Is CBGM a type of AI?

I am wondering exactly how CGBM works. I have not seen any simple easy to understand explanations as to how it works. Is it using a form of AI? And if not, how does it differ from AI?

Searching the Logos website only returns one book on the topic which is dated 2017. Now that this tool is actually being used to make textual decisions which will eventually influence our English translations, I would think that more people would want to understand exactly how it works.

Does anyone on the forum know of any good resources on the topic? (Good, meaning easy or possible to understand for the average person)

Find more posts tagged with

Resources

Comments

Doc B

LOL, no, not even close.

It is a statistical method of building a "family tree" based on the variants found in manuscripts. And it doesn't come to conclusions, it simply tells the researcher where to look. Good old fashioned text-critical work then has to happen before a textual decision is made.

CBGM is not a process that is simple to explain (if it was easy, it wouldn't take the computational power of a good CPU to pull it off). The most thorough explanations are on the INTF website where the scholar who developed it, Gerd Mink, works.

Perhaps the easiest explanation to read is the book by Wasserman and Gurry, A New Approach to Textual Criticism: An Introduction to the Coherence-Based Genealogical Method. (This is probably the book you found on Logos.)

Don't expect a lot of new books on it, as the procedure is not new…it's been around since the early 90s.

MJ. Smith

Do you mean CBGM (Coherence-Based Genealogical Method)? This is a method that takes advantage of the fact that a computer can deal with more data and consider more possibilities than one can do manually. While older methods (Lachmannian stemmatics) dealt primarily with errors in the text, CBGM deals with both agreements and disagreements in the text. It can distinguish between merely statistical agreement and genealogical coherence that follows a flow of manuscripts. Differences (agree, disagree, derived from) can be assigned directionality based on historic likelihoods. This results in a network rather than a simple tree for the manuscript genealogy. Clustering may show differences that arose independently that were previously considered linked. But the process still requires human judgment/input at various points so it is a tool for humans not independent artificial intelligence. I think of it as enabling what a scholar always wished they could do but could not accomplish in practice - all known manuscripts, all agreements and differences, considering all plausible manuscript relationships … although when one considers fragments especially lectionaries … there is still room for improvement and even more refinement of the results.

Joseph Sollenberger

Disclaimer: My professional training is in the field of chemistry. I have a lay interest in language, manuscripts, and huge curiosity about how the scriptures came about. On top of that add a compelling desire to figure out what the words in these documents are communicating. I approach this topic as a large data source and I am trying to make sense of it. ;)

CGBM is much more akin to genetic genealogy in crime solving than what is happening with Large Language Models (the most common form of AI). In forensic genealogy a document (the DNA analysis) is the known, but the ID of the document is not. By comparing the characteristics of the sample document to other documents (DNA reports) having ID attached, it may be possible to determine the identity belonging to the document under study. This process requires much data searching, pattern matching, and logical puzzle solving. The success of the query depends on the quality and size of the dataset of known documents (DNA reports) matched with identity (name of individual) as well as genealogical information.

The crime solving example is attempting to use past complete documents (DNA analysis plus identity) to generate a new document (DNA analysis) that happens to match a document that has no identity attached. Now imagine the opposite process in which one has complete documents (DNA analysis and identity) and is tasked with working out the family tree without having a full record of who is related to whom. Oh, and the ultimate goal is to generate a document (DNA analysis) of the oldest ancestor possible that is able to explain all descendant documents and variations from the earliest deducible documents to the present.

CGBM is attempting to deduce the earliest manuscript content that could be used to deduce all manuscripts we have based on how manuscripts get changed through processing of creating copies of manuscripts. If you are thinking about how a similar process of change naturally takes place in the transmission of spoken language from one generation to another, check out Grimm's Law relating to developments in Germanic languages.

Change is the only constant in life — Heraclitus