5.2.2 Taro-coding siyasa manifestos

Coding siyasa manifestos, wani abu yawanci yi da masana, za a iya yi da wani mutum ƙidãyar aikin sakamakon mafi reproducibility da kuma sassauci.

Similar to Galaxy Zoo, akwai mutane da yawa yanayi inda zamantakewa bincike so code, rarraba, ko Label wani image ko yanki na rubutu. Wani misali na irin wannan bincike ne coding siyasa manifestos. A lokacin zaben, jam'iyyun siyasa nuna manifestos kwatanta su da manufofin matsayi da shiryarwa falsafancin. Alal misali, a nan ne wani yanki na manifesto da Labor Party a Great Britain daga 2010:

"Miliyoyin mutanen da aiki a cikin jama'a da sabis embody mafi kyau dabi'u na Birtaniya, taimaka karfafawa mutane su sa mafi yawan da rãyukansu, alhãli kuwa kare su daga kasada ya kamata su ba su da kai a kan nasu. Kamar yadda muke bukatar mu zama bolder game da rawar da gwamnati a yin kasuwanni aiki fairly, muna kuma bukatar ya zama m gyãrãwa na gwamnati. "

Wadannan manifestos dauke da muhimmanci data siyasa masana kimiyya, musamman wadanda nazarin zaben da ƙarfafa muhimmancin manufofin muhawara. Domin tsare cire bayanai daga wadannan manifestos, masu bincike halitta The Shelar Project , wanda shirya siyasa masana kimiyya to Code 4,000 manifestos daga kusan 1,000 jam'iyyun a 50 ƙasashe. Kowane magana a kowace manifesto da aka masu lamba da gwani ta amfani da 56-category makirci. A sakamakon wannan ha] in} o} arin shi ne m dataset rungumar bayanai saka a cikin wadannan manifestos, kuma wannan dataset da aka yi amfani a cikin fiye da 200 kimiyya takardunku.

Kenneth Benoit da kuma abokan aiki (2015) ya yanke shawarar ya dauki manifesto coding aiki wanda ya rigaya an yi da masana da kuma juya shi a cikin wani ɗan adam ƙidãyar aikin. A sakamakon haka, sun halitta a coding tsari shi ne mafi reproducible kuma mafi m, ba a ma maganar mai rahusa da kuma sauri.

Aiki tare da 18 manifestos generated lokacin shida 'yan zaben a Birtaniya, Benoit da kuma abokan aiki amfani da tsaga-Aiwatar-hada dabarun da ma'aikata daga wani micro-aiki aiki kasuwa (Amazon Mechanical Turk da CrowdFlower ne misalai na micro-aiki aiki kasuwanni, gama more a micro-aiki aiki kasuwanni, ka duba babi na 4). The masu bincike ya kowane manifesto kuma raba shi a cikin sentences. Next, mutum rating aka amfani da su a kowace jumla. Musamman ma, idan magana da hannu a siyasa sanarwa, an lamba tare biyu girma: m (daga hagu zuwa sosai sosai dama) da kuma zamantakewa (daga m to ra'ayin mazan jiya) (Figure 5.5). Kowane magana da aka masu lamba ta game da 5 mutane daban-daban. A karshe, wadannan ratings aka hade ta amfani da ilimin kididdiga model cewa lissafta biyu mutum rater effects da wahala da magana effects. A duk, Benoit da kuma abokan aiki tattara 200,000 ratings daga game da 1,500 ma'aikata.

Figure 5.5: coding makirci daga Benoit et al. (2015) (Fig 1).

Domin tantance ingancin da taron coding, Benoit da kuma abokan aiki kuma na da game da 10 masana-furofesoshi, kuma digiri na biyu dalibai a Kimiyyar Siyasa-rate guda manifestos ta amfani da irin wannan hanya. Ko da yake ratings daga mambobi ne na taron kasance mafi m fiye da ratings daga masana, da yarjejeniya taron rating da gagarumar yarjejeniyar da yarjejeniya gwani rating (Figure 5.6). Wannan gwadawa ya nuna cewa, matsayin da Galaxy Zoo, mutum ƙidãyar ayyukan iya samar da high quality results.

Figure 5.6: Gwani kimomi (x-axis) da kuma taron kimomi (y-axis) sun kasance a ƙwarai yarjejeniya a lokacin da coding 18 jam'iyyar manifestos daga Great Britain (Benoit et al. 2015) . The manifestos masu lamba suna daga uku jam'iyyun siyasa (Conservative, Labour, Liberal Democrats) da kuma shida zaben (1987, 1992, 1997, 2001, 2005, 2010).

Gina a kan wannan sakamakon haka, Benoit da kuma abokan aiki amfani da taron-coding tsarin yin bincike cewa ya yiwuwa da Shelar Project. Alal misali, Shelar Project bai code da manifestos a kan topic na shige da fice domin cewa ba a salient topic a lõkacin da coding makirci da aka ɓullo da a cikin tsakiyar 1980s. Kuma, a wannan lokaci, shi ne logistically infeasible ga Shelar Project koma da sake code su manifestos kama da wannan bayani. Saboda haka, zai bayyana cewa masu bincike sha'awar nazarin siyasa na shige da fice ne daga luck. Duk da haka, Benoit da kuma abokan aiki sun iya amfani da mutum ƙidãyar tsarin yi haka coding-musamman ga bincike tambaya-sauri da kuma sauƙi.

Domin nazarin shige da fice siyasa, su masu lamba da manifestos takwas jam'iyyun a 2010 zaben a Great Britain. Kowane magana a kowace manifesto da aka masu lamba ya ga ko shi da alaka da shige da fice, kuma idan haka ne, ko shi pro-shige da fice, tsaka tsaki, ko anti-shige da fice. A cikin 5 hours of ƙaddamar da aikin, sakamakon sun kasance a. Sun tattara fiye da 22,000 martani a total kudin $ 360. Bugu da ari, kimomi daga taron nuna ƙwarai yarjejeniya tare da wani a baya binciken masana. Sa'an nan kuma, kamar yadda a karshe gwajin, watanni biyu daga baya, da masu bincike reproduced su taron-coding. A cikin 'yan sa'o'i, suka halitta sabon taron-masu lamba dataset cewa a hankali dace da ainihin taron-masu lamba data sa. A wasu kalmomin, mutum ƙidãyar sa su don samar da coding siyasa texts cewa amince da gwani kimantawa kuma ya reproducible. Bugu da ari, domin mutum ƙidãyar ya sauri da kuma cheap, shi ne mai sauƙi a gare su, su siffanta su data tarin ga takamaiman bincike tambaya game da shige da fice.