Urban data: harnessing subjective sociocultural data from local newspapers

SPECIAL COLLECTION: DATA POLITICS IN THE BUILT ENVIRONMENT RESEARCH CORRESPONDING AUTHOR: Filipe Mello Rose Digi al Ci y Science, Ha enCi y Uni e si y Hambu g, Hambu g, DE; Labo a o y o Knowledge A chi ec u e, Technical Uni e si y D esden, D esden, DE [email p o ec ed] KEYWORDS: ci ies; da a-based go e nance; da a poli ics; da a ica ion; local jou nalism; planning; sociocul u al da a; subjec i e da a; u ban da a; u ban go e nance TO CITE THIS ARTICLE: Mello Rose, F., & Chang, J. (2023). U ban da a: ha nessing subjec i e sociocul u al da a om local newspape s. Buildings and Ci ies, 4(1), pp. 369–385. DOI: h ps://doi.o g/10.5334/bc.300 U ban da a: ha nessing subjec i e sociocul u al da a om local newspape s FILIPE MELLO ROSE JUIWEN CHANG ABSTRACT As da a-based go e nance becomes mains eam, social and cul u al in e ac ions ha cha ac e ise u ban li e a e a isk o being igno ed in decision-making p ac ices i only supposedly objec i e, quan i iable da a a e used. In his con ex , his a icle concep ualises subjec i e sociocul u al da a as a da a o m ha conside s a ci y’s in angible and unquan i iable social and cul u al aspec s. A me hodology is p oposed o collec ing and using subjec i e sociocul u al da a by highligh ing local p ess a icles as a po en ial da a sou ce. A pilo applica ion conduc ed in Hambu g, Ge many, demons a es a po en ial in eg a ion o subjec i e sociocul u al da a in o u ban planning p ocesses by analysing o e 2500 local newspape a icles. The indings e eal ha local jou nalism can be a da a sou ce o unde s anding di e se social and cul u al in e ac ions be ween ci izens and u ban places. This s ee -le el in o ma ion om local newspape a icles can (1) p o ide u ban planne s wi h an o e iew o newspape men ions o any speci ic u ban a eas, (2) suppo he iden i ica ion o local deba es, and (3) aid in he obse a ion o eme ging places o sociocul u al in e ac ions. This app oach can suppo he di e se go e nmen and non-go e nmen s akeholde s engaged in da a-based go e nance o be e accoun o in angible sociocul u al aspec s o u ban li e. PRACTICE RELEVANCE This esea ch suppo s go e nance ac o s in dealing wi h he epis emological limi a ions o pu posely ga he ed and/o objec i e da a by concep ualising a new—cu en ly un apped—da a ype: subjec i e sociocul u al da a sou ced om local jou nalism. By using geog aphical ex analysis on local newspape a icles, u ban planne s and decision-make s gain access o a weal h o s ee -le el in o ma ion, local deba es and empo al dynamics o u ban issues. This app oach p o ides a comp ehensi e unde s anding o in angible and unquan i iable aspec s o u ban li e, allowing o mo e in o med and con ex -sensi i e decision-making. The p ac ical bene i s include iden i ying di e se uses o u ban spaces, cap u ing local public deba es, and acking he eme gence and disappea ance o places in he public sphe e, possibly leading o mo e e ec i e and inclusi e u ban planning p ac ices. *Au ho a ilia ions can be ound in he back ma e o his a icle 370Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 1. INTRODUCTION: BETTER DATA FOR URBAN GOVERNANCE Many new ends in u ban planning, such as algo i hmic go e nance, sma ci ies and digi al wins, es on da a-based u ban go e nance. In his ype o u ban go e nance, ‘a complex sys em o ac o s, ela ionships, p ocesses, and echnologies’ is designed o decide on u ban issues based on di e se se s o a ailable and analysable da a (Ma ei e al. 2020: 124). Da a-based planning is a c ucial ins umen o add essing he upcoming en i onmen al and socie al challenges (Ba y & Yang 2022: 146) as i allows comp ehensi e and in e disciplina y analyses o complex p oblems by making hem ‘mo e “legible” o go e ning’ (Mejias & Could y 2019: 4). A he same ime, he da a ica ion o u ban go e nance aises a ious issues conce ning da a alidi y, as well as he eliabili y, ep esen a i eness and in e ope abili y o a ailable da a (Bunde s & Va ó 2019). C ucially, da a-based u ban go e nance is limi ed by all u ban p oblems no being ‘equally knowable and sol able’ by da a-based go e nance p ac ices (Bunde s & Va ó 2019). Many key in angible and unquan i iable aspec s o u ban li e (e.g. he ole o amily-owned co ne shops in a local neighbou hood) play a i al ole in he day- o-day u ban expe iences o ci izens (Jacobs 1961/1992). Likewise, u ban places such as pa ks o pa emen s ‘mean no hing di o ced om hei p ac ical, angible uses’ (Jacobs 1961/1992: 111). This in o ma ion is ha dly quan i iable and is hus igno ed in da a-based go e nance. These limi a ions o da a-based go e nance can a leas be mi iga ed by including b oade and mo e a iega ed da ase s (Ba y 2019). In his sense, he p esen a icle concep ualises subjec i e sociocul u al da a as an impo an ye equen ly igno ed o unde used inpu o da a-based go e nance. Wi h his concep ualisa ion and an exempla pilo me hodology, he aim o his pape is o suppo he di e se s akeholde s engaged in da a-based go e nance (including go e nmen and non-go e nmen ac o s) o be e accoun o in angible and unquan i iable aspec