Baki Project Team. Gülşah Taşkın (Boğaziçi U), Sarah Ketchley (U of Washington), Selim S. Kuru (U of Washington), Natasha Dietzler (Senior UX Designer at Nordstrom), Brad Holland (Senior Software Development Engineer), Rutvi Patel (Software Engineer at HBO) “SCRIBE/KÂTİB: A Tool for Digitizing Arabic Alphabet-Based Ottoman Turkish Texts”

View presentation: https://echo360.ca/media/7303e22d-0875-466f-8716-94fd14cdf053/public

View transcript/additional materials: https://dhsi.org/wp-content/uploads/sites/7317/2023/03/Taskin_R2L_Transcript_05032021-Sarah-Ketchley.pdf

Abstract: SCRIBE/KÂTİB is an open source Arabic to Latin transcription tool developed as a part of The Baki Project to create an exemplary digital edition of an exceptionally famous Ottoman poet’s works preserved in manuscripts and now ready to be employed for other texts written in Ottoman Turkish. SCRIBE/KÂTİB has been constantly evolving since its initial development in 2016, as The Baki Project transcribers and other users identify issues and suggest improvements for the tool. The core feature of SCRIBE/KÂTİB is that every transcription decision (i.e. every character typed) is captured together with metadata about the character. The metadata contains information about the character including, but not limited to: the Turkish transcription alphabet symbol, the Ottoman script extended Arabic alphabet symbol, the symbol’s order of importance and any need for disambiguation (for distinguishing this character from other visually similar characters). This paper will discuss tool development to date, showcase manuscript transcription output, and present future iterations of SCRIBE/KÂTİB, which focus on building out collaborative features, such as authentication, a project management system, and integration of other digitized texts with The Baki Project’s manuscript database through communication with other right-to-left transcription tools.

Till Grallert “Open Arabic Periodical Editions: Contributing to the Digital Commons Without the Help We Cannot Get”

View presentation: https://echo360.ca/media/12ff281a-10ec-45f7-853a-fce39f0762d9/public

Abstract: The proposed short paper introduces the project Open Arabic Periodical Editions (OpenArabicPE) as a case study of minimal computing to address the specific challenges and affordances of digital humanities in the Eastern Mediterranean. I will outline the socio-technical challenges for digital approaches to the cultural heritage of societies outside the Global North. I pose that a digital episteme deeply rooted in 20th-century, english-speaking, American neoliberal capitalism causes a neo-orientalist silencing of the material heritage of predominantly Arabic-speaking Islamicate societies. This places the onus of corpus building on individual scholars and citizens and requires mitigation strategies on every level of the digital workflow. Influenced by minimal computing and pirate care and based on the guiding principles of accessibility, simplicity, sustainability and credibility, OpenArabicPE originated in 2015 from an attempt to address the core problem of corpus building: the lack of functional Arabic OCR without which one would have to resort to human transcribers. Yet, we could produce reliable ground truth for improving and—with the advent of machine-learning based approaches—training OCR algorithms by combining existing transcriptions from shadow libraries, such as al-Maktaba al-Shāmila, with openly available digital facsimiles for the purpose of validating the former. The proposed paper will introduce the workflows to produce standard-compliant, machine-actionable, scholarly digital editions of Arabic periodicals by creatively repurposing open and free-to-use tools and platforms, which allow us to currently host full-text TEI XML editions of some 630 issues from six journals and more than 7 million words without any funding and resources beyond our own labour time and laptop computers.

Merve Tekgürler (Stanford U) “Ottoman Transkribus: Training HTR+ Models for Eighteenth-Century Ottoman Turkish Paleography“

View presentation: https://echo360.ca/media/d3d37e15-e3d2-41a5-bf13-d491198a097b/public

