Overview

• Problematic state of Arabic periodical studies
• Methodologic proposals to computationally address this state
• My approaches
  • bibliometrics
  • (social) network analysis
  • stylometric authorship attribution

Problems

1. periodicals are commonly perceived as a source but not a subject in its own right
2. the history of the periodical press
   • focusses on a few “core” publications from Cairo and Beirut
   • remains largely unexplored
   • is full of untested hypotheses
   • is heavily biased by national(ist) narratives

One of the many reasons

Map: geographic distribution of library holdings of al-Muqtabas

Open questions:

micro level

• how were individual periodicals produced?
• who authored the vast majority of anonymous articles?

meso level

• What are the core nodes (authors, periodicals, other works) in the ideosphere of the late Ottoman Eastern Mediterranean?

macro level

• How do we have to revise the nahḍa narrative / late Ottoman intellectual history if we include the full ideosphere of the press beyond Cairo and Beirut?

Methodology: computational approaches

• distant reading
  • bibliometrics
  • social network analysis
  • stylometric authorship attribution
• requires a digital corpus -> corpus building

corpus building

It’s labour and resource intensive. It really is!

1. get the data
   • text: transcription, train OCR/HTR
   • facsimiles: scanning
   • bibliographic metadata: transcription, validated iterative generation
2. transform the data into a human and machine readable edition
   • model the source
   • identify entities and link them to authority files
3. host, share and preserve the data

corpus building: Open Arabic Periodical Editions

1. ideas:
   • unite available facsimiles and transcriptions in a standard-compliant format
   • harvest, generate, validate and share open metadata
2. aims
   • validate the transcription against the facsimiles
   • improve the transcription with the help of the “crowd”
   • make everything citable for scholars, linkable for machines
   • share all data, metadata and tools with the broadest possible licences to facilitate access and re-use
3. principles
   • re-purpose available and established tools, technologies, and material
   • preference for open and simple formats and tools

corpus building: Open Arabic Periodical Editions

1. Open licences: CC BY-SA 4.0 (TEI, MODS, BibTeX), MIT license (XSLT, XQuery)
2. Social digital editions hosted on GitHub: gradually improve transcription and mark-up
3. Releases are archived at Zenodo: receive a DOI for reliable citation
4. Static web-view: provides side-by-side view of facsimiles and text
5. Access to bibliographic metadata through a public Zotero group

OpenArabicPE’s corpus

OpenArabicPE’s corpus

network of referenced periodicals: data sources

والأصح الدرعية بلام التعريف (راجع <bibl subtype="journal" type="periodical">مجلة <title level="j" ref="oape:bibl:3 oclc:1034545644" xml:id="title_16.d2e2291">الزهور</title> المصرية  <biblScope unit="volume" from="2" to="2">٢</biblScope> :  <biblScope unit="page" from="292">٢٩٢</biblScope></bibl>)

وانتخب <persName>فؤاد أفندي الدفتري البغدادي</persName> و<bibl><editor><persName>نوري أفندي</persName></editor> راس كتاب <textLang otherLangs="ota">القسم التركي</textLang> في <bibl type="periodical" subtype="newspaper">جريدة <title ref="oape:bibl:532">الزهور</title></bibl> البغدادية</bibl> نائبين عن <placeName ref="oape:place:372 geon:94824">كربلاء</placeName>.

network of referenced periodicals

Figure: Network of periodicals mentioned al-Ḥaqāʾiq, al-Ḥasnāʾ, Lughat al-ʿArab and al-Muqtabas; weights per issue

network of referenced periodicals: the core

Figure: Core nodes in the network of periodicals mentioned al-Ḥaqāʾiq, al-Ḥasnāʾ, Lughat al-ʿArab and al-Muqtabas; weights per issue

