Ai and the Needle

To historical societies and museums Ai is a new and inquisitive area unfortunately untouched according to this museum’s inquiries — well, certainly in the region of northern New South Wales, so we decided to be inquisitive.

A prelude to our efforts is the Murrurundi and District Historical Society Inc is in the long and laborious process of scanning our region’s newspaper relevance thankfully with past filed copies in some cases going back to the mid and late 1800s.

The process raised the inquiry into handling copious quantities of PDFs, databases and text recognition.

To complicate matters further, being newspapers often the copies were scarcely inked so printed photographs were underdone to say the least.

None of this was any good if the computer has trouble reading the PDF and understanding often blurred text. Just have to look at some of the yields from national Canberra-based databases.

In fairness, scanning techniques, hardware and old database software which was, in its day, revolutionary, governs their output.

The advent of so-called Ai is really the advent of faster computer processing power and programming to match, thereby causing a rethink across a broad range of industries and organisations.

The result so far is lightning fast processors, difference source software and new hardware processing capabilities.

For a little museum at the top of the Hunter Valley with stretched volunteer sources and little money for such experiments, there was still inquisitiveness.

So we embraced the medium with a degree of caution.

Just for your information we tested a little known Ai company called Needle with 10 readable PDFs of varying quality and pics.

Then we asked the questions which centered around family and situation research.

We asked a simple question like What do you have on Mrs June Smith?

To our amazement the results fed back to us included the caption of a very hard to read photograph pertaining to Mrs Smith along with the main points in articles relating to Mrs Smith and copies of the PDF.

Lessons? Our question’s reply picked what Ai thought was the main points about Mrs Smith. She went to Sydney for a holiday, her old man died in 1926 and she had a daughter named Julie.

Fine but with family research the families want to know how long she went to Sydney for? Were there any other children and what were the relevant dates?

We wanted to know finer details to gauge what life was like in those days including with what Mrs Smith wore, did she go to church and what clubs she might have been a member of?

With Ai, it all lies in the way you ask the question as to the results. That, we might add, is almost an art within itself. However, reading the clipping hard copy line for line did fill in details ignored by Ai.

Two lessons, it’s how you ask the