James Boyer's Statistical Studies - more recent studies?

Post Reply
Jonathan Robie
Posts: 4158
Joined: May 5th, 2011, 5:34 pm
Location: Durham, NC
Contact:

James Boyer's Statistical Studies - more recent studies?

Post by Jonathan Robie »

ἐξίσταντο δὲ πάντες καὶ διηποροῦντο, ἄλλος πρὸς ἄλλον λέγοντες, τί θέλει τοῦτο εἶναι;
http://jonathanrobie.biblicalhumanities.org/
MAubrey
Posts: 1090
Joined: May 6th, 2011, 8:52 pm
Contact:

Re: James Boyer's Statistical Studies - more recent studies?

Post by MAubrey »

Technically most of the stuff I've done has been like this, but I tend to get lazy in documenting my data in order to make it available.

The perfect
prohibitions
clitic pronouns
discontinuous noun phrases
vendlerian verb classes

They all exist in poorly organized spreadsheets that I'm too embarrassed to share publicly...sorry.
Mike Aubrey, Linguist
SIL International
Koine-Greek.com
Stephen Carlson
Posts: 3350
Joined: May 11th, 2011, 10:51 am
Location: Melbourne
Contact:

Re: James Boyer's Statistical Studies - more recent studies?

Post by Stephen Carlson »

Jonathan Robie wrote:When working with corpora, it's often useful to compare results to what others have computed. Here are some useful papers by James Boyer:

What other resources do you know of that can be used to query corpora and compare results to known results from other sources? Who has been doing this kind of work in the decades after Boyer?
Boyer is a good illustration of how not to do it. He uses statistics, true, but there are merely descriptive and he has no model for evaluating the results, so statements like "only 37%" are pseudo-scientific.

Though not in New Testament Greek, those interested in corpus studies should read researchers like John Sinclair, Patrick Hanks, Adam Kilgarriff, etc., who worked extensively in corpus studies of English (where the money and the big data are).
Stephen C. Carlson, Ph.D.
Melbourne, Australia
Jonathan Robie
Posts: 4158
Joined: May 5th, 2011, 5:34 pm
Location: Durham, NC
Contact:

Re: James Boyer's Statistical Studies - more recent studies?

Post by Jonathan Robie »

Stephen Carlson wrote:Boyer is a good illustration of how not to do it. He uses statistics, true, but there are merely descriptive and he has no model for evaluating the results, so statements like "only 37%" are pseudo-scientific.
As I mentioned in the OP, what I am looking for is a way to compare results (of treebank queries) to known results from other sources, and Boyer provides some very helpful data for that. I'm very slowly slogging through the process of comparing one treebank to another, and both to published results.

I'm not thrilled with his statistical analysis and probabilistic approach to linguistics either, but that's not what I am using these papers for.
Stephen Carlson wrote:Though not in New Testament Greek, those interested in corpus studies should read researchers like John Sinclair, Patrick Hanks, Adam Kilgarriff, etc., who worked extensively in corpus studies of English (where the money and the big data are).
English results won't be comparable, of course, and aren't easily used for finding and correcting errors in treebanks or providing completish lists of grammatical phenomena in known sets of ancient Greek texts. That's what I'm looking for.

And big data techniques are fairly different from queries on treebanks. In the long run, exploiting more corpus linguistic techniques from these other sources would be great, but I'm not there yet. And working with a small corpus of very well known texts, we may choose somewhat different techniques than the big data folks - even if we add in everything in the TLG, we're not really talking big data.
ἐξίσταντο δὲ πάντες καὶ διηποροῦντο, ἄλλος πρὸς ἄλλον λέγοντες, τί θέλει τοῦτο εἶναι;
http://jonathanrobie.biblicalhumanities.org/
Stephen Carlson
Posts: 3350
Joined: May 11th, 2011, 10:51 am
Location: Melbourne
Contact:

Re: James Boyer's Statistical Studies - more recent studies?

Post by Stephen Carlson »

OK, you're doing something different then. (Basically, debugging your treebank, right?)
Stephen C. Carlson, Ph.D.
Melbourne, Australia
Jonathan Robie
Posts: 4158
Joined: May 5th, 2011, 5:34 pm
Location: Durham, NC
Contact:

Re: James Boyer's Statistical Studies - more recent studies?

Post by Jonathan Robie »

Stephen Carlson wrote:OK, you're doing something different then. (Basically, debugging your treebank, right?)

If you expand the notion of "debugging" to include not only fixing obvious problems, but also making sure that the analysis is sufficiently expressive to represent the kinds of syntactic questions that come up when reading various works, e.g. Rijksbaron's book on the verb, an intermediate syntax, etc.

Right now, we're starting the process of comparing query results with equivalent queries on PROIEL and GBI, and working with various groups to create what we hope will be a better format that will be used by multiple groups. We are in early stages of that. Concrete use cases help us identify exactly what kinds of information need to be recorded, and whether we are recording it accurately.
ἐξίσταντο δὲ πάντες καὶ διηποροῦντο, ἄλλος πρὸς ἄλλον λέγοντες, τί θέλει τοῦτο εἶναι;
http://jonathanrobie.biblicalhumanities.org/
Post Reply

Return to “Syntax and Grammar”