Matches in SemOpenAlex for { <https://semopenalex.org/work/W3204920203> ?p ?o ?g. }
- W3204920203 abstract "The principle of independent causal mechanisms (ICM) states that generative processes of real world data consist of independent modules which do not influence or inform each other. While this idea has led to fruitful developments in the field of causal inference, it is not widely-known in the NLP community. In this work, we argue that the causal direction of the data collection process bears nontrivial implications that can explain a number of published NLP findings, such as differences in semi-supervised learning (SSL) and domain adaptation (DA) performance across different settings. We categorize common NLP tasks according to their causal direction and empirically assay the validity of the ICM principle for text data using minimum description length. We conduct an extensive meta-analysis of over 100 published SSL and 30 DA studies, and find that the results are consistent with our expectations based on causal insights. This work presents the first attempt to analyze the ICM principle in NLP, and provides constructive suggestions for future modeling choices. Code available at https://github.com/zhijing-jin/icm4nlp" @default.
- W3204920203 created "2021-10-11" @default.
- W3204920203 creator A5002316432 @default.
- W3204920203 creator A5016724158 @default.
- W3204920203 creator A5024604043 @default.
- W3204920203 creator A5044005697 @default.
- W3204920203 creator A5047867068 @default.
- W3204920203 creator A5074163178 @default.
- W3204920203 creator A5086287200 @default.
- W3204920203 date "2021-01-01" @default.
- W3204920203 modified "2023-10-18" @default.
- W3204920203 title "Causal Direction of Data Collection Matters: Implications of Causal and Anticausal Learning for NLP" @default.
- W3204920203 cites W1089141 @default.
- W3204920203 cites W1482778976 @default.
- W3204920203 cites W1493730910 @default.
- W3204920203 cites W1597032530 @default.
- W3204920203 cites W1638203394 @default.
- W3204920203 cites W1811279891 @default.
- W3204920203 cites W1915251500 @default.
- W3204920203 cites W1978341344 @default.
- W3204920203 cites W2002586403 @default.
- W3204920203 cites W2005097301 @default.
- W3204920203 cites W2012012028 @default.
- W3204920203 cites W2025768430 @default.
- W3204920203 cites W2034368206 @default.
- W3204920203 cites W2037603696 @default.
- W3204920203 cites W2048679005 @default.
- W3204920203 cites W2062989416 @default.
- W3204920203 cites W2083689856 @default.
- W3204920203 cites W2095844573 @default.
- W3204920203 cites W2099153428 @default.
- W3204920203 cites W2101460669 @default.
- W3204920203 cites W2107008379 @default.
- W3204920203 cites W2132119275 @default.
- W3204920203 cites W2133556223 @default.
- W3204920203 cites W2135625884 @default.
- W3204920203 cites W2144020560 @default.
- W3204920203 cites W2145837098 @default.
- W3204920203 cites W2147881172 @default.
- W3204920203 cites W2153903004 @default.
- W3204920203 cites W2153929442 @default.
- W3204920203 cites W2157331557 @default.
- W3204920203 cites W2172268343 @default.
- W3204920203 cites W22168010 @default.
- W3204920203 cites W2238429470 @default.
- W3204920203 cites W2251358367 @default.
- W3204920203 cites W2252272516 @default.
- W3204920203 cites W235810693 @default.
- W3204920203 cites W2463241543 @default.
- W3204920203 cites W2486125749 @default.
- W3204920203 cites W2513458147 @default.
- W3204920203 cites W2561274697 @default.
- W3204920203 cites W2739858629 @default.
- W3204920203 cites W2773436024 @default.
- W3204920203 cites W2787244810 @default.
- W3204920203 cites W2801890059 @default.
- W3204920203 cites W2804017338 @default.
- W3204920203 cites W2912924812 @default.
- W3204920203 cites W2950339735 @default.
- W3204920203 cites W2953291403 @default.
- W3204920203 cites W2963101081 @default.
- W3204920203 cites W2963403868 @default.
- W3204920203 cites W2963748441 @default.
- W3204920203 cites W2964308564 @default.
- W3204920203 cites W2970038984 @default.
- W3204920203 cites W2985165968 @default.
- W3204920203 cites W3034716087 @default.
- W3204920203 cites W3035013535 @default.
- W3204920203 cites W3035083705 @default.
- W3204920203 cites W3043997204 @default.
- W3204920203 cites W3089984751 @default.
- W3204920203 cites W3091670682 @default.
- W3204920203 cites W3098824823 @default.
- W3204920203 cites W3099396524 @default.
- W3204920203 cites W3104976898 @default.
- W3204920203 cites W3111757881 @default.
- W3204920203 cites W3114610051 @default.
- W3204920203 cites W3137687154 @default.
- W3204920203 cites W3154779166 @default.
- W3204920203 cites W3157753698 @default.
- W3204920203 cites W3196375010 @default.
- W3204920203 cites W3212480953 @default.
- W3204920203 cites W879220392 @default.
- W3204920203 cites W2922417627 @default.
- W3204920203 doi "https://doi.org/10.18653/v1/2021.emnlp-main.748" @default.
- W3204920203 hasPublicationYear "2021" @default.
- W3204920203 type Work @default.
- W3204920203 sameAs 3204920203 @default.
- W3204920203 citedByCount "3" @default.
- W3204920203 countsByYear W32049202032022 @default.
- W3204920203 countsByYear W32049202032023 @default.
- W3204920203 crossrefType "proceedings-article" @default.
- W3204920203 hasAuthorship W3204920203A5002316432 @default.
- W3204920203 hasAuthorship W3204920203A5016724158 @default.
- W3204920203 hasAuthorship W3204920203A5024604043 @default.
- W3204920203 hasAuthorship W3204920203A5044005697 @default.
- W3204920203 hasAuthorship W3204920203A5047867068 @default.
- W3204920203 hasAuthorship W3204920203A5074163178 @default.
- W3204920203 hasAuthorship W3204920203A5086287200 @default.
- W3204920203 hasBestOaLocation W32049202031 @default.