I’ve got a couple of faculty members who are doing research that involves old newspapers and magazines. The captures are of varied complexity and quality. Some of them are from the nwespaper.com site, some are phone photographs.
I hear that Acrobat’s OCR function does a good job with normal scans of documents. I did a demo video for applying it to all the files in a folder for those who have success with it. I was intending to turn it into a folder action with Automator or Apple Script but it did such a poor job with the example files I had that I decided to test it against some AI options.
The sample scan I used is below. There’s some distortion in the curve of the page. The are three columns which makes it more complex. The fonts are relatively simple.
To test things, this was my prompt.
“Extract the text from the following image. It has three main columns. Process the entire column before moving to the next one. Ignore the heading content.”
A JPG from the newspaper site was uploaded as the source material. The Adobe content was generated from a PDF from the same source.
The results are below. The AI options did far better than Acrobat. Gemini went crazy though and there are weird tweaks that would be easy to miss. You can see the full OCR data here.
|
|
Hand Transcription |
Claude |
ChatGPT |
Gemini |
Adobe OCR |
|
2nd paragraph, 1st sentence |
Denunciation and defense of the educated negro was an interesting feature of the debate preced-ing the vote by which the Murray bill appopriat-ing $2,500 to assist the Halochee industrial in-stitute, a branch of Tuskogee, at Taft, Okla., was killed in committee of the whole in the house today. |
Designation and defence of the educated negro was an interesting feature of the debate preceding the vote by which the Murray bill appropriating $2,500 to assist the Hatchetor Industrial Institute, a branch of Tuskegee, at Taft, Okla., was killed in committee of the whole in the house today.
|
Denunciation and defense of the educated negro was an interesting feature of the debate preceding the vote by which the Murray bill appropriating $2,000 to assist the Halochee Industrial Institute, a branch of Tuskegee, at Taft, Okla., was killed in committee of the whole in the house today.
|
However, The Daily Oklahoman an educated newspaper an interesting feature of the debate precedents it in the state and nation. It has gone to the extent to confront the Halechoe Industrial institute at Ardmore by publishing the full correspondence of committee of the whole to his house in that.
|
• oenunrl.itJon anti tlctense of the educated negro -,11 an l11h’r(‘stlng feature of the debate preced- ll!f tbt~ rntc by which the Murray bill ar~proprlal· tog: S! ;;f>II to nsalst the Ha.locb~e industrial ln- ttltute, n branch of Tuakogoe, at Taft, Okla., was Ulled In (‘ommlttce of the whole Jn tho house lOdar.
|
|
1st sentence under “Let the Faithful Be Our Guard” heading |
There is a well organized political faction in the national democratic party that is try-ing to work up sufficient opposition to pre-vent Mr. Bryan’s nomination on the first ballot. |
There is a well organized political faction in the national democratic party that is trying to force an artificial opposition to prevent Mr. Bryan’s nomination on the first ballot. |
There is a well–organized political faction in the national democratic party that is trying to work up sufficient opposition to prevent Mr. Bryan’s nomination on the first ballot. |
There is a well organized political faction in Mississippi, consisting of those whose aim is to work up sufficient opposition to prevent Senator Money’s reelection, but they will fail. |
‘J’hert’ is n well orgnuized politienl faction in the 11ntio11nl democrutic p11rty thnt is tr.r• ing to work up rmfficit•ut opposition to pre- ,·e11I .\Ir. Hrynn’N nominntion on thl’ fin.t lmllol. |
|
Last sentence |
Secretary Taft ought to study the trust question a little more, or discuss it less, for each speech reveals his lack of familiarity with the subject or his lack or sympathy with the people at large. |
Secretary Taft ought to study the trust question a little more, or please less, for each speech reveals his lack of familiarity with the subject or his lack of sympathy with the people at large.
|
Secretary Taft ought to study the trust question a little more, or discuss it less, for each speech reveals his lack of familiarity with the subject or his lack of sympathy with the people at large.
|
Money is to interfere with the trusts. The question is a little more, or discuss it less, far more momentous than the trusts, and with the subject or his lack of sympathy with the people at large.
|
Secretary Tuft ought to study the truat question a little more, or di1JeuHs it le11 for each F!pccch rev Hls liiM lack of famili~rity with the 1mbjeet or hiH lnek or sympathy with the people ut large.
|
I also did some experiments with more complex newspaper layouts like this one. I didn’t do the full treatment like I did above because I have other stuff to do.
This is a much harder image to read. It’s harder to describe the layout with headlines that span multiple columns, more varied type, decorative elements, etc. The results show that. Breaking out section of the paper was far more successful.