2.4.1.3 Censorship ntawm kev tawm los ntawm lub tsoom fwv Suav

Soj ntsuam scraped Suav social xov xwm qhov chaw kawm censorship. Lawv hais txog nrog incompleteness nrog latent-kev zoo losis phem inference.

Nyob rau hauv tas li ntawd mus rau lub loj cov ntaub ntawv siv nyob rau hauv ob yav dhau los piv txwv, kev soj ntsuam kuj yuav sau lawv tus kheej observational cov ntaub ntawv, raws li tau wonderfully taw qhia los ntawm Gary King, Jennifer Yias, thiab Molly Roberts ' (2013) kev tshawb fawb ntawm censorship los ntawm tsoom fwv Suav.

Social xov xwm posts nyob rau hauv Tuam Tshoj yog censored los ntawm ib tug enormous lub xeev apparatus uas yog xav kom muaj xws li kaum tawm txhiab tus neeg. Soj ntsuam thiab cov pej xeem, li cas los xij, muaj me ntsis kev txiav txim zoo li cas cov censors txiav txim siab li cas cov ntsiab lus yuav tsum raug muab tshem ntawm cov kev sib raug zoo xov xwm. Kws tshawb fawb ntawm Tuam Tshoj yeej muaj kev tsis sib haum kev ntshaw txog uas hom posts yog feem ntau yuav tau txais kev deleted. Ib txhia xav tias censors siab rau posts uas yog cov tseem ceeb ntawm lub xeev thaum lwm tus neeg xav hais tias lawv tsom rau posts uas txhawb kom tibneeg tus coojpwm, xws li tawm tsam. Khom lub tsheb tawm uas ntawm cov kev cia siab yog muaj tseeb muaj qhov cuam tshuam rau yuav ua li cas soj ntsuam to taub Tuam Tshoj thiab lwm yam authoritarian tsoom fwv uas koom nyob rau hauv kev tswj kav. Yog li ntawd, King thiab lug txhawb cov miv xav los sib piv posts uas raug luam tawm thiab tom qab deleted rau posts uas raug luam tawm thiab yeej tsis deleted.

Sau cov posts koom tes nrog cov amazing engineering feat ntawm nkag ntau tshaj 1,000 Suav kev tawm websites-txhua nrog txawv nplooj ntawv layouts-nrhiav cov posts, thiab ces revisiting cov posts saib uas tau tom qab deleted. Nyob rau hauv tas li ntawd mus rau lub qub engineering teeb meem txuam nrog rau loj scale web-nkag, qhov no peb tes num muaj lub ntxiv kev sib tw hais tias nws yuav tsum tau mus yuav tsis tshua muaj ceev ceev vim hais tias muaj ntau yam censored posts yog npaum li cas nyob rau hauv tsawg tshaj li 24 teev. Nyob rau hauv lwm yam lus, ib tug qeeb crawler yuav nco ntau ntau posts uas tau censored. Ntxiv mus, lub crawlers yuav tsum tau ua tag nrho cov no cov ntaub ntawv sau thaum evading nrhiav kom tau tsam kev tawm websites thaiv kev nkag tau mus los yog lwm yam hloov lawv tej kev cai nyob rau hauv teb mus rau txoj kev tshawb no.

Thaum no loj heev engineering ua hauj lwm twb tiav, King thiab lug txhawb cov miv twb tau txais txog 11 lab posts on 85 txawv tej yam uas twb pre-teev raws li lawv tsim theem ntawm rhiab heev. Piv txwv li, ib lub npe ntawm siab rhiab heev yog Ai Weiwei, lub dissident artist; ib lub npe ntawm nruab nrab rhiab heev yog txaus siab thiab devaluation ntawm Suav txiaj, thiab ib lub npe ntawm tsawg rhiab heev yog lub ntiaj teb khob. Ntawm cov 11 lab posts txog 2 lab tau censored, tab sis posts on rhiab heev txog tej yam tau censored tsuas yog me ntsis ntau dua posts on nruab nrab thiab tsawg rhiab heev yam. Nyob rau hauv lwm yam lus, suav censors yog txog li yuav mus censor ib tug ncej uas hais Ai Weiwei raws li ib tug ncej uas hais lub ntiaj teb khob. Cov kev tshawb nrhiav tsis tau phim lub simplistic lub tswv yim hais tias tsoom fwv censors tag nrho cov posts nyob rau hauv rhiab yam.