s o u ban li e in hei decision-making. Subjec i e da a a e a pa icula da a o m ha explici ly in ol es human judgmen in i s collec ion. Sociocul u al da a depic aspec s ‘ ela ed o he di e en g oups o people in socie y and hei habi s, adi ions, and belie s’ (Camb idge Uni e si y P ess n.d.). Join ly, subjec i e sociocul u al u ban da a depic sel - epo ed and e alua ed accoun s o a ci y’s social and cul u al p ac ices (i.e. he in e ac ions o ming a ci y’s social and cul u al ab ic). In pu suing he esea ch goal o b oadening he da a sou ces o u ban go e nance, his a icle concep ualises he u ili y o subjec i e sociocul u al u ban da a and p oposes a me hodology o collec ing and using his da a ype. This concep ualisa ion and me hodology es on a c i ical e iew o he li e a u e on he di e en ypes o da a used in u ban go e nance and a syn hesis o concep ual discussions on subjec i e da a. Mo eo e , local p ess a icles we e iden i ied as a po en ial and la gely un apped sou ce o subjec i e sociocul u al u ban da a, which, once geo- e e enced and syn hesised, can in o m local go e nance decisions. Empi ically, his s udy e lec s on a pilo s udy ha aims o in eg a e subjec i e sociocul u al da a in o u ban planning by d awing on a esea ch p ojec ca ied ou wi h he s a e-owned company o eal es a e managemen and land asse s o he Ge man ci y o Hambu g: he Landesbe ieb Immobilienmanagemen und G und e mögen Hambu g (LIG). The pilo s udy de eloped a geo- pa sing module wi h subjec i e sociocul u al da a ha aims o complemen ‘objec i e’ da a used in decision-making ega ding public eal es a e policies. This s udy d aws on da a om 2539 local newspape a icles om Hambu g published online be ween 2018 and 2022. The a icles we e au oma ically analysed o spa ial e e ences in Hambu g (i.e. geo-pa sed and geocoded) and clus e ed a ound 10 algo i hmically gene a ed opics. The pilo use cases show ha local jou nalism can (1) p o ide u ban planne s wi h an o e iew o newspape men ions o any speci ic u ban a eas, (2) suppo he iden i ica ion o local deba es, and (3) aid in he obse a ion o eme ging places o sociocul u al in e ac ions. In concep ualising local p ess a icles as a sou ce o subjec i e sociocul u al u ban da a o u ban go e nance, his s udy d aws on and con ibu es o esea ch on da a ica ion p ocesses in go e nance. Howe e , as ‘da a ica ion is c oss-disciplina y in na u e’ (Lombo g e al. 2020), his s udy also d aws on mul iple academic ields. Fi s , i d aws on media s udies o a gue o he gene al u ili y o (local) jou nalism as a da a sou ce o he da a-based go e nance (Ba is ini e 371Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 al. 2013; Sa ın & Uluğ ekin 2019). Second, i d aws on digi al na u al language p ocessing (NLP) me hods o geog aphical ex analysis and opic modelling (Cai 2021; Middle on e al. 2018; Po e e al. 2015). This a icle is s uc u ed as ollows. Following his in oduc ion, he pape concep ualises and si ua es subjec i e sociocul u al da a among o he u ban da a o ms used o u ban planning. The me hodology is hen desc ibed and he esul s o an empi ical pilo applica ion a e epo ed. This empi ical sec ion se s ou he NLP me hods and h ee po en ial use cases o he ou lined echniques. The las sec ion e iews and discusses he use o subjec i e sociocul u al da a o u ban go e nance. 2. EXPLICITLY SUBJECTIVE DATA FOR DATA-BASED GOVERNANCE 2.1 FROM GOVERNMENT DATA TO DATA-BASED GOVERNANCE In con as o he adi ional go e nmen s’ use o sel -collec ed da a, da a-based go e nance d aws on a a ie y o da a sou ces. These ange om ou comes o algo i hms and he In e ne o Things (Cole a & Ki chin 2017), as well as la ge da ase s, dashboa ds o su eillance sys ems o c ea e a mo e e icien adminis a ion and go e nance o places (Bunde s & Va ó 2019; Ki chin & McA dle 2017; Valdez e al. 2018; Zubo 2019). Mo eo e , wi h he inc easing p e alence o sou ces in da a-based go e nance, many s akeholde s a e able and needed o p o ide o analyse da a. Mos no ably, in e na ional pla o m co po a ions now also in o m decision-making and policy de elopmen wi h hei da ase s and analy ical ools (e.g. Re be g 2020). Along wi h nume ous o he ac o s, his da a ica ion has hus u he ad anced he shi om adi ional go e nmen o go e nance by beyond- he-s a e ‘ho izon al associa ional ne wo ks o p i a e (ma ke ), ci il socie y […] and s a e ac o s’ (Swyngedouw 2005: 1992). Da a-based go e nance, hus, coo dina es his ‘beyond- he-s a e go e ning’ a angemen by p o iding and analysing signi ican and di e se da ase s. This da a ica ion imp o es he capaci y o an icipa o y planning in a go e nance sys em ha includes an e e g ea e a ie y o ac o s (beyond adi ional go e nmen adminis a ions) by using da a ‘ o e olu ionize he p ocess o policy analysis’ (Ma ei e al. 2020: 124). While da a- based go e nance can hus ep esen new business oppo uni ies o companies collec ing da a as a by-p oduc (e.g. Re be g 2020), i can also allow esea che s o ci il socie y o ganisa ions o pu go e nmen s and co po a ions unde g ea e sc u iny (e.g. Dal on 2019). C ucially: da a ica ion enables us o deal wi h lines o inqui y ha we e di icul i no impossible o pu sue be o e. (Lombo g e al. 2020: 207) In his sense, da a-based go e nance can also os e a mo e collabo a i e and inclusi e app oach o go e nance, whe e a ne wo k o ac o s wo k oge he o iden i y and add ess complex socie al challenges. By p o iding and analysing da a ha a e o in e es o u ban go e nance, hese non- go e nmen ac o s hus, di ec ly o indi ec ly, engage in expanding ne wo ks o da a-based go e nance. 