Replication Data for: CASM: A Deep-Learning Approach for Identifying Collective Action Events with Text and Image Data from Social Mediadoi:10.7910/DVN/SS4LNNHarvard Dataverse2019-06-251Pan, Jennifer; Zhang, Han, 2019, "Replication Data for: CASM: A Deep-Learning Approach for Identifying Collective Action Events with Text and Image Data from Social Media", https://doi.org/10.7910/DVN/SS4LNN, Harvard Dataverse, V1, UNF:6:26KE5u8/rqgZAoNS8X8wXg== [fileUNF]Replication Data for: CASM: A Deep-Learning Approach for Identifying Collective Action Events with Text and Image Data from Social Mediadoi:10.7910/DVN/SS4LNNPan, JenniferZhang, HanHarvard DataversePan, JenniferZhang, PanPan, Jennifer2019-06-25Social Sciencescollective action, deep learning, event data, social media, ChinaProtest event analysis is an important method for the study of collective action and social movements and typically draws on traditional media reports as the data source. We introduce collective action from social media (CASM)—a system that uses convolutional neural networks on image data and recurrent neural networks with long short-term memory on text data in a two-stage classifier to identify social media posts about offline collective action. We implement CASM on Chinese social media data and identify more than 100,000 collective action events from 2010 to 2017 (CASM-China). We evaluate the performance of CASM through cross-validation, out-of-sample validation, and comparisons with other protest data sets. We assess the effect of online censorship and find it does not substantially limit our identification of events. Compared to other protest data sets, CASM-China identifies relatively more rural, land-related protests and relatively few collective action events related to ethnic and religious conflict.We recommend you view files in Tree structure and begin with the readme.txt<a href="http://creativecommons.org/publicdomain/zero/1.0">CC0 1.0</a>Zhang Han and Jennifer Pan. 2019. “CASM: A Deep-Learning Approach for Identifying Collective Action Events with Text and Image Data from Social Media” Sociological Methodology 49: 1-59.10.1177/0081175019860244Zhang Han and Jennifer Pan. 2019. “CASM: A Deep-Learning Approach for Identifying Collective Action Events with Text and Image Data from Social Media” Sociological Methodology 49: 1-59.protest_events.tab1363301text/tab-separated-valuesUNF:6:wbuMQn7wm4tl11Io3al6oQ==keyword_search_placebo_count.tab1801text/tab-separated-valuesUNF:6:ZxN0FDr+XlhoWkcNZctu9w==Also used for plot/SuppFigure7firststage_c1_text_image_precision_recall.tab60463text/tab-separated-valuesUNF:6:hxAPly5qKIGA1G6HfLHv1Q==c1c2.tab24083text/tab-separated-valuesUNF:6:rL3sAGsTp2RrS1yDdkgrcQ==c1c2_precision_recall_crossvalidation.tab64133text/tab-separated-valuesUNF:6:mkyvOgenEhyzR7efni3q8Q==secondstage_c2_text_image_precision_recall.tab38534text/tab-separated-valuesUNF:6:NKAiaoqeTkNJgn5gbpy7Cg==keyword_vs_num_event_700000_fig4.tab45003text/tab-separated-valuesUNF:6:BOr7j3rkVv5QTcQtfK23Tw==precision_vs_keyword.tab25603text/tab-separated-valuesUNF:6:qHEFiyMlqC8j0XwBEChMJQ==firststage_c1_SVM_NB.tab60383text/tab-separated-valuesUNF:6:VueXQuSkwhSwntSBqd6gWg==event_id forms issuesUNF:6:wbuMQn7wm4tl11Io3al6oQ==month count_all month_numeric quarter Dataset