“The Scriptural BCP”
Newsletter editor Diana Verseghy recently interviewed Arlie Coles, the developer of a new online resource designed to demonstrate how thoroughly Scriptural the Prayer Book is. Arlie tells how this project grew out of her interest in linguistics and computer science.
Please describe this new project for us.
“The Scriptural BCP” is a digitization and web upload of an 1839 publication, The Book of Common Prayer: With Marginal References to Texts in the Holy Scriptures, which was produced by the Society for Promoting Christian Knowledge (SPCK) of the Church of England. It is basically an annotated copy of the 1662 BCP. Alongside each collect, exhortation, canticle, versicle, etc., there is a column containing exhaustive chapter-and-verse citations of all the scriptural passages that are referenced in it. However, the original book is not very user-friendly, since readers must look up each of the numerous references themselves and attempt to hold them in their minds while considering the connection to the BCP text. The online format of “The Scriptural BCP” instead allows the user to click on the text of the BCP content and at once read all the related scriptural references together. It can be found here.
Wow, what inspired you to take on such an enormous task, and how did you accomplish it?
It actually wasn’t as difficult as it might seem, and it was really quite an absorbing project! To backtrack a bit: some time ago I was helping to design a new service booklet for my parish in Montreal with the goal of adding some gentle explanation — the parish in question (St. John the Evangelist’s) is quite high-church, and knowing when to stand, sit, kneel, as well as why who is doing what, can be a bit opaque for a newcomer. I showed a draft to my mother in Texas, who had become an Anglican coming from a non-liturgical tradition. She suggested that one way to make the booklet more accessible would be to add a few scriptural references, since people new to high-church liturgy might be wary of unjustified ritual. I thought this was a good idea, remembering for example the clear quotations of Isaiah chapter 6 in the Sanctus, of Matthew chapter 15 in the Prayer of Humble Access and so on. For fun and to jog my memory of anything else obvious to add, I went looking for any BCP annotated with scriptural references and found the 1839 book, which had already been scanned and made available online at archive.org. But the quantity of references was just overwhelming! Plus, the book was hard to use, since the reader had to look up each reference in turn.
It struck me that this was a fantastic resource in a non-ideal medium (of course, that was all that was available at the time it was written). I had a good enough grasp of some computer science techniques that made a transfer to the web easy. The only manual work involved was transcribing the marginal citations of the 1839 book. (I tried to make use of machine learning, that is, digital pattern recognition techniques, to process the text automatically, but that was unsuccessful due to typos, inconsistencies and ambiguities in the original book.) However, once I had done this, it was quite straightforward to program software to use the citations to fetch copies of the Bible texts from online sources, and generate web pages from this. So to me, the project was just the right intersection of things I love and want to perpetuate, and things I knew how to do.
Tell us a bit more about your background. How did you come to acquire your enthusiasm for the Prayer Book tradition?
Having had parish homes first in the Episcopal Church of the USA and later in the Anglican Church of Canada, I can say that the BCP heritage has made a great difference to my life. I am a cradle Episcopalian, and was shaped by the Prayer Book without knowing it. As a very young child I grew up in a parish in Texas that used Rite II (that’s the contemporary-language rite) of the 1979 American Prayer Book, and from there I climbed the “high-church ladder” as a matter of my own taste, gravitating towards Rite I. And we were given the space to climb that ladder, and there was always instruction about how to do it. I came to admire how the BCP held the high and low church parishes together, and Rite I opened up the history of the whole prayer book tradition to me. In retrospect, I really didn’t realize how fortunate I was to grow up in such a supportive environment! And for the past several years that I’ve lived in Montreal, I have been a member at the Church of St. John the Evangelist, which uses the 1962 Canadian BCP as the basis of its Anglo Catholic liturgy. I love it.
And how did you end up coming to Montreal?
Well, McGill University had exactly what I wanted to study. They had a really excellent programme for computer science together with linguistics. And there was also a financial incentive; it was actually cheaper to do this great programme in Canada than it would have been to find something of similar quality in the States. After I earned my bachelor’s degree, I discovered an outstanding M.Sc. program in machine learning on offer at the University of Montreal. So it was the right place, right time for me to continue on in grad school in the same vein. And after I graduated I landed a job as a research scientist at Nuance Communications, a company with a longstanding presence in Montreal, where I work in deep learning research for speech and language applications.
To digress for a few moments – many readers of this article will not be familiar with the terms “machine learning” and “deep learning”. Could you just briefly explain what they are and how they work?
Sure. Machine learning is a form of artificial intelligence that involves building a computational tool to handle huge volumes of known data, automatically learn its patterns, and then apply what it has learned to unknown data to make inferences about it for a specific task. A tool that has been well “trained” on large volumes of data can notice subtleties or common elements in new data that humans might have more trouble seeing. For example, it has been used to help identify which works were written by the same author. What makes some machine learning “deep learning” is the use of certain “layered” algorithms, such as neural networks, that can learn to detect more and more complex information – patterns within patterns. The area that I specialize in, which is connected with linguistics, is actually a very large sub-field within the deep learning / artificial intelligence world. I work at a company that does a lot of business in automatic speech recognition, specifically for the medical context. So for example when a doctor and a patient meet for an appointment, they can record their conversation and then pass the recording on to the technology, which can then separate out what the doctor and the patient each said, and generate all the required paperwork. It leaves more space for the human connection.
You mentioned earlier that you had tried to apply machine learning to your “Scriptural BCP” project. A tool like that could also be used to produce something like Biblical concordances, couldn’t it – you know, those huge books that used to contain enormously detailed Scriptural cross-references?
Yes indeed. I do think there’s a large overlap between what we can do with this type of technology and what we can do for fields that have a long textual history. The Christian tradition is one of these, and in our Anglican world for example we have the Biblical texts and the Prayer Book texts that are all intertwined. As I said, one thing that the deep learning lets us do is handle enormous quantities of text and extract information about them in some way. I haven’t really been able to dig into as many side projects on this as I would have liked. But I think there’s a lot of unexplored territory as to what AI could do in a church context, by way of pointing out links between things, or just helping us study them better. For example, there’s a variety of theological themes that could be explored – it may be interesting to train a machine at some point to take a particular allusion in some part of a text and delve into what theological connections it may have to other texts that on the surface might not be obviously related. Machines can tell us things like this if they’re constructed and trained in the right way. So that’s one example.
Your other area of specialization, alongside computer science, has been linguistics; does that also resonate with your appreciation of the Prayer Book?
Very much so. As a linguist I’ve gained a deep appreciation for the importance of one’s native language. People have a special relationship with their mother tongue. You know, we are built to do language. When babies hear the sounds of human speech, those sounds are literally built into their brains. If certain sounds aren’t heard on an ongoing basis, as in the case of a different language that they only hear for a while, those sounds can get pruned away, labeled as “not my language”, at very early points in the development of their brains as they grow. But the sounds of the mother tongue really get baked in, in a significant biological way. So it’s not just my aesthetic interest or emotional response that makes me enjoy hearing the liturgy in English; it has a core meaning for me exactly because of that visceral attachment to it. The Reformation impulse to get things into the vernacular has really come alive to me; I’ve realized what an intimate gift it is, facilitated by people like Tyndale and Cranmer, to worship God in my own language using some extremely well-trod paths — a tradition that, in adulthood, I better understood that not all have had! I firmly believe that the Prayer Book tradition is unifying, intimate, and well-justified.