Abstract: Handwritten text recognition (HTR) has emerged as an important method for digital humanities practice in the last decade, especially for the field of history. Transkribus, an EU-funded software package that is designed to train machine-learning based HTR models, has seen great success for multiple, primarily European, languages, However, Transkribus has certain difficulties when it comes to training models using RTL languages, which they are trying to address based on user feedback. This talk will reflect on the processes of model training for eighteenth century Ottoman Turkish (OT) paleography, using Transkribus, with a particular focus on RTL support after the reverse text integration update of March 16, 2021. Due to the fact that OT was written in Arabic script and modern historical transcription conventions use latinized transcription alphabets, researchers needed to either reverse their text or mirror their images before uploading them to Transkribus to train a model that follows academic transcription guidelines. This created additional challenges and barriers in model training, such as taking extra steps to reverse the transcribed texts before uploading that this update aims to amend. By comparing the experiences of data uploading and processing before and after this update, this talk offers insights into how RTL languages can be successfully supported in digital resources.

Max Dugan (U Pennsylvania) and Elliot Montpellier (U Pennsylvania) “Multiple Scripts: Regularizing Social Media Discourse in Urdu and its (Many) Transliterations”

View presentation: https://echo360.ca/media/edc23c16-2408-4e98-9129-fcb1a55c3900/public

View transcript/additional materials: https://dhsi.org/wp-content/uploads/sites/7317/2023/03/EMM_MJD_RTL_Transcript-Max-Dugan.pdf

Abstract: This paper outlines our methodology for regularizing the transliteration of Urdu-based discourse in digital spaces for analysis with various DH methods. Scholars, acknowledging social media as an increasingly central discursive space and the irregularity of Urdu-based online transliterations, have taken generative steps in this regard (Irvine and Ann 2012, Sharf and Ur Rahman 2017). At the same time, these methods require substantial technical knowledge and do not account for English words or Arabic-based Islamic expressions (e.g. ما شاء الله) and their variants (e.g. mashAllah, ma sha’ Allah, mash’Allah, mA, etc.). This paper advocates the use of spreadsheet applications as an accessible and adaptable tool for regularizing variability in language, script, and orthographic practices in Urdu-based social media discourse. Drawing from challenges faced in our respective research projects, we focus on a small set of piously affective terms and phrases that highlight the applicability of this work for DH projects in Muslim digital worlds. The paper argues that shortening the technical learning curve improves the robustness of DH work with RTL languages, particularly in social media spaces. We elaborate how regularizing orthographic variability in the digital culture in Islamic South Asia and its diasporas widens the analytical scope for scholars working with multiscript, polylingual, and religiously-inflected datasets.

Nada Ammagui (NYU Abu Dhabi) “Shaping Old Sharjah: A Geospatial Analysis of Biennial Art Exhibitions in Old Sharjah from 2003–2019”

View presentation: https://echo360.ca/media/ca464af4-ee41-4de6-a389-8ac2ea1b7d7b/public

Abstract: Scant data is collected about cultural institutions in Sharjah, UAE, especially for the purpose of conceptualizing its burgeoning art scene. While there is no paucity of papers about the “world-class” museums in nearby Abu Dhabi and their roles in shaping a national identity (Ajana, McClellan, Ouroussoff), rarely do we see critics engage with the home-grown institutions that make up Sharjah’s cultural landscape. In my presentation, I argue that using geospatial artwork data from the Sharjah Biennial, a contemporary art exhibition held in Sharjah, offers a unique path to visualizing the rapid spatial growth of this exhibition and the impact of its spatiality on local residents. In my talk, I will retrace the history of the Sharjah Biennial, discuss the layout of the exhibitions using catalogue-generated maps, and consider the impact of governmental support for cultural and artistic institutions in Sharjah. In addition, I will reflect on my methods of data collection, challenges in working with bilingual exhibition catalogues, and the use of mapping tools to represent spaces in the Arabian Peninsula. The aim of my talk is to shed light on geospatial data collection in the UAE and to think about opportunities for the integration of RTL cultural data into global Digital Humanities circles.

