Final Report for Ottoman Recogito (OttoRec)

As was detailed in the interim report, the chosen source for toponym extraction were the so-called ‘Registers of Important Affairs’ (mühimme defterleri). The particular sources are composed of summaries of imperial orders sent to the provinces: as a result they are an extremely valuable source for the purposes of the present exercise since they are particularly rich in toponyms. There are in total 419 volumes covering the years 1553–1915, many of which are already published published in a digital format (typically as PDFs of MA or PhD theses). This meant that we had access to an important corpus that was already transcribed and available online.

Among the different volumes and periods available, we decided to focus on volume no. 12 covering the years between 1570-1572. Our choice was made according to the following criteria: that this was a period of rapid and important change at the geopolitical level, with wars on different fronts and imperial orders dispatched to meet those challenges. Furthermore, it was during this period that the limits of the empire’s expansion were set and its borders consolidated. Such a source therefore allows us to identify the locations with which the imperial centre was corresponding at a critical juncture in its history. Thus, as both a quantitative (i.e., the number of toponyms contained) and qualitative dataset, the particular source represents an ideal pilot study for shedding light on the internal and external networks that Istanbul had with other places. The register contains more than 900 pages and 1,283 orders (for a sample of such orders see figure 1). With the initial upload of the text to Recogito, the index was removed in order to avoid duplicates.

Figure 1: A sample page from Volume 12 of the Registers of Important Affairs.

Each team member was allocated a section of the text and started annotating place names. We quickly encountered various issues with bulk annotation of toponyms that appear multiple times in a text, and provided feedback to the Recogito to address the various issues we encountered. At an early stage we realized one fundamental mistake during the uploading of the document: each imperial order is preceded by a summary in Modern Turkish. This meant that we could not use bulk annotation since place names appear in both summaries and the orders themselves, thus creating multiple occurrences that would distort the picture of how often a particular toponym appears in the original source.

Another task necessary for building a gazetteer involved identifying toponyms from a published Ottoman Gazetteer (Tahir Sezen, Osmalı Yer Adları [Istanbul, second edition 2017] with existing URIs from other gazetteers in an effort to enrich our identifies toponyms. Using a matching script developed by Vagelis Papadias (PhD candidate at Harokopio University, Athens), we were able to automatically find coordinates for 8,458 toponyms as they were recorded in Osmalı Yer Adları (see Figure 2). We then had to manually verify each entry. Of the first 3,000 entries checked about 2,200 were incorrect, due to transcription issues and the form in which these toponyms appeared. For this reason this exercise had to be aborted and we decided that the entries of Osmalı Yer Adları would have to be manually identified – a task to be undertaken at a later stage.

Figure 2: Automatically-generated georeferencing of toponyms from Osmanlı Yer Adları

Going back to annotating Volume 12 of the Important Affairs Registers, a spreadsheet was created in order to collect all the non-identified toponyms from the existing Recogito gazetteers, whereby all relevant information would be entered (see Figure 3).

Figure 3: The spreadsheet with non-identified toponyms

Figure 4 shows the map with all the toponyms of Volume 12 of the Important Affairs Registers that are included in the Recogito gazetteers. Overall, there were:

• 10,812 annotations
• 8,213 place annotations (7,217 verified, 996 not identified)
• 2,347 people annotations
• 647 toponyms identified
• 390 toponyms not identified
• 1,037 toponyms in total in the source

Figure 4: Map showing the toponyms of Volume 12 of the Important Affairs Registers that are part of the existing Recogito gazetteers

Table 1 shows the top ten of toponyms:

Table 1: The ten most-encountered toponyms in Volume 12 of the Important Affairs Registers

Particularly useful was the “Tag” function in Recogito. This allowed us to annotate thematic spatial entities that are not necessarily toponyms. For example, Figure 5 shows the castles and fortresses mentioned in the source.

Figure 5: Map showing the castles and fortresses mentioned in Volume 12 of the Important Affairs Registers

Other topics of interest which were annotated using the “Tag” function were: tribes; temples; churches; mosques; religious schools (medreses); shrines; ships; rivers; rebels/rebellions; prisons; ports; nomadic communities; mines; mountains; marketplaces; pious foundations (vakıfs); caravanserais; bridges. We also found tagging to be a particularly useful tool for including the administrative dependency of a toponym as it appears in our primary source, thus greatly facilitating the gazetteer compilation.

The final result of about 1,000 toponyms was below expectations. This is due to the fact that 1570-72 was the period of the war of Cyprus and the Battle of Lepanto. These two military events have greatly affected the distribution and frequency of appearance of particular toponyms, especially those adjacent to these places. Other volumes of the Registers of Important Affairs will not necessarily have this bias towards particular toponyms. We fully expect that annotation of different registers will help us gradually enrich the list of toponyms and meet our aim of building towards an Ottoman Gazetteer.