Dear Arthur,

one possible explanation runs as follows:

(1) Linguistic features: Every text from the DCS is split into sections of equal sizes (say 100 lines). Frequencies of approx. 1,000 linguistic features such as the accusative, certain Sandhi, metre, ... are counted in each of these sections.

(2) Dates: I collect approximate dates for each text from secondary literature. They include "exact" ones, as for some texts after 1000 CE, as well as quite wide temporal ranges for Puranas (and almost all the rest of Sanskrit literature).

(3) Now, I use a mathematical model (neural networks) that tries to predict the approximate dates (2) on the basis of the feature frequencies (1). Basically the same as is done, for instance, for weather forecasting: Collect some information about the weather of today, and try to predict the temperature in the next 24 hrs.

The uncertainties arising from (2) are hopefully handled in a follow-up of the follow-up, which I am currently working on.

Best, Oliver


On 28/10/2018 11:51, Artur Karp wrote:
Dear Oliver, 

As a retiree of - already - some seniority, I have difficulty in ingesting phraseological solids, such as <<neural networks digesting all kinds of linguistic features and making temporal predictions based on them>>.

Could you please explain what hides behind these scholarly terms?

Regards, 

Artur


2018-10-28 9:45 GMT+01:00 Oliver Hellwig via INDOLOGY <indology@list.indology.info>:
Dear all,

just a short PS for the Mbh stratification thread (Andrew, Joydeep and others): A follow-up paper to appear in 2019 (neural networks digesting all kinds of linguistic features and making temporal predictions based on them) seems to suggest that we can distinguish temporal strata in the Mbh. A detail study of the Bhishmaparvan produces results that are not too far away from some outcomes of philological research:
 * BhG + the introductory chapters of Book 6: first few centuries CE (rather 200 CE++)
 * Battle passages quite constantly placed in the centuries BCE

Best, Oliver

---
Oliver Hellwig, SFB 991 Düsseldorf, IVS Zürich

_______________________________________________
INDOLOGY mailing list
INDOLOGY@list.indology.info
indology-owner@list.indology.info (messages to the list's managing committee)
http://listinfo.indology.info (where you can change your list options or unsubscribe)