>\n<\/section>\n<\/section>\n
    \n

    \u76f4\u64ad\u4fe1\u606f<\/strong><\/strong><\/p>\n<\/section>\n

    \n

    <\/section>\n

    <\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


    <\/p>\n

    \u65f6\u95f4\uff1a<\/span><\/strong><\/p>\n

    2022\u5e742\u670827\u65e5\u665a\u4e0a20: 00-22: 00<\/span><\/p>\n


    <\/span><\/section>\n

    \u53c2\u4e0e\u65b9\u5f0f\uff1a<\/span><\/strong><\/p>\n