1

Explicit Question (TL;DR): I exemplify with English, because it has changed more than e.g. the Romance Languages, for a given time period. For brevity, I use the noun 'disquisition' to mean any decent, reputable piece of writing.

What linguistic concepts reveal how a disquisition (from one time period), differs from another (written from another time period)? To wit, in which linguistic concepts do changes in writing differ?

Optional Reading and Supplement: I reify the above: Suppose you encounter a disquisition with no background information (ie: the author, context, or date). As you start to read this disquisition, you will be able to estimate its time period, spontaneously and subconsciously. (You may not able to distinguish English writing in the 1500s from that in 1600s, but you can probably distinguish writing in Early Modern English from Late Modern English (1800s).) Please help me understand this subconscious ability: What precisely enables the reader to date a disquisition?

For example, distinction is helped by:

diction: former mandatory study of Latin (and Greek) explains the prevalence of Latinate over Germanic words in writing of the 1700s.

syntax: sentences were longer in the past.

but not grammar: authors may use different verbs, but tenses remain the same, except the past's preference for the subjunctive mood and literary tenses (eg French, Spanish).

1 Answers1

1

The question is very strangely worded, but here's my try:

You can train any decent machine learning algorithm (e.g., a Support Vector Machine) to recognise the time period where a text was written. Describing the text as a "bag of words" (i.e., transforming it into a a frequency list) is fully sufficient for that task. You can normalise the text to modern orthography and grammar without loosing much of precision and recall.

Adding more features (like sentence length, n-grams, etc.) improves precision and recall.

Sir Cornflakes
  • 30,154
  • 3
  • 65
  • 128
  • It seems like this questioner wants to know how humans do it, but I'm personally happy with the hypothesis that it's a similar statistics-tracking process. – Jeremy Needle Jul 07 '15 at 13:08
  • @JeremyNeedle Thanks, yes, I wish to know humans determine the changes. –  Jul 07 '15 at 14:33
  • @JeremyNeedle To jknappen too: Please feel free to edit my OP to improve it. –  Jul 07 '15 at 14:34