The Social World of John Chrysostom

Wendy Mayer and David O'Brien

Centre for Early Christian Studies, Australian Catholic University

Introduction (Slides 1-4)

What David and I will present to you today is a quick overview of the web database that we are developing. I will discuss in brief the background to the project, David (who handles the technical side) will explain how it currently works, will show how we've dealt with the issue of presenting Greek text across different platforms, and will then concentrate in the remainder of the paper on how we currently structure the data (in a three-level hierarchy) in relation to how, we now realise, the data can better be organised to facilitate searching (using multiple levels linked according to conceptual graph theory).

It is important to stress first that the project is very much in the developmental stage. Funding (supplied by the Australian Research Council) is limited and, while it commenced in mid 2001, the two research assistants involved, David O'Brien and Silke Trczionka, are able to work on the database on only a very part-time basis.

What are we trying to capture in the database and how did the project come about? In 1991 my then doctoral supervisor, Pauline Allen, in recognition that early christian homilies can contain useful data about everyday life in the late antique world, began a project systematically analysing all of the Constantinopolitan homilies up to 700 CE for such data. The homilies of John Chrysostom thought to have been delivered in that city (398-403/4), some 153 in number (out of a total of 823), formed the backbone of the data entered at the time in a primitive computer database (see P. Allen and W. Mayer, Computer and Homily: Accessing The Everyday Life Of Early Christians, Vigiliae Christianae 47 [1993] 260-280). As our research on the social world of Constantinople and Antioch as seen through John Chrysostom's sermons continued it became clear that it was important to analyse all of the homilies, not just a subset, and that we needed to develop a more efficient way of tagging and recording the data, not to mention finding a way to guarantee that a search would locate all instances of what we were looking for. It also became clear that with the advances in computer technology, we would be able to make this research tool widely available, if the database were mounted on the web. A side benefit of this decision was the facility for entry of data via the web by personnel located in different places around Australia.

The aim of the database is thus to model the social world of late antique Antioch and Constantinople as seen through the eyes of John Chrysostom. The site is directed toward the specialist user and historians of late antiquity and postgraduate students. In terms of size, it is anticipated that when the database is complete, it will hold around 70,000 entries.

To give you some idea of how we tagged and organised the data originally, we started out with a three level hierarchy (Slide 2). The top level (domain) organised the data under a broad category, such as the social or private world; the mid level (class) broke the broad domain down into a more specific sub-class (e.g. plants); while the third level (category) provided a yet more specific label for the data (e.g. trees). In addition to the three searchable levels of labelling, brief information regarding the location of the passage in the homily and a short description of what the passage contained was recorded. While we retained the idea of organising the internal structure of the database hierarchically, we have made substantial changes to accommodate the needs of the web and demands of the user, as well as refining the original entries from 1991-1993 to bring them in line with the current project. These changes are reflected in the current version of the user interface (Slide 3). I now hand over to David to take you on a brief tour of how the database currently works. I stress the word "currently." As David will explain, we are about to change substantially the internal structure, with corresponding changes to the user interface.

We recognised that the usefulness of a search database largely depends on the ease in retrieving its information. To this end we opted for two search strategies allowing the user to choose between selecting a search item via pull-down menus and allowing the user to enter in a keyword. The first search strategy is the most accurate though it does require that the user traverse up to 3 such pull-down menus as he/she progressively narrows a search from a broad domain to a specific category. The latter strategy has the advantage of being potentially quicker but has the disadvantage of being rather hit-or-miss. The hit-or-miss nature of this search strategy can largely be ameliorated by a table of synonyms linked to each category.

The design layout of the web pages is intentionally simple. Our focus has been on efficient searching and quick download times taking into account the needs of vision-impaired users and those who have low-spec computers and low internet bandwidths. We felt that we could not assume that every user had the processing power of Java on their local machine and for that reason we have opted for a server-side solution based on PHP and MySQL. Our intention is that the web database be as cross-platform as possible (Windows, Mac OS and Linux).

All information databases, including this one, raise the problem of interpretation (Slide 4). Our task essentially is to build an ontology, that is, a hierarchy of categories that make up the social world of Late Antiquity. However, this data ought not to be interpreted in isolation. Researchers using this web database need to appreciate that this ontology does not stem from an allegedly, objective viewpoint of a modern-day sociologist or cultural historian but rather reflects the biases of a fourth century representative of the social elite, a male, Christian bishop. Our intention is to represent the social world as seen through John Chrysostom's eyes. It is the task of the user of the web database to compensate for these biases in his or her own reconstructions.

To facilitate this we have included several brief scholarly essays by Dr Mayer on the social background of the world in which John Chrysostom lived, a biography of his life, the background of the cities of Antioch and Constantinople where Chrysostom conducted his ministry, and a selective bibliography which aims to give the user a more informed understanding from which one may interpret the data.

Demonstration of a search (Slides 5-12)

Let's say that we wish to list all the references to fasting in Chrysostom’s homilies (Slide 5). In this demonstration we shall follow the strategy where one begins at a broad domain and progressively narrows the search to the specific category of fasting.

