[B-Greek] The Unreasonable Effectiveness of Data: An Argument for a Large Corpus Size (i.e., Reading WIdely)
Elizabeth Kline
kline_dekooning at earthlink.net
Thu Mar 25 17:26:11 EDT 2010
On Mar 25, 2010, at 1:04 PM, John M. Linebarger wrote:
> OK, this one may seem to be coming out of left field, but let me make my case. Here is a URL to a paper called "The Unreasonable Effectiveness of Data" in IEEE Intelligent Systems. Each of the authors works for Google. "What in the world does that have to do with B-Greek?", you may be wondering. Well, a theme of the paper is that in a world characterized by enormous amounts of data, more traditional rule- or heuristic-based approaches to translation or semantics are being replaced by statistical ones, which perform much better. In other words, in the presence of absolutely huge amounts of data, correlations become amenable to probabilistic solutions.
>
> To me, this provides yet another confirmation of Carl's constant dictum that in order to know Biblical Greek well, one must read widely *outside* of Biblical Greek. In other words, the bigger the corpus as the context in which to understand Biblical Greek, the better. In this argument, the human becomes the probabilistic information processor, and does increasingly better pattern matching and statistical correlations as the corpus size of Greek that has been read increases. Rules may bootstrap the process, but the context of a huge corpus trumps rules with statistics.
Yes read everything and particularly read difficult texts, like a chorus from Sophocles or speeches from Thucydides. Reading simple narrative, where you can guess in advance what is going to be said based on the preceding context and the majority of verbs are familiar action words, this sort of reading does not force you to push out beyond the walls of intermediate grammar and semantics.
Elizabeth Kline
More information about the B-Greek
mailing list