More AI for local historians

One of the best sources for historians of the 19th and 20th centuries, be they family, local or academic, are the newspapers from those years. The problem is that images with which sites such as Ancestry and Find My Past work are often very poor, especially for those from the 19th century.

The National Library of Scotland, which is at the forefront of research in the use of artificial intelligence (AI) in libraries and archives, tested the transcription accuracy of AI against the optical character recognition (OCR) program it normally uses on newspapers from its collection.

Using OCR returned an accuracy rate of just 60.1%, whereas AI managed 93.9%.

I decided to test ChatGPT and the OCR program I normally use (PDF24) on transcribing the newspaper cutting of the letter from Daniel Mayor that featured in a post yesterday.

This is PDF24โ€™s OCR transcript of the cutting:

mmm mmm

SUFFRAGISTS AND THE PRESTON POLICE

[TO THE FHITOR

OF THE โ€œ LANCASHIRE DAILY POST.โ„ข) Sin, โ€”~1 see from your report of the suffragist

meeting on Saturday night โ€”which, | am told by *

| supporter who was present, 1s a perfectly correct

, | report โ€”that Muss Hewitt denied that Mm. Rigby|

and the other lads with her obstructed the police

in the discharge of their duties. Allow me w say that | was with these ladies from the time they need. lessly and foolwhly forced themselves into the thickest |

| of the crowd at the twp of Wharf street until their

, arrest at the bottom of Fleet-street, and that they|

| obstructed and defied the officers of the Chief Con. ‘.

| stable practically all the time. Where Miss Hewitt { | was at this time 1s not known to me. {

| The exhibition they made of themselves wes such that | have seldom in my life been so ashamed and humbled gs on that occasion. The patience, and I will even say the gentleness, with which the pole

| treated them was beyond praise, and when at last , โ€˜they did arrest them, | felt relieved and thankful –

mot only for the sake of the public peace, but for| their own sakes

Laterally โ€”and | mean exactlyโ€” they compelled the police to arrest them; and now, forsooth, they wish |

(10 pose as martyrs in a glernous cause! Yours, &c., |,

K

}

w Ww “Fy

DANIEL MAYOR, SEN. 6, Wharf street, Preston, Dec. 14th, 1909.

And this is ChatGPTโ€™s transcript:

SUFFRAGISTS AND THE PRESTON POLICE

[TO THE EDITOR OF THE โ€œLANCASHIRE DAILY POST.โ€]

Sir,โ€”I see from your report of the suffragist meeting on Saturday nightโ€”which, I am told by a supporter who was present, is a perfectly correct reportโ€”that Miss Hewitt denied that Mrs. Rigby and the other ladies with her obstructed the police in the discharge of their duties. Allow me to say that I was with these ladies from the time they needlessly and foolishly forced themselves into the thickest of the crowd at the top of Wharf-street until their arrest at the bottom of Fleet-street, and that they obstructed and defied the officers of the Chief Constable practically all the time. Where Miss Hewitt was at this time is not known to me.

The exhibition they made of themselves was such that I have seldom in my life been so ashamed and humbled as on that occasion. The patience, and I will even say the gentleness, with which the police treated them was beyond praise, and when at last they did arrest them, I felt relieved and thankful not only for the sake of the public peace, but for their own sakes.

Literallyโ€”and I mean exactlyโ€”they compelled the police to arrest them; and now, forsooth, they wish to pose as martyrs in a glorious cause!โ€”Yours, &c.,
DANIEL MAYOR, Sen.
6, Wharf-street, Preston, Dec. 14th, 1909.

I would say thereโ€™s no contest. I will be using ChatGPT in future. If I had had it over the past few years, it would have spared me hours and hours of work correcting transcripts or simply having to rekey the texts.

Source:ย ‘Navigating Artificial Intelligence for Cultural Heritage Organisations’ pp 195-8.ย 


Discover more from preston history

Subscribe to get the latest posts sent to your email.

One thought on “More AI for local historians

  1. The following script produces two impressively accurate transcriptions of an image, I tried it on newspaper adverts. In ChatGPT terms the ‘diplomatic’ version retains the original look should you wish to recreate a transcribed version of the original, whilst in the ‘normal reading’ version the text flows continuously.

    “Convert image to text, exact line-for-line diplomatic transcription, retaining original capitalization, punctuation, and hyphenation exactly as printed, with line breaks matching the image. Provide a normalized reading text parallel to the diplomatic version, retaining original capitalization, punctuation, and hyphenation.”

    Hope this helps.

Leave a Reply