2.2 ONTOLOGY OF DATA SOURCES OF DATA-BASED GOVERNANCE Be o e digi alising public adminis a ion, economic and social ac i i ies, da a-based go e nance elied p ima ily on pu posely collec ed da a, such as s uc u ed su eys and obse a ion da a. This ype o da a sou ce includes da a delibe a ely p oduced by go e nmen ins i u ions (e.g. da a om census and s a is ics o ices) o non-go e nmen o ganisa ions suppo ing da a-based go e nance wi h hei da a. Fo ins ance, da a-based go e nance o en d aws on specialised su ey ins i u es o uni e si ies o collec da a wi h di e se o mal me hods (online su eys, in-pe son in e iews, online acking, e c.). A ai o hese pu posely collec ed da a o su ey- ype da a is, howe e , ha i s con olled collec ion p ocesses educe he da a’s ‘scope, empo ali y, and size, and a e qui e in lexible in hei adminis a ion and gene a ion’ (Ki chin & McA dle 2016: 2). 372Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 Digi alising social, economic and go e nmen al ac i i ies has enabled he collec ion o usable da a as a by-p oduc . As public adminis a ions become mo e digi ised, mo e da a a e p ocessed digi ally in go e nmen bu eauc acies and become a po en ial new da a sou ce (Ki chin & McA dle 2016). This ype o da a, which a e ga he ed as a by-p oduc , is ei he obse ed (i.e. as a esul o people using echnology) o in e ed (i.e. consolida ed in o ma ion om exis ing da a sou ces) (Taylo & Rich e 2015). Da a ha o igina e as a by-p oduc o go e nmen ac i i ies also include eco ds, such as budge s and spending da a, and da a on asse s as well as om egis ies (e.g. egis a ion o esidences, inancial ac i i ies o ehicles). C ucially, his da a sou ce ype is no limi ed o any ex-an e speci ica ion o pu pose in da a-based go e nance, as he possible knowledge gains a e de ined a e he da a a e collec ed du ing da a p epa a ion (Ki chin & McA dle 2016) (Table 1). All da a (sou ces) ha e na u al epis emological limi s. Pu posely collec ed da a a e limi ed o add essing socie al p oblems known be o e he da a collec ion based on scien i ic heo y (Ki chin 2014). While mos ly ee o sampling and s a is ical e o s, his da a ype has a low po en ial o ul il unan icipa ed usages (Ki chin & McA dle 2016). Uns uc u ed da a collec ed as a by-p oduc , in con as , a e p one o sampling and s a is ical e o s and biases o igina ing om unequal uses o a se ice o p oduc bu allow explo a o y esea ch. T adi ional da a sou ces in u ban go e nance gene ally elied on he accessibili y o pu posely collec ed da a, such as su eys and censuses. As pu posely collec ed da a equi e an ex-an e de ini ion o esea ch p oblems (i.e. be o e da a collec ion), da a-based go e nance mus also conside da a sou ces allowing g ea e explo a o y esea ch o iden i y new esea ch p oblems. 2.3 SUBJECTIVE AND SOCIOCULTURAL DATA Sociocul u al da a depic ‘socie y and hei habi s, adi ions, and belie s’ (Camb idge Uni e si y P ess n.d.). They ea u e s uc u ed and s a is ical cul u al, e hnic, eligious o demog aphic in o ma ion, as well as (mo e subjec i e) da a on cul u al p ac ices, day- o-day ac i i ies and ou ines, and belie s. In his a icle, sociocul u al da a a e essen ially da a ha allow one o unde s and be e a gi en a ea’s social and cul u al ab ic. Da a-based go e nance p ima ily d aws on supposedly ‘objec i e’ da a as impe sonal and neu al ‘ aw in o ma ion’ ha a e bes collec ed wi h he leas ‘subjec i e’ human in ol emen possible (Riede & Simon 2016). This applies o sociocul u al da a ha a e o en educed o a iables o e hnic, eligious o class iden i ica ion which can be assessed somewha objec i ely. Subjec i e da a, in con as , signi ican ly and explici ly in ol e human judgmen in i s p oduc ion. Despi e a g ea e isk o biases, his da a ype is used as pu posely collec ed da a o measu e social phenomena such as happiness, a ec ion o wellbeing (e.g. Kahneman & K uege 2006; Macků e al. 2020). Subjec i e measu es a e ele an o wo main easons. Fi s , in he absence o (mo e) objec i e measu es, subjec i e ones can (a leas pa ially) depic social phenomena on which da a a e sca ce. Second, subjec i e measu emen s allow one o: cap u e changes in bo h he explici and he implici componen s o he a iable being measu ed and, he e o e, […] can be be e sui ed o he s udy o b oadly de ined concep s. (Jahedi & Méndez 2014: 3) GOVERNMENT DATA NEW DATA-PRODUCERS IN DATA- BASED GOVERNANCE Pu posely collec ed da a (S uc u ed) Pe iodically collec ed da a on he poli y, e.g. census da a, su eys o s a is ics adminis a ions Da a om non-s a e da a collec ion companies and esea ch ins i u es, i.e. ma ke esea ch, uni e si y esea ch Da a as a collec ed as a by-p oduc (Uns uc u ed) Da a collec ed