network of authors: data sources

<person>
    <persName><roleName type="pseudonym">ساتسنا</roleName></persName>
    <persName><roleName type="pseudonym">أمكح</roleName></persName>
    <persName><roleName type="pseudonym">فهر الجابري</roleName></persName>
    <persName><roleName type="rank">الأب</roleName> <forename>أنستاس</forename> <forename>ماري</forename> <surname><addName type="nisbah">الكرملي</addName></surname></persName>
    <persName><forename>أنستاس</forename> <forename>ماري</forename> <addName type="nisbah">الألياوي</addName> <surname><addName type="nisbah">الكرملي</addName></surname></persName>
    <persName><forename>بطرس</forename> <addName type="nasab">بن <forename>جبرائيل</forename></addName> <forename>يوسف</forename> <surname>عواد</surname></persName>
    <idno type="VIAF">39370998</idno>
    <idno type="oape">227</idno>
    <idno type="wiki">Q4751824</idno>
    <birth><date source="viaf" when="1866-08-05">1866-08-05</date> in <placeName ref="oape:place:216 geon:98182">Baghdad</placeName></birth>
    <death><date source="viaf" when="1947-01-07">1947-01-07</date> in <placeName ref="oape:place:216 geon:98182">Baghdad</placeName></death>
</person>

network of authors

Figure: Network of authors with bylines in al-Ḥaqāʾiq, al-Ḥasnāʾ, Lughat al-ʿArab and al-Muqtabas

network of authors: the core

Figure: Core nodes in the network of authors with bylines in al-Ḥaqāʾiq, al-Ḥasnāʾ, Lughat al-ʿArab and al-Muqtabas

network of authors: geography

Map: Locations in bylines in al-Muqtabas (Cairo and Damascus)

Map: Locations in bylines in al-Ḥaqāʾiq (Damascus)

Map: Locations in bylines in al-Ḥasnāʾ (Beirut)

Map: Locations in bylines in Lughat al-ʿArab (Baghdad)

problem: network reflects only 17% of articles

state of the field

• question of authorship attribution has not received much attention
• implicit and commonly accepted hypothesis: editors authored all anonymous articles themselves

problems:

• hypothesis remains untested
• we don’t know the potential candidates for authorship even within the hypothesis
• it is unlikely that a single person authored everything

computational authorship attribution: stylometry

• this has not yet been applied to Arabic
• method: compare stylistic features to gain a numerical measure of difference
  • distance measure is sensitive to composition of corpus
• stylistic features: most frequent words (MFWs)
  • have been shown to be extremely effective for authorship attribution
• chunking/sampling impacts results
  • features: run multiple iterations and have them vote
  • texts: minimum length of 4000-5000 words

stylometry: 5000 words

Figure: bootstrap consensus network of articles (length >= 5000 words, 100–1000 MFWs), colours by modularity group

stylometry: how to deal with the minimal length requirement?

Figure: bootstrap consensus network of sections and articles (length >= 5000 words, 100–1000 MFWs) in al-Muqtabas, coloured by modularity

stylometry: how to deal with the minimal length requirement?

Figure: bootstrap consensus network of sections of articles in 123 periodical issues based on consensus of 100–1000 MWFs, coloured by modularity

Summary

• corpus building: modelling is indispensible for article-level analysis
• network of periodicals: confirms importance of Cairo
  • shows vast differences between journals: parochial, regional, trans-regional
  • should guide corpus building
• network of authors: surprising
  • importance of Iraq and Iraqis
  • very limited overlap between journals from the same city
  • many authors not mentioned in standard accounts
• stylometry: promising
  • problem: 5000 word minimum length

Thank you!

• Contributors to OpenArabicPE: Jasper Bernhofer, Dimitar Dragnev, Patrick Funk, Talha Güzel, Hans Magne Jaatun, Jakob Koppermann, Xaver Kretzschmar, Daniel Lloyd, Klara Mayer, Tobias Sick, Manzi Tanna-Händel and Layla Youssef
• Links:
  • Slides: https://OpenArabicPE.github.io/slides/2020-idh/
  • Paper (pre-print, submitted): https://doi.org/10.5281/zenodo.1413610
  • Project URL: https://www.github.com/OpenArabicPE
  • Project blog: https://openarabicpe.github.io
  • Twitter: @tillgrallert
  • Email: grallert@orient-institut.org
• Licence: slides and plots are licenced as CC BY-SA 4.0