Qhov no yooj yim muab xam ntawm censorship tus nqi los ntawm lub ntsiab lus yuav yog yuam kev, txawm li cas los. Piv txwv li, tsoom fwv tej zaum yuav censor posts uas txhawb nqa Ai Weiwei, tab sis tawm posts uas yog cov tseem ceeb ntawm nws. Nyob rau hauv thiaj li yuav paub qhov txawv ntawm posts ntau kom zoo zoo, cov kev soj ntsuam yuav tsum ntsuas lub sentiment ntawm txhua post. Yog li, ib txoj kev xav txog nws yog hais tias lub sentiment ntawm txhua post nyob rau hauv ib qho tseem ceeb latent feature ntawm txhua post. Tu siab, txawm npaum li cas ua hauj lwm, tag nrho cia li txoj kev sentiment nrhiav kom tau siv pre-uas twb muaj lawm dictionaries yog tseem tsis zoo heev nyob rau hauv ntau zaus (xav hais tias rov qab mus rau cov teeb meem tsim muaj ib qho kev xav ncua sij hawm ntawm lub Cuaj hlis 11, 2001 los ntawm Section 2.3.2.6). Yog li ntawd, King thiab lug txhawb cov miv yuav tsum tau ib txoj kev rau daim ntawv lo lawv 11 lab kev tawm posts raws li seb lawv 1) ib qho tseem ceeb ntawm lub xeev, 2) txhawb ntawm lub xeev, los yog 3) sob los yog tseeb lus ceeb toom txog cov txheej xwm. Qhov no suab zoo li ib tug loj heev txoj hauj lwm, tab sis lawv solved nws siv ib tug haib ua kom yuam kev; ib tug uas yog ib qho nyob rau hauv ntaub ntawv kev kawm tab sis tam sim no kuj tsis tshua muaj nyob rau hauv social science.

Ua ntej, nyob rau hauv ib kauj ruam feem ntau hu ua pre-ua, tus neeg tshawb fawb hloov dua siab tshiab rau hauv lub social xov xwm posts mus rau hauv ib daim ntawv-lub sij hawm matrix, nyob qhov twg muaj ib tug kab rau txhua daim ntawv thiab ib kem uas kaw seb tus ncej muaj ib lo lus (piv txwv li, kev tawm tsam, tsheb, thiab lwm yam). Tom ntej no, ib tug pab pawg neeg ntawm kev tshawb fawb pab txhais tes xuas lub sentiment ntawm ib tug qauv ntawm ncej. Ces, King thiab lug txhawb cov miv siv no tes xuas ntaub ntawv mus rau kwv yees ib tug tshuab kev kawm cov qauv uas yuav xav hais tias lub sentiment ntawm ib tug ncej raws li nws tus yam ntxwv. Thaum kawg, lawv siv qhov no tshuab kev kawm qauv los laij rau tus sentiment ntawm tag nrho cov 11 lab posts. Yog li, es tsis manually nyeem ntawv thiab labeling 11 lab posts (uas yuav tau yog ib qho uas tsis yooj yim sua), lawv manually sau tias ib tug me me muaj pes tsawg tus ntawm cov posts thiab ces siv li cas cov ntaub ntawv zaum yuav hu rau saib xyuas kev kawm los laij rau tus pawg ntawm tag nrho cov posts. Tom qab ua tsab ntawv no tsom xam, King thiab lug txhawb cov miv twb tau los xaus hais tias, dog dig kuj ceeb tias, qhov tshwm sim rau ib tug ncej raug deleted yog txheeb ze rau seb nws yog ib qho tseem ceeb ntawm lub xeev los yog txhawb ntawm lub xeev.

Daim duab 2.3: Yooj yim zog schematic rau txoj kev siv nyob rau hauv King, yias, thiab Roberts (2013) rau kwv yees cov sentiment ntawm 11 lab Suav kev tawm posts. Ua ntej, nyob rau hauv ib kauj ruam feem ntau hu ua pre-ua, tus neeg tshawb fawb hloov dua siab tshiab rau hauv lub social xov xwm posts mus rau hauv ib daim ntawv-lub sij hawm matrix (saib Grimmer thiab Stewart (2013) kom paub ntau ntxiv). Thib ob, cov kev soj ntsuam ntawm tes-coded lub sentiment ntawm ib tug me me qauv ntawm posts. Peb, tus neeg tshawb fawb kev kawm ib saib xyuas kev kawm cov qauv los cais lub sentiment ntawm posts. Plaub, kev soj ntsuam siv cov saib xyuas kev kawm cov qauv los laij rau tus sentiment ntawm tag nrho cov posts. Saib King, yias, thiab Roberts (2013), Appendix B rau ib tug xav paub ntxiv.