Wafa Fatima Isfahani (Rutgers U) “Expanding Islamicate Digital Humanities: A Network Analysis of the Sufi Naqshbandi Order”

View presentation: https://echo360.ca/media/53db90ab-9fb4-4405-bf11-4caf849e397f/public

View transcript/additional materials: https://dhsi.org/wp-content/uploads/sites/7317/2023/03/Presentation-Transcript_Isfahani-wafa-isfahany.pdf

Abstract: My seed project is a network analysis of Sufi orders and silsilah (lineages) locating them in space and time. I am using a number of different primary and secondary sources to collect, curate and finally visualize data that will help scholars of Sufism to recapture some of the structure and shape of its dynamic intellectual networks. Specifically, I am looking at the Sufi Naqshbandi order that has its roots in Bukhara in what is now Uzbekistan, starting from the Sufi saint, scholar and founder of the order Hazrat Khwaja Muhammad Baha’uddin Shah Naqshband (Algar), and spreading to South Asia. The Naqshbandi Sufi order exists today in many parts of the world through a network of variegated relationships. In effect, there are at least two attributes that stand out in my network analysis of the Naqshbandi Sufi order: the pedigree (of religious significance), and the master-disciple chain. Sufi orders have been neglected in the field of South Asian, Middle Eastern and Islamicate (Hodgson) digital humanities, and through my project I argue for the enormous potential they hold in furthering the study of the Islamicate world, specifically viz. the transmission and cross-fertilization of ideas. It is similar to the more popular Hadith studies scholarship, in so far as it looks at teacher-disciple transmissions called ‘citation networks,’ but that popularity is in large part because of the availability of extensive citations in comprehensive encyclopedias. Sufi orders, on the other hand, suffer from the digital humanist’s neglect owing to the lack of OCRed Tazkirah (memorandum) texts. My project is therefore an effort towards bringing Sufism into the fold of digital humanities research, and creating opportunities to reframe research questions on Islamic intellectual networks from analogue to digital form.

Tariq Yousef (U Leipzig) “Textual Alignment of Persian Poetry”

View presentation: https://echo360.ca/media/e3bb0c35-5364-4af3-b92f-08ec7931186c/public

View transcript/additional materials: https://dhsi.org/wp-content/uploads/sites/7317/2023/03/RTL-Transcribtion-Tariq-Yousef.pdf

Abstract: Text alignment is an essential task in Digital Humanities and NLP; it compares texts to find correspondences between different textual units. Text alignment can be employed to achieve various purposes, such as collation, text reuse detection, and translation alignment (Yousef/Jänicke). We will present our work on two different alignment-related topics, detecting and visualizing textual reuse in Persian poetry and Translation alignment of Persian Poetry. The corpus we studied comes from the medieval period between the 10th and 14th centuries. It consists mainly of Shahnameh, Khayyam quatrains, Masnavi, and Divan Hafez. Persian literature is one of the richest and oldest literature in the world (Marashi), and Persian poetry has been translated into many languages. Most of the existing automatic alignment algorithms are unsupervised such as Giza++ and IBM Models (Brown et al); we aimed to benefit from the manually created translation pairs to align other poems. Our corpus contains a manually aligned English translation of Divan Hafez at word/phrase level, the alignment has been done by our colleagues Mariam Foradi and Saman Rezaie using Alphios (Almas/Beaulieu) and Ugarit (Yousef) alignment editors. This alignment has generated 24000 unique translation pairs, and we used these data to align Masnavi and Khayyam quatrains with their English translations. In this paper, we will present our methods of the textual alignment of Persian poetry, discuss the corpus- and language-related difficulties we faced, and show our research results. master-disciple chain. Sufi orders have been neglected in the field of South Asian, Middle Eastern and Islamicate (Hodgson) digital humanities, and through my project I argue for the enormous potential they hold in furthering the study of the Islamicate world, specifically viz. the transmission and cross-fertilization of ideas. It is similar to the more popular Hadith studies scholarship, in so far as it looks at teacher-disciple transmissions called ‘citation networks,’ but that popularity is in large part because of the availability of extensive citations in comprehensive encyclopedias. Sufi orders, on the other hand, suffer from the digital humanist’s neglect owing to the lack of OCRed Tazkirah (memorandum) texts. My project is therefore an effort towards bringing Sufism into the fold of digital humanities research, and creating opportunities to reframe research questions on Islamic intellectual networks from analogue to digital form.