as a by-p oduc o go e nmen adminis a ions, e.g. pe mi da a, da a om public companies (i.e. heal h insu ance, unemploymen insu ance, public anspo ), budge da a Da a collec ed as a by-p oduc o o dina y economic and social ac i i ies, e.g. social media companies, a ge ed ad e isemen s, cus ome e iews Table 1: Concep ualisa ion o di e en da a sou ces and collec ion o ms. 373Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 Da a on he sociocul u al aspec s o u ban li e mee bo h easons o using subjec i e measu es. Fo one, sociocul u al ac o s o u ban li e o e ew al e na i e ‘objec i e’ measu es. Mos u ban in e ac ions and p ac ical, angible uses o places a e di icul o quan i y objec i ely wi hou using (some imes a - e ched) p oxy measu es. Fo ano he , he sociocul u al aspec s o u ban li e a e a b oadly de ined concep ha e e s o he in angible aspec s, such as he di e se social and cul u al in e ac ions en iching u ban li e. Allowing di e en ac o s o de e mine he ele an social and cul u al in e ac ions will likely be e depic he local u ban social and cul u al ab ic. Mo eo e , while s a is ical and sys ema ically quan i ied sociocul u al da a allow la ge-scale analyses, small- scale subjec i e da a can help planne s o be e unde s and he social condi ions shaping an a ea’s u ban (sociocul u al) ab ic. U ban sociocul u al da a depic day- o-day ac i i ies, ou ines, alues and public spaces’ p ac ical, angible uses. In his sense, his a icle unde s ands subjec i e sociocul u al da a as collec ions o desc ip ions o social in e ac ions and cul u al li e in an a ea. This includes accoun s o cul u al and ci ic e en s, he uses o public spaces, o popula meanings o places (e.g. he impo ance o adi ional co ne shops o ba s o a communi y) ele an o unde s anding he local u ban ab ic. Combining di e en ypes o da a sou ces allows o iangula ing in o ma ion and complemen ing analyses wi h di e en pe spec i es. In his sense, adding da a ga he ed as a by-p oduc o da a- based go e nance eases explo a o y, unexpec ed lines o enqui y, while subjec i e sociocul u al da a expand da a-based go e nance o be e accoun o a gi en a ea’s social and cul u al ab ic. 2.4 LOCAL JOURNALISM AS A DATA SOURCE FOR URBAN GOVERNANCE This s udy concep ualises and empi ically es s he use o local p ess a icles as a sou ce o subjec i e sociocul u al u ban da a. In doing so, i d aws on mul iple s udies in which p ess a icles se e as an explici ly subjec i e sou ce o geog aphical and sociocul u al da a. Fo ins ance, Voukela ou e al. (2021: 300) ind ha ‘news da a a e a new p omising da a sou ce o he u he explo a ion o subjec i e well-being’. Ba is ini e al. (2013: 157) use online news eeds as da a sou ces o e icien ly map geohaza ds (i.e. ea hquakes, landslides and loods) unde he assump ion ha whene e geohaza ds ha e ‘ ele an consequences, news is epo ed on he In e ne ’. G ego y & Pa e son (2020) elabo a e on his p ac ice unde he i le o geog aphical ex analysis, o which he schola s geo-pa se his o ical p ess a icles o map his o ical po e y in he UK. The schola s’ s udy on his o ical po e y demons a es how analysing ex ual da a can p o ide insigh s in o geog aphical pa e ns (G ego y & Pa e son 2020). Ozgun & B oekel (2021) use egional newspape s o enqui e abou he egional di e ence in a i udes o inno a ion by quan i ying he men ions and sen imen s o di e en egional newspape a icles. Howe e , while mos s udies ocus on he na ional o a leas egional le el, Pa e son (2020: 66) poin s ou ha : e e ences o place, especially a mo e local le els, can make he impac (s) o abs ac concep s […] appea mo e conc e e by si ua ing hem wi hin de inable geog aphical bounda ies. O he s udies ha e used simila me hodologies o mobilise social media da a o da a analysis in en i onmen al (Ghe mandi & Sinclai 2019) and ou ism esea ch (Chen e al. 2021). Howe e , in con as o social media da a, p ess a icles ha e a leas a minimal deg ee o quali y con ol. Mo eo e , while in e ac ions on social media a e highly di e se and equen ly use colloquial o g oup-speci ic language, newspape a icles a e ypically w i en o b oade audiences and equi e less con ex -speci ic language o knowledge abou he au ho s o be in elligible. In his spi i , applying geog aphical ex analysis o local p ess a icles appea s as an auspicious way o in oducing new sou ces o subjec i e sociocul u al da a o da a-based u ban go e nance. The use o newspape a icles as a sou ce o sociocul u al da a is g ounded on he assump ion ha he con en p oduced by newspape s a leas gene ally depic s cul u al p ac ices, day- o-day ac i i ies and ou ines, and belie s o a gi en place. The way media ou le s ‘ ame an issue shapes how people unde s and and emembe i ’ (Ozgun & B oekel 2021: 3). Mo eo e , media ou le s 374Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 ha e nes ed in e es s ha in luence hei aming and ca ego isa ions o wha is ‘newswo hy’. Howe e , while he p ess migh ad ance speci ic agendas, ‘ hey do no do so independen o hei audience’ (3). As mos media o ganisa ions a e comme cially d i en, he choice o newswo hiness and he one o epo ing news is s ongly linked wi h a emp s o gain and e ain he la ges possible audiences (Scheu ele 1999; Gen zkow & Shapi o 2010; Agi das 2015). As people seem o ‘selec i ely expose hemsel es o a i ude-consis en in o ma ion’ (Ea le & Hodson 2022: 9), aligning o he poli ical o ien a ion and epo ing p io i ies o a a ge eade ship is a comme cial necessi y and a esponse o public demand (Gen zkow & Shapi o 2010). While na ional news ou le s end o become inc easingly pa isan o win audience loyal y in a poli ically di ided ma ke (Gen zkow & Shapi o 2010), his is no he case o local newspape s due o hei limi ed geog aphic compe i ion (Agi das 2015). 