Daim duab 2.3: Yooj yim zog schematic rau txoj kev siv nyob rau hauv King, Pan, and Roberts (2013) rau kwv yees cov sentiment ntawm 11 lab Suav kev tawm posts. Ua ntej, nyob rau hauv ib kauj ruam feem ntau hu ua pre-ua, tus neeg tshawb fawb hloov dua siab tshiab rau hauv lub social xov xwm posts mus rau hauv ib daim ntawv-lub sij hawm matrix (saib Grimmer and Stewart (2013) kom paub ntau ntxiv). Thib ob, cov kev soj ntsuam ntawm tes-coded lub sentiment ntawm ib tug me me qauv ntawm posts. Peb, tus neeg tshawb fawb kev kawm ib saib xyuas kev kawm cov qauv los cais lub sentiment ntawm posts. Plaub, kev soj ntsuam siv cov saib xyuas kev kawm cov qauv los laij rau tus sentiment ntawm tag nrho cov posts. Saib King, Pan, and Roberts (2013) , Appendix B rau ib tug xav paub ntxiv.

Nyob rau hauv lub kawg, King thiab lug txhawb cov miv sab tias tsuas yog peb hom posts tau tsis tu ncua censored: saib duab liab qab, kev thuam ntawm censors, thiab cov neeg uas muaj collective txiav txim tej zaum (ie, tau ua kom loj-scale tawm tsam). Los tsom kwm ib tug lossis loj tus naj npawb ntawm cov posts uas tau deleted thiab posts uas twb tsis tshem, King thiab lug txhawb cov miv twb tau mus kawm yuav ua li cas lub censors ua hauj lwm cia li los saib thiab suav. Nyob rau hauv tom ntej kev tshawb fawb, lawv tau ncaj qha intervened rau hauv Suav cov kev tawm ecosystem los ntawm kev tsim posts nrog systematically txawv cov ntsiab lus thiab xab uas tau censored (King, Pan, and Roberts 2014) . Peb yuav kawm ntxiv txog kev sim le caag nyob rau hauv Tshooj 4. Ntxiv mus, foreshadowing ib lub ntsiab uas yuav tshwm sim thoob plaws phau ntawv, cov latent-attribute inference teeb meem-uas yuav tej zaum yuav daws tau nrog saib xyuas kev kawm-tig mus ua heev heev nyob rau hauv kev sib raug zoo kev tshawb fawb nyob rau hauv lub cov muaj hnub nyoog. Koj yuav pom cov duab zoo heev rau daim duab 2.3 nyob rau hauv tshooj 3 (Nug cov lus nug) thiab 5 (Tsim loj sib koom tes); nws yog ib tug ntawm ob peb lub tswv yim uas zoo nkaus li nyob rau hauv ntau yam chapters.

Tag nrho peb ntawm cov piv txwv-cov ua hauj lwm tus cwj pwm ntawm tsheb tavxij tsav tsheb nyob rau hauv New York, kev phooj ywg tsim los ntawm cov tub ntxhais kawm, thiab kev sib raug zoo xov xwm censorship cwj pwm ntawm tsoom fwv Suav-qhia tias kuj yooj yim suav ntawm observational cov ntaub ntawv yuav pab kom soj ntsuam los mus ntsuam xyuas theoretical txoj hau kev. Nyob rau hauv tej rooj plaub, loj cov ntaub ntawv enables koj yuav tau ua qhov no suav kuj ncaj qha (raws li nyob rau hauv cov ntaub ntawv ntawm New York Taxis). Nyob rau hauv lwm tus neeg mob, kev soj ntsuam yuav tsum tau mus sau lawv tus kheej observational cov ntaub ntawv (raws li nyob rau hauv cov ntaub ntawv ntawm Suav censorship); nrog incompleteness los ntawm cov phiajcim cov ntaub ntawv ua ke (xws li nyob rau hauv cov ntaub ntawv ntawm network evolution); los yog kev ua tau zoo ib co daim ntawv ntawm latent-kev zoo losis phem inference (raws li nyob rau hauv cov ntaub ntawv ntawm Suav censorship). Raws li kuv vam thiab cia siab no piv txwv qhia, rau kev soj ntsuam uas yog tau hais nthuav lus nug, loj tuas cog lus zoo heev.