The first step is to choose the homily/homilies to search (Slide 6). One is able to search one homily or all homilies. In the next version of the web database we shall look at the option of selecting more than one individual homilies to search.

The second step is to select [Broad Search] and to choose one of the 8 domains from the pull-down menu which best fits the category we are searching for; in this case, Asceticism best fits our search for fasting (Slide 7).

We are then directed to the pull-down menu for a narrowed search (Slide 8). Asceticism yielded 11 classes (out of a total of 70). From this pull-down menu we choose Behaviour as the most suitable class for fasting.

We are then directed to the pull-down menu for a specific search (Slide 9). Behaviour yielded 40 categories (out of a total of 700) and there we can select fasting.

The results page for our search on fasting is then displayed (Slide 10). Each result entry includes an informed description enabling the researcher to assess quickly whether that particular result is relevant to his or her research. These results can either be printed or emailed. Each result entry also includes a Source Reference. The user can click on the Source Reference to display the Greek context for that passage.

The Greek context is ten lines of TLG text from the relevant passage of the homily (Slide 11). It is based on the TLG CD ROM Format which follows the ISO 9600:1988 (E) standard for volume and directory structure. This standard imbeds indexing information in the text. We created a PHP programme to read the Chrysostom file in binary mode into a MySQL database which is indexed on volume, page and line number for more efficient access.

The Greek is displayed as a unicode 16bit glyph or character. In Normal Greek (without accents) the hexidecimal block is hex 0370-03ff and can be displayed on any browser with unicode support.

The Greek context can also be displayed in Greek Extended mode which displays all accents and diacritical marks (Slide 12). Greek Extended is based on Arial Unicode MS which is able to display polytonic Greek fonts. This font set is supplied with MS Office 2000 and Office XP. We understand that Mac workstations are also able to display Greek Extended with the fonts Gentium and GentiumAlt and Unix similarly with CASLON. The Greek Extended unicode is mapped to the hexidecimal block 1f00-1fee.

Shortcomings of the first version (Slides 13-14).

While the web database in its current form would already be extremely useful for researching aspects of the social world of Late Antiquity gleaned from the homilies of John Chrysostom there are several ways that the search engine can be improved and made more efficient.

First, the current search engine is too restrictive in being unidirectional. One is only able to search from broad to specific. It could be advantageous to be able to search the other way, from specific to broad, and then down to another specific category, for example.

Second, the current search engine still yields irrelevant data. Even though the current search engine greatly simplifies the task of searching a desired category (in the case of fasting it gives a list of 40 categories from 700 possible categories in the pull-down menu) it still presents irrelevant search items corresponding to the general class of Behaviour. For example, we see that crowd behaviour and drunken behaviour are listed from the class Behaviour but which one would not expect under the domain of Asceticism (Slide 14). There must be a way of fine-tuning searches to yield only revelant options.

Ways of improving the search engine (Slides 15-19)

Both shortcomings are the by-product of an inadequate data structure. This is clearly demonstrated by means of conceptual graph theory. Conceptual graph theory is a way of representing an ontology based on the hierarchy of its categories or types. In this hierarchy, all types inherit from zero or more supertypes. That is a subtype inherits all the properties from each of its supertypes. At the topmost level is a special type called an entity from which all types inherit. At the bottom-most level is a special type called an absurdity. This bottom-most type will inevitably inherit contradictory or incongruous properties and hence its name.

In the example of fasting (slide 15) we clearly see the inadequacy of our data structure. Fasting is a proper subtype of Behaviour but Behaviour is not a proper subtype of Asceticism (fasting behaviour but behaviour, asceticism). One could imagine a behaviour that is not ascetic. An obvious example of this is, according to one's perspective, drunkenness.

A clear solution to this particular example is, based on the recognition that all supertype-subtype relations are transitive, the introduction of an intermediate type Ascetic Behaviour and the promotion of Behaviour to the top level (Slide 16). This yields the hierarchy (fasting < ascetic behaviour < asceticism) and the hierarchy (fasting < ascetic behavour < behaviour). Having normalized the hierarchy to conform to conceptual graph theory we can now allow for bi-directional searches and eliminate any irrelevant search categories.

A further improvement to the data structure is to bypass the artificial constraint of the topology being restricted to 3 levels (Slide 17). By adopting an n-level hierarchy our data model will be more representative of the ontology of the social world of Late Antiquity and less artificial than if we forced this ontology to conform to a 3-level hierarchy.

Another improvement to our search engine is to enable the use of compound queries (Slide 18). Suppose we wished to search on holy man who lives on lentils. We cannot make living on lentils a subtype of holy man as clearly holy women can also live on lentils. Thus we suggest that a compound queries whereby one can search for two or more categories, in this case, holy man and living on lentils. This would necessitate that each entry in the main table be associated with 1 or more categories.

Where to from here (Slide 19).

Version two of the Chrysostom web database shall incorporate the following three modifications.

to revamp totally the type-hierarchy in line with an ontology of the social world of Late Antiquity in order to to allow bi-directional searching and more accurate search results.
to enable searches on >1 homilies at a time.
to allow compound queries to facilitate searches on more than one category.