yearUNF:6:ZxN0FDr+XlhoWkcNZctu9w==labelUNF:6:/WKB02PwmgYsUaAvUUl4WA==precision6046.00.3928149639659638.0.00.31799858101135190.169027845627747921.00.21272211487785242UNF:6:HMKRgUbpYsQmCO9gH/xqng==recall0.01.00.95953757225433520.24434564628902133.0.83168779243526030.0028901734104046246046.0UNF:6:XkzU2089TYovplWPOCq1sg==labelUNF:6:VbmLr5BCRwmU99POlnD+WQ==precision2408.00.198430023818076670.65617841329540820.38625204582651390.00.6143394726876077.1.0UNF:6:ePL9lu2FbYZ9gjX7w8j01A==recall0.81779661016949160.002118644067796610.02408.01.00.2784397034783831.0.7100924179289374UNF:6:uiEp/74k1CTmxZPLsd+HmA==labelUNF:6:fPFJY05MN31qzjjPmzkzpA==precision0.93731685565040136413.00.35019241341396373.0.20563193790730970.00.81417759296936851.0UNF:6:bBv3mtaY3MeaqBDnvSY4VA==recall0.99966592427616930.72204899777282850.31650989626401830.0015082956259426848.0.00.62087280059567526413.0UNF:6:K8kuyfoC2yYRo65HurFLlg==labelUNF:6:RvEjpk9rT+6KKdRzq5zt9Q==precision1.00.6848123751991867.0.190640820141634913853.00.39656652360515020.66472303206997090.0UNF:6:c3Q63cuB0AIvGzLYCeTvOg==recall1.0.0.81376518218623480.70297714447049740.281927102387562250.00215982721382289430.03853.0UNF:6:4Fcvgiujxg8/SJElDdlYzg==threshold1.09190918688468130.77051880975919680.00.81481574743221973853.0.0.204374435828867830.16134203970432281UNF:6:9T2CstOvsCNfzAarpKYidA==monthUNF:6:O/XMkrySXj7/stLtO+oGrw==nunique3126.01306.6455555555684500.00.02.0.1437.0775.2087042562923UNF:6:IRufL0DUEM+jNmdzQo6MJg==rank4500.0.1.014.43247338691515625.550.00.025.5UNF:6:ikVCsyzjacSJi7zCXrf62Q==precision0.89241124488594610.12220860158924130.98310810810810810.93821567488809410.24090203404994343.0.02560.0UNF:6:5yicMrkeNYov1JxVvD9UEw==rank10.00.014.144898569077962560.030.0.50.030.0UNF:6:19Zex4EAgUSxVumQ8R8b6w==recall1.00.65704015955187780.212254159398533370.031785909339158932560.00.6223105027688090.0.UNF:6:rYMxh1EOHDUGy5TlCF94cQ==labelUNF:6:TWro7UTlIZoER+ZhoImqnw==precision0.01.00.0.0.32414329091960670.36985653121232636038.00.16754473718512444UNF:6:tBzZODFBTMf0oJ7mvsPsSw==recall0.9768786127167631.00.00.83096171262160560.00.261592059414755336038.0.UNF:6:I3gaMv8h51OylDenGlcv2g==readme.txttext/plainrequirements.txttext/plainCASM_c1_deep_text.pytext/x-python-scriptCASM_c2_deep_text.pytext/x-python-scriptCASM_generate_predicted_probability_image.pytext/x-python-scriptCASM_generate_predicted_probability_text.pytext/x-python-scriptcommon_operations.pytext/x-python-scriptdependency.pytext/x-python-scriptLSTM_text_dependency.pytext/x-python-scriptword_preprocessing.pytext/x-python-scriptimage.jsonapplication/jsontext-stage1.jsonapplication/jsontext-stage2.jsonapplication/jsonweights_image.h5application/x-h5weights_text-stage1.h5application/x-h5weights_text-stage2.hdf5application/x-hdf5protest_posts.csvtext/csvprotest_region_count_prefecture_log_scale.Rtype/x-r-syntaxCN-shi-A.dbfapplication/dbfCN-shi-A.prjapplication/prjCN-shi-A.sbnapplication/sbnCN-shi-A.sbxapplication/sbxCN-shi-A.shpapplication/shpCN-shi-A.shxapplication/shxCASM_divide_irrelevant.Rtype/x-r-syntaxCASM_Wickedonna.Rtype/x-r-syntaxwicked_events_2018-11-29.csvtext/csvfirststage_c1_text_image_precision_recall.pdfapplication/pdffirststage_c1_text_image_precision_recall.rtype/x-r-syntaxc1c2_vs_c1_precision_recall.pdfapplication/pdfc1c2_vs_c1_precision_recall.rtype/x-r-syntaxc1c2_vs_cross_validation_precision_recall.pdfapplication/pdfc1c2_vs_cross_validation_precision_recall.rtype/x-r-syntaxkeywordCoverageSet.txttext/plainkeyword_coverage_wickedonna.pdfapplication/pdfsupfig1.rtype/x-r-syntaxkeywordCoverageSet_keyword_year.txttext/plainkeyword_coverage_wickedonna_by_year.pdfapplication/pdfsupfig2.rtype/x-r-syntaxkeywordCoverageSet_keyword_prov.txttext/plainkeyword_coverage_wickedonna_by_province.pdfapplication/pdfsupfig3.rtype/x-r-syntaxkeyword_vs_num_event_700000.pdfapplication/pdfsupfig4.rtype/x-r-syntaxprecision_vs_keyword.pdfapplication/pdfsupfig5.rtype/x-r-syntaxfirststage_c1_SVM_NB_precision_recall.pdfapplication/pdfsupfig6.rtype/x-r-syntaxCASM_keyword_irrelevant_count.pdfapplication/pdfsupfig7.Rtype/x-r-syntaxhigh_frequency_protest_words.txttext/plainjieba.dict.big.txttext/plainpropaganda_media_words.txttext/plainstopwords1.txttext/plainvillage_level_dict.txttext/plainvillage_level_dict_reversed.txttext/plainvocab_pos_grievance.dictapplication/octet-streamvocab_pos_KGP_50000.dictapplication/octet-stream