3. EMPIRICAL PILOT APPLICATION: SOURCING LOCAL JOURNALISM FOR URBAN GOVERNANCE In he ollowing, his s udy desc ibes and discusses an empi ical pilo applica ion o he possible uses and sou ces o sociocul u al da a o da a-based go e nance concep ualised abo e. While his pilo applica ion has been de eloped in pa ne ship wi h a s a e-owned company o a speci ic case (i.e. o imp o e da a-based go e nance o public eal-es a e asse s), he me hodology (including he used algo i hms) is a ailable as open-sou ce code on Gi Hub. This way, he p ac ical implemen a ion o sou cing local jou nalism o u ban go e nance can be es ed and imp o ed in o he con ex s. 3.1 CONTEXT OF THE EMPIRICAL PILOT APPLICATION This s udy is si ua ed wi hin a b oade esea ch p ojec ha in es iga es new modes o da a-based planning conduc ed wi h Hambu g’s s a e-owned company o eal es a e managemen and land asse s, he LIG. The Ge man ci y o Hambu g is no excep ion o he end o da a ica ion o u ban go e nance, and he LIG aims o imp o e i s go e nance ac i i ies wi h new digi al ools o spa ial analysis. To de elop a me hod o digi al da a-based eal es a e and land managemen in he ci y-s a e, a collabo a ion be ween he LIG and Ha enCi y Uni e si y has engaged in he c ea ion o a digi al pla o m o he e alua ion o land pa cels. This pla o m will consolida e a ious da a sou ces and in o m decisions ega ding he acquisi ion and sale o land by public adminis a ions. Mo e p ecisely, he digi al ool allows he compa ison o land pa cels based on hei su oundings (i.e. he p oximi y o a ious u ban ameni ies) and hei connec i i y (i.e. isoch ones in he ci y) in he ‘LIG-Finde ’ module. In addi ion, he ‘geo-pa sing’ module p o ides a spa ial o e iew o geocoded ex s. This geo-pa sing module displays geocoded documen s om he s a e pa liamen and mul iple egional and local newspape s. This includes he Elbe Wochenbla , a hype local ee newspape ha his s udy uses as a pilo applica ion. Elbe Wochenbla is a hype local ad e isemen -based newspape om Hambu g. I s a icles a e dissemina ed online and in se en weekly neighbou hood-le el p in edi ions, dis ibu ed ee o cha ge o all households (unless hey op ou ) wi hin a speci ic deli e y zone (Figu e 1). As an ad e isemen -based newspape , i p o ides a pla o m o and depends on local businesses o ad e ise hei p oduc s and se ices. Elbe Wochenbla ea u es sho a icles using colloquial language on local poli ics, spo s, ci ic li e, cul u al e en s and businesses. Nex o Hambu ge Wochenbla , Elbe Wochenbla is he second la ges ee newspape in Hambu g wi h app oxima ely 300,000 weekly p in ed copies.1 Howe e , he media landscape in Hambu g is also—and p obably mo e subs an ially—shaped by egional and Hambu g-based na ional newspape s, such as MOPO/ BILD, Hambu ge Abendbla and Zei .2 Elbe Wochenbla ’s accessibili y and high olume o publica ions om 2018 o 2021 make i a aluable es case o he pilo applica ion o he LIG-Finde ’s geo-pa sing module. Du ing ha pe iod, he newspape was majo i i ely owned by Madsack Medieng uppe, wi h a mino i y s ake held by Funke Medieng uppe. In 2021, he newspape was ully acqui ed by Funke Medieng uppe, which in eg a ed he p e iously independen edi o ial o ice wi h o he local newspape s om 375Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 Hambu g (no ably wi h Hambu ge Wochenbla ) in Janua y 2023. While Madsack Medieng uppe is egionally ocused on no he n Ge many, Funke Medieng uppe is he 10 h la ges media g oup in Ge many (Ins i u ü Medien- und Kommunika ionspoli ik 2022). This s udy uses all 3511 newspape a icles published online (and in a leas one o he se en neighbou hood p in edi ions) om 2018 o 2021. The ime ame was selec ed due o he newspape ’s abo e-a e age publishing ac i i y. The ocus is on a icles au ho ed by he newspape ’s edi o ial s a o sa egua d a consis en o ma and ocus on he newspape ’s a ge a eas. This excludes submissions and commen a ies om eade s ha a y as ly in o ma , making an au oma ed co pus analysis di icul a his pilo ing s age. Elbe Wochenbla was selec ed o his pilo applica ion because o (1) i s hype local ocus and desc ip ions o neighbou hood li e in hei epo ing and (2) he a ailabili y o a la ge co pus o a icles. A he ime o da a collec ion, mos o he neighbou hood newspape s in Hambu g co e ed smalle a eas. In his sense, using local jou nalism as sociocul u al da a is exempli ied using Elbe Wochenbla as a pa adigma ic case (Fly bje g 2006: 232). This s udy does no aim o discuss he media p ac ices o he newspape