Arab tavasz és nlp

Mostanában nagyon rákattantam a "computational social science" témára, mivel úgy látom nagyon gazdag terület, amiben az nlp-nek is sok szép szerep jut(hat). Trey Causey a University of Washington PhD hallgatója az autoriter államok médiáit vizsgálja empirikus alapon, ehhez pedig nagyban támaszkodik a látens dirichlet allokációra (latent dirichlet allocation - LDA) a korpuszelemzésben. Disszertációjának összegzése szerint:

I model these meaning contests using Latent Dirichlet Allocation, a Bayesian method for discovering the latent topic structure of text corpora. Using an original dataset of more than 25,000 Egyptian and international news articles in both Arabic and English from November 2010 to March 2011, I build a model of regime-friendly, semi-independent, and independent media content before, during, and after the so-called Arab Spring uprisings in Tunisia, Egypt, Libya, and Bahrain. In the months leading up to the uprisings, while Egyptian state media focused on economic development projects and minimized attention to rising discontent int he region, non-state media offered extensive coverage of investigations of businessmen linked to the regime and fraudulent elections. When events in Tunisia ignited political imaginations in the region, state media consistently downplayed these events relative to their independent counterparts and employed conspiracy, foreign interference, and instability themes in their coverage. However, as the fall of the Mubarak regime drew closer, media coverage converged distinctly toward the language of independent media. I argue that this divergence and subsequent convergence around highly salient issues provides important information to potential collective actors about the ability and likely success of pro- and anti-regime groups to exert influence.

The Battle for Bystanders: Information, Meaning Contest, and Collective Action in the Egyptian Uprising of 2011 című tanulmányában a disszertációnál tömörebben foglalja össze Causey téziseit és módszertanát - érdemes elolvasni, hiába "draft", van benne valami. Úgy gondolom, hamarosan sok hasonló vizsgálódás lát majd napvilágot és bizonyosan az ipar is rákap az ilyen kutatások ízére.

Dzsihád és nlp

Tegnap futottam bele a Mancs honlapján egy interjúba Miklósi Lászlóval a Történelemtanárok Egyletének elnökével. Engem is meglep, hogy pont a fiatal bölcsészek tolódnak egyre inkább jobbra, de rögtön beugrott két tanulmány. Jeskó József, Bakó Judit és Tóth Zoltán A radikális jobboldal webes hálózatai című írása nagyon szépen összefoglalja, miképp formálódott a radikális jobb webes jelenléte, ami kulcsfontosságú a fiatalok elérésében (érdemes elolvasni, már csak azért is, mert eddig ez az egyetlen általam is ismert hazai adatvezérelt társadalomtudományi cikk). Ugyanakkor hiányzik még annak megértése, miképp válnak radikálissá pont humán beállítottságú emberek. Itt kerül képbe a dzsihád.  Rich Nielsen a Harvard Department of Goverment PhD hallgatója azt vizsgálja miért válnak radikálissá egyes muszlim vallás- és jogtudósok (ulema). Még nem publikált, de már elérhető tanulmánya a Jihadi Radicalization of Muslim Clerics remek példa arra, hogyan alkalmazható a számítógépes nyelvészet a társadalomtudományi tartalomelemzésben. Az absztrakt remekül összefoglalja a tanulmányt (elvégre ez a dolga):

This paper explains why some Muslim clerics adopt the ideology of militant Jihad while others do not. I argue that clerics strategically adopt or reject Jihadi ideology because of career incentives generated by the structure of cleric educational networks. Well-connected clerics enjoy substantial success at pursuing comfortable careers within state-run religious institutions and they reject Jihadi ideology in exchange for continued material support from the state. Clerics with poor educational networks cannot rely on connections to advance through the state-run institutions, so many pursue careers outside of the system by appealing directly to lay audiences for support. These clerics are more likely to adopt Jihadi ideology because it helps them demonstrate to potential supporters that they have not been theologically coopted by political elites. I provide evidence of these dynamics by collecting and analyzing 29,430 fatwas, articles, and books written by 91 contemporary clerics. Using statistical natural language processing, I measure the extent to which each cleric adopts Jihadi ideology in their writing. I combine this with biographical and network information about each cleric to trace the process by which poorly-connected clerics become more likely to adopt Jihadi ideology.

A The Economist cikke röviden és érthetően összegzi Nielsen eredményeit.  A mi szempontunkból legalább annyira érdekes a kutatás módszertana.

There is no way to determine the population of Muslim clerics, so randomly sampling clerics for analysis is not a possibility. Instead, I focus the analysis on contemporary Sunni clerics writing in Arabic. In order to compare Jihadi clerics to the those that were most likely to become Jihadi but ultimately did not, I intentionally over-sample two key groups of clerics: (1) Jihad clerics and (2) conservative Salafi clerics who share similar beliefs to Jihadis but reject the ideology of militant Jihad. Practically, this means the analysis is focused on conservative clerics primarily from Saudi Arabia (58 percent) and Egypt (22 percent) who are currently living or lived in the last century. [...] To measure the ideology of these clerics, I collect their books, articles, and fatwas with special emphasis on the latter.

Nielsen célja az volt, hogy beazonosítsa a radikális és nem-radikális ulemákat elérhető szövegeik alapján. Ehhez a gépi tanulást hívta segítségül:

In principle, identifying Jihadi ideology is often as simple as distinguishing between these two texts. However, with 29,430 texts from 91 clerics, close reading of each is infeasible. Instead, I measure cleric ideology by applying supervised learning methods from the statistical machine learning literature (Hastie, Tibshirani and Friedman, 2009) to the documents. Most of the previous work on measuring the ideology of political actors has used roll-call votes to estimate actor ideal points (Poole and Rosenthal, 1985; Martin and Quinn, 2002). This is not possible for Muslim clerics (because they do not vote on a common set of proposals), so instead, I directly scale the texts to estimate cleric support for militant Jihadi ideology. My method uses two sets of training documents, one of which is assumed to be Jihadi and the other of which is assumed to be non-Jihadi. The Jihadi corpus consists of 765 texts of various genres that are available on Jihadi web forums as the “Mujahid’s bookbag” (ﺎﻫﺪĐا ﺣﻘﻴﺒﺔ).13 These documents are specifically selected by Jihadis themselves as sources of spiritual instruction and advice, as well as mixed political and religious commentary.14 By using a set of known Jihadi documents as my training data, I avoid the difficult task of deciding which texts are most authentically “Jihadi”. Instead, I let Jihadis themselves identify the texts that are most representative of Jihadi ideology.

Lehet erőltetett párhuzamot látni egy radikális ulema és egy radikális bölcsész között, de úgy gondolom érdemes lenne Nielsen vizsgálatához hasonló módon elemezni a radikális jobboldal netes szövegeit és elkezdhetnénk empirikusan közelíteni ahhoz, hogy miért és hogyan radikalizálódnak a fiatalok a bölcsészkarokon.