When google just released Ngram, I looked at similar things. ;))
But caution: seemingly promising early results (before 1750) for short words may well be inaccurate and misleading, since old manuscripts used fonts that are frequently misinterpreted by google scanners.
may well be inaccurate and misleading this is true, since the sample is relatively very small (which is seen from the curve's behavior). Another important caveat: smoothing (other than 0) can mislead by implied exaggeration of the size of the sample. fonts that are frequently misinterpreted by google scanners with such one-way consistency? sounds unlikely. True, an old version of s might be taken for f; but I checked out with other words, and the results are qualitatively identical.
Comments 3
Reply
But caution: seemingly promising early results (before 1750) for short words may well be inaccurate and misleading, since old manuscripts used fonts that are frequently misinterpreted by google scanners.
Reply
fonts that are frequently misinterpreted by google scanners with such one-way consistency? sounds unlikely. True, an old version of s might be taken for f; but I checked out with other words, and the results are qualitatively identical.
Reply
Leave a comment