beyond a necessa y con ex ualisa ion. 3.2 NATURAL LANGUAGE PROCESSING (NLP) OF LOCAL JOURNALISM As uns uc u ed da a and a by-p oduc o economic and social ac i i ies, local newspape a icles equi e signi ican and me iculous p epa a ion o p o ide angible bene i s o u ban decision- make s as sociocul u al da a. The p epa a ion and cu a ion o he uns uc u ed da a d aw on he p ac ices o he eme ging ield o NLP, ad ancing i s use in da a-based u ban go e nance (e.g. Cai 2021). NLP is an algo i hmically enabled analy ical p ac ice ha combines linguis ics, compu e science and a i icial in elligence o ‘s uc u e la ge olumes o uns uc u ed da a’ (Cai 2021: 1). This s udy uses wo p ima y me hods o NLP o ans o m uns uc u ed newspape da a in o a sou ce o subjec i e sociocul u al u ban da a: geo-pa sing and opic modelling. 3.2.1 Geo-pa sing The geo-pa sing p ocess allows he ex ac ion o loca ion names and coo dina es om he newspape a icles’ uns uc u ed ex ual da a. The ‘pa sing’ p ocess sepa a ed ex s in o a compu e - eadable o ma o de ec e ms ha can be linked o geog aphical iden i ie s. In o he wo ds, geo-pa sing in ol es wo subp ocesses ha i s ex ac loca ion iden i ie s (i.e. place names) and hen geocodes hem (Wang e al. 2020). Fi s , oponym ecogni ion and geo agging iden i y names o geog aphical loca ions in he ex based on name en i y ecogni ion. This ask loca es and classi ies ‘named en i ies’ in uns uc u ed ex in o p ede ined ca ego ies (e.g. pe son names, o ganisa ions, loca ions, ime exp essions, quan i ies, mone a y alues and pe cen ages). A e compa ing mul iple name en i y ecogni ion classi ica ion lib a ies, his s udy used spaCy, an open-sou ce so wa e lib a y a ailable o o e 70 languages.3 Second, his s udy uses Nomina im4—an open-sou ce geocode o con e place names in o geocoo dina es— o geocode he geog aphical en i ies iden i ied in he p e ious s ep. The algo i hm builds a s a is ical model es ima ing he p obabili y o a named en i y e e ing speci ic geoloca ed place (Middle on e al. 2018). En ies wi h a sco e below a p obabili y o 0.95 we e checked manually. The geocoding o he 3511 newspape a icles iden i ied 10,320 men ions o 1718 places in Hambu g. A e he geo-pa sing p ocess, he da abase o newspape a icles equi ed u he p epa a ion and da a cleaning. A signi ican po ion o place men ions e e s o en i ies linked o a eas ac oss mul iple neighbou hoods (i.e. dis ic s, egions o Hambu g as a whole). As hese place names a e somewha a bi a ily geocoded o a poin wi hin he a ea hey ep esen , hese place men ions clu e neighbou hood-le el analysis wi h in o ma ion ela ed o a eas ou side he neighbou hoods.5 Fo his eason, he s udy excludes all men ions o place names e e ing o geog aphical a eas abo e he neighbou hood le el. This s ep excludes 1665 men ions o ‘Hambu g’, 356 men ions o he gene al pa s o Hambu g (e.g. Wes Hambu g, Sou h Hambu g) and 845 men ions o one o he se en dis ic s. This il e ing educes he da abase o 7903 men ions o 1689 places. 376Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 Elbe Wochenbla is dis ibu ed in an a ea ha co e s he en i e wes and sou h o Hambu g, including 34 o 104 o icial neighbou hoods (Figu e 1). This a ea accoun s o 41% o Hambu g’s o al a ea and 35.3% o Hambu g’s popula ion. The neighbou hoods se iced by he newspape a e highly di e se ac oss a ious me ics anging om popula ion densi y o employmen . Fo ins ance, he a ea o deli e y includes he h ee weal hies and he h ee poo es neighbou hoods o he ci y. The geocoding con i ms he hype local ocus o he newspape ’s epo ing, as o e 89.5% o all place men ions in he da abase a e in he deli e y zone. Mo eo e , he as majo i y (2539 a icles, o 72.3%) o all 3511 analysed a icles men ion a leas one place ha can be loca ed wi hin a neighbou hood whe e he newspape ’s p in edi ion is deli e ed. The e o e, he ollowing s udy ocuses on he a eas in which he newspape ’s p in e sion is dis ibu ed. The ‘cleaned’ co pus o newspape a icles hus includes 2539 a icles men ioning 1424 di e en u ban places in he deli e y a ea a o al o 6665 imes. Tes s o he ex co pus and he pa sed and geocoded places indica e ha he numbe o imes places in each neighbou hood a e men ioned signi ican ly co ela es o ha neighbou hood’s popula ion (Pea son co ela ion = 0.667; p < 0.01) and a ea (Pea son co ela ion = 0.546; p < 0.01). No signi ican s a is ical ela ionship exis s be ween he numbe o place men ions in an a ea and he a ea’s median income, housing p ices o o he es ed socioeconomic ac o s (i.e. sha e o mig an s, sha e o highe educa ion deg ees). 3.2.2 Topic modelling Topic modelling is an impo an s ep in NLP ha allows he iden i ica ion o di e en opics and opic dis ibu ions in la ge co po a o ex . While he e a e se e al opic modelling me hods, he mos popula a e la en di ichle alloca ion (LDA) (Blei e al. 2003) and s uc u ed opic modelling (STM) (Robe s e al. 2019). The p ocess ollows he gene al p ocedu e o clus e ing co-occu ing lis s o keywo ds in o a p ede ined numbe o opics in he i s s ep wi hou needing any p elabelled da a. To imp o e he ce ain y o iden i ying he co ec opic, his s udy used bo h me hods o iden i y 10 opics (i.e. co-occu ing lis s o keywo ds) in he ex co pus o local newspape a icles. While bo h majo opic modelling me hods p oduce sensible esul s, a close analysis o he esul ing 10 keywo d lis s o ming each opic sugges ed he use o STM, as i allows he inclusion o addi ional Figu e 1: Spa ial dis ibu ion o he place men ions in he local newspape . 377Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 in o ma ion. In he case o his s udy, sen imen sco es ega ding he a icle’s s yle and one we e added o he analysis. These sen imen sco es es on ‘Tex BlobDE’6 and he ‘Valence Awa e Dic iona y and Sen imen Reasone ’.7 In he second s ep, he opic modelling algo i hm calcula es he likelihood o an indi idual a icle being linked o each o he 10 opics (i.e. ‘ opic sco e’). This analysis was based on he en i e y o each a icle. As he sum o he 10 opic sco es is equal o 1, he a icles ha ing a opic sco e > 0.5 (mo e han all o he nine opics combined) in any gi en opic a e conside ed o be signi ican ly ela ed o ha opic. A icles wi hou a opic sco e > 0.5 a e unde s ood no o be signi ican ly linked o any o he 10 opics (e.g. combining mul iple opics). Table 2 shows he opics iden i ied in he opic modelling, he numbe o a icles (wi h a espec i e opic sco e o a leas 0.5) and he numbe o places men ioned in a icles conce ned wi h ha opic. The numbe o place men ions a ies signi ican ly ac oss di e en opics. Fo example, while opic 1, ‘pandemic’, was he mos equen opic in he co pus, a icles clea ly linked o opic 4, ‘mobili y’, ha e he mos place men ions. Due o he high h eshold o be conside ed ela ed o a pa icula opic, abou 45% o a icles a e no linked o any opic. 3.3 EXEMPLAR USE CASES OF SOURCING LOCAL JOURNALISM FOR URBAN GOVERNANCE The ollowing h ee use cases illus a e how he applica ion o NLP o local newspape a icles can gene a e da a o be e comp ehend a gi en a ea’s social and cul u al ab ic. While he i s use case ocuses on geo-pa sing, he second demons a es he applica ion o opic modelling ID TOPIC LABEL ARTICLES PLACE MENTIONS KEYWORDS IN ORDER OF LIKELIHOOD (ROOT WORDS) 1 Pandemic 325 416 co ona; accine; i us; es ; pool; in ec ; cen e; o e ; hygiene; ship; ip; numbe ; mask; pandemic; dis ance 2 Heal h and social ca e 275 1,318 heal h; hospi al; help; also; ca e; wo k; homeless; employee; o ice; need; suppo ; se ice; cen e; con ac ; will 3 Mobili y 233 1,515 a ic; b idge; pa k; s a ion; ailway; cons uc ; ca ; oad; plan; s ee ; ee; esidence; wo k; a ea; can 4 Spo s 231 960 eam; club; spo ; game; league; will; oo ball; championship; playe ; play; ime; coach; goal; poin ; s a 5 Poli ics 223 638 dis ic ; g een; SPD; o ice; CDU; ci izen; membe ; elec ; ini ial; boa d; ede al; poli ic; associa ion; pa liamen ; le 6 Schools and educa ion 183 531 school; s uden ; child en; wo k; young; elemen a y; p ojec ; class; lesson; pa en ; eache ; educa e; dis ic ; young; lea n 7 Housing 151 391 build; new; plan; a ea; dis ic ; house; enan ; de elop; en ; apa men ; cen e; eu o; p ope y; cons uc ; o ice 8 Conce s and e en s 134 385 music; icke ; a is ; hea e; pe o m; s age; a is ; play; conce ; show; band; musician; ilm; choi 9 Po ai s o neighbou hood igu es 93 277 like; a he ; pho o; ime; wi e; iend; emembe ; e en; s ill; jus ; now; always; li e; wan ; know 10 Ci ic li e 79 234 es i al; museum; o’clock; child en; o e ; ca é; o gan; chu ch; e en ; open; ma ke ; cul u e; place; isi o ; ake 0No clea opic 1,588 2,671 Table 2: Resul s o he opic modelling. No e: CDU = Ch is ian Democ a ic Union (Ch is lich Demok a ische Union); SPD = Social Democ a ic Pa y (Sozialdemok a ische Pa ei). 384Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 Gen zkow, M., & Shapi o, J. M. (2010). Wha d i es media slan ? E idence om U.S. daily newspape s. Econome ica, 78(1), 35–71. DOI: h ps://doi.o g/10.3982/ECTA7195 Ghe mandi, A., & Sinclai , M. (2019). Passi e c owdsou cing o social media in en i onmen al esea ch: A sys ema ic map. Global En i onmen al Change, 55, 36–47. DOI: h ps://doi.o g/10.1016/j. gloen cha.2019.02.003 G ego y, I. N., & Pa e son, L. L. (2020). English language and his o y: Geog aphical ep esen a ions o po e y in his o ical newspape s. In The Rou ledge handbook o English language and digi al humani ies (pp. 418–439). Rou ledge. DOI: h ps://doi.o g/10.4324/9781003031758-22 Ins i u ü Medien- und Kommunika ionspoli ik. (2022). Ranking—Die zehn g öß en deu schen Medienkonze ne 2021 (June). S a is a. h ps://de.s a is a.com/s a is ik/da en/s udie/194686/um age/ die-10-g oess en-medienkonze ne-in-deu schland/ Jacobs, J. (1992). The dea h and li e o g ea Ame ican ci ies [1961]. Vin age/Penguin Random House. Jahedi, S., & Méndez, F. (2014). On he ad an ages and disad an ages o subjec i e measu es. Jou nal o Economic Beha io & O ganiza ion, 98, 97–114. DOI: h ps://doi.o g/10.1016/j.jebo.2013.12.016 Kahneman, D., & K uege , A. B. (2006). De elopmen s in he measu emen o subjec i e well-being. Jou nal o Economic Pe spec i es, 20(1), 3–24. DOI: h ps://doi.o g/10.1257/089533006776526030 Ki chin, R. (2014). Big da a, new epis emologies and pa adigm shi s. Big Da a & Socie y, 1(1), 2053951714528481. DOI: h ps://doi.o g/10.1177/2053951714528481 Ki chin, R., & McA dle, G. (2016). Wha makes big da a, big da a? Explo ing he on ological cha ac e is ics o 26 da ase s. Big Da a & Socie y, 3(1), 2053951716631130. DOI: h ps://doi. o g/10.1177/2053951716631130 Ki chin, R., & McA dle, G. (2017). U ban da a and ci y dashboa ds: Six key issues. In Da a and he ci y (Vol. 1). Rou ledge. DOI: h ps://doi.o g/10.4324/9781315407388-9 Lombo g, S., Dencik, L., & Moe, H. (2020). Me hods o da a ica ion, da a ica ion o me hods: In oduc ion o he Special Issue. Eu opean Jou nal o Communica ion, 35(3), 203–212. DOI: h ps://doi. o g/10.1177/0267323120922045 Macků, K., Caha, J., Pász o, V., & Tuček, P. (2020). Subjec i e o objec i e? How objec i e measu es ela e o subjec i e li e sa is ac ion in Eu ope. ISPRS In e na ional Jou nal o Geo-In o ma ion, 9(5), a . 5. DOI: h ps://doi.o g/10.3390/ijgi9050320 Ma ei, S., Leoni, F., & Villa i, B. (2020). Da a-d i en an icipa o y go e nance. Eme ging scena ios in da a o policy p ac ices. Policy Design and P ac ice, 3(2), 123–134. DOI: h ps://doi.o g/10.1080/25741292.2020. 1763896 Mejias, U. A., & Could y, N. (2019). Da a ica ion. In e ne Policy Re iew, 8(4). h ps://policy e iew.in o/ concep s/da a ica ion. DOI: h ps://doi.o g/10.14763/2019.4.1428 Middle on, S. E., Ko dopa is-Zilos, G., Papadopoulos, S., & Kompa sia is, Y. (2018). Loca ion ex ac ion om social media: Geopa sing, loca ion disambigua ion, and geo agging. ACM T ansac ions on In o ma ion Sys ems, 36(4), 40:1–40:27. DOI: h ps://doi.o g/10.1145/3202662 Ozgun, B., & B oekel, T. (2021). The geog aphy o inno a ion and echnology news—An empi ical s udy o he Ge man news media. Technological Fo ecas ing and Social Change, 167, 120692. DOI: h ps://doi. o g/10.1016/j. ech o e.2021.120692 Pa e son, L. L. (2020). Mapping aus e i y: Geog aphical ex analysis o UK place-names in The Gua dian and The Daily Teleg aph. In Mul imodal app oaches o media discou ses. Rou ledge. DOI: h ps://doi. o g/10.4324/9780367332907-4 Po e , C., A kinson, P., & G ego y, I. (2015). Geog aphical ex analysis: A new app oach o unde s anding nine een h-cen u y mo ali y. Heal h & Place, 36, 25–34. DOI: h ps://doi.o g/10.1016/j. heal hplace.2015.08.010 Re be g, J. W. (2020). Si ua ed da a analysis: A new me hod o analysing encoded powe ela ionships in social media pla o ms and apps. Humani ies and Social Sciences Communica ions, 7(1), a . 1. DOI: h ps://doi.o g/10.1057/s41599-020-0495-3 Riede , G., & Simon, J. (2016). Da a us : O , he poli ical ques o nume ical e idence and he epis emologies o big da a. Big Da a & Socie y, 3(1), 2053951716649398. DOI: h ps://doi. o g/10.1177/2053951716649398 Robe s, M. E., S ewa , B. M., & Tingley, D. (2019). s m: An R Package o s uc u al opic models. Jou nal o S a is ical So wa e, 91, 1–40. DOI: h ps://doi.o g/10.18637/jss. 091.i02 Sa ın, P., & Uluğ ekin, N. (2019). Analyzing newspape maps o ea hquake news h ough ca og aphic app oach. ISPRS In e na ional Jou nal o Geo-In o ma ion, 8(5), a . 5. DOI: h ps://doi.o g/10.3390/ ijgi8050235 Scheu ele, D. (1999). F aming as a heo y o media e ec s. Jou nal o Communica ion, 49(1), 103–122. DOI: h ps://doi.o g/10.1111/j.1460-2466.1999. b02784.x 385Mello Rose and Chang Buildings and Ci ies DOI: 10.5334/bc.300 TO CITE THIS ARTICLE: Mello Rose, F., & Chang, J. (2023). U ban da a: ha nessing subjec i e sociocul u al da a om local newspape s. Buildings and Ci ies, 4(1), pp. 369–385. DOI: h ps://doi.o g/10.5334/bc.300 Submi ed: 10 Feb ua y 2023 Accep ed: 05 June 2023 Published: 30 June 2023 COPYRIGHT: © 2023 The Au ho (s). This is an open-access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (CC-BY 4.0), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal au ho and sou ce a e c edi ed. See h p://c ea i ecommons.o g/ licenses/by/4.0/. Buildings and Ci ies is a pee - e iewed open access jou nal published by Ubiqui y P ess. Swyngedouw, E. (2005). Go e nance inno a ion and he ci izen: The Janus ace o go e nance-beyond- he- s a e. U ban S udies, 42(11), 1991–2006. DOI: h ps://doi.o g/10.1080/00420980500279869 Taylo , L., & Rich e , C. (2015). Big da a and u ban go e nance. In J. Gup a, K. P e e , H. Ve es , & M. Ros-Tonen (Eds.), Geog aphies o u ban go e nance (pp. 175–191). Sp inge . DOI: h ps://doi. o g/10.1007/978-3-319-21272-2 Valdez, A.-M., Cook, M., & Po e , S. (2018). Roadmaps o u opia: Tales o he sma ci y. U ban S udies, 55(15), 3385–3403. DOI: h ps://doi.o g/10.1177/0042098017747857 Voukela ou, V., Gab ielli, L., Miliou, I., C esci, S., Sha ma, R., Tesconi, M., & Pappala do, L. (2021). Measu ing objec i e and subjec i e well-being: Dimensions and da a sou ces. In e na ional Jou nal o Da a Science and Analy ics, 11(4), 279–309. DOI: h ps://doi.o g/10.1007/s41060-020-00224-2 Wang, J., Hu, Y., & Joseph, K. (2020). Neu oTPR: A neu o-ne oponym ecogni ion model o ex ac ing loca ions om social media messages. T ansac ions in GIS, 24(3), 719–735. DOI: h ps://doi.o g/10.1111/ gis.12627 Zubo , S. (2019). The age o su eillance capi alism: The igh o a human u u e a he new on ie o powe . PublicA ai s.