On 24-Feb-10, Lars Aronsson wrote:

[ENGLISH SUMMARY: How should we build a better website for
 large size newspapers? Is it possible to link to a position
 in a large image? Which websites have solved this problem?]

Om man länkar till en sida i Sundsvalls Tidning, så har man sex spalter att leta i, innan man hittar den artikel eller annons som man ville länka till. Lät mig ge ett exempel på problemet:

Hitta artikeln "Häftig snöstorm i Newyork" på den här sidan, http://runeberg.org/sundtidn/1888/0143.html

Det går inte helt fort att hitta. Artikeln finns i spalt 4, nedanför Stanley-expeditionen, Ryske tronföljaren och Sur mjölk.

I en vanlig webbsida (HTML) kan man sätta in "ankare" och länka till dem med # i URLen. Men hur gör man för att länka till en viss position i en stor bild? Finns det några bra lösningar för detta? Några förebilder?

Kort sagt: Finns det några webbsajter med inscannade dagstidningar, som har lyckats lösa detta pä ett bra sätt?

Detta utgör inget problem i vanliga böcker och tidskrifter, i små tidningar.... Det är når man kommer till dagstidningar i stort format med 6 eller 8 spalter som problemet ger sig till könna.

Dear Lars,

[SVENSKA SAMMANFATTNING: Korrekturläsning af tidningens plakatsidor utgör en särskild utmaning. Karakter-baserade PDF-filer kan innehålla lokala lenker, men bildbaserade PDF-filer kan icke. Arbeta i en textredigerare och Acrobat kan vara enklare än att använda Runebergs gränssnittet.]

As an exercise, I proofread this page last night and uploaded the result to the Runeberg project. I agree that it is not easy to deal with such large pages. I simplified my task by copying the entire page of OCR text into a text file which I enlarged to 16 point Courier (so that I could easily tell the difference between capital letter I, lowercase letter l and the numeral 1, for instance).

The image was, indeed, a very large file to look through. Other Runeberg proofreaders may also try making a local copy of the page scan (right-click on the Print/PDF button near the top left of Runeberg's page) and viewing it from within Adobe Acrobat. This can be helpful, especially with viewing the small type, since Acrobat allows for much bigger enlargements than Runeberg's default maximum of 200 percent.

I was pleased to see that the OCR text was arranged in logical order as opposed to strict physical column order. That was definitely helpful. Since some preprocessing of the OCR text must have been necessary to achieve this, perhaps the scanning person could divide up future page images into smaller chunks, each representing a complete article or a group of ads and/or filler material. I think this is how of copies of New York Times articles from the 1800s appear when one does a Google search for them. (Unfortunately, it seems that access to such archived PDF articles now requires a paid subscription.)

The alternative might be to arbitrarily break up future page scans (before OCR) into perhaps four smaller PDFs, representing the top and bottom halves of three columns each. This has the advantage of making the proofreading task more manageable (since each sub-page can be completed in less time and with less demand on one's computer), but also has two disadvantages. First, the text files resulting from sub- page scans might not be complete articles, since it is possible for an article to run the whole height of a column or to span the two middle columns, and thereby not be complete on a single sub-page. Secondly, there would need to be some overlap, at least in the top and bottom sub-pages, which would result in a little bit of duplicated effort as well as subsequent processing to combine the scanned and proofread text from two or more sub-pages into a single text file.

Other Runeberg proofreaders may also try making a local copy of the page scan (right-click on the Print/PDF button near the top left of Runeberg's page) and viewing it from within Adobe Acrobat. This can be helpful, especially with viewing the small type, since Acrobat allows much bigger enlargements than Runeberg's default maximum of 200 percent.

Using a page makeup program such as Adobe InDesign or Quark XPress, one can produce a compact character-text-only PDF or a text-and- imported-illustrations PDF that replicates the look of the original scanned-image-only PDF. With a text-based PDF generated from such a typographic page layout program, it is easy to incorporate any desired links. However, I know of no way to link to particular areas of a scanned bitmap-only PDF, which is really only a wrapper for a downsampled TIFF or JPEG file produced by the scanner.

Proofreading broadsheet newspaper pages represents a special challenge; it may be that the complexity of the task cannot be readily eased. Good luck to those who volunteer to help.

Kind regards / Med venlig hilsen, Erik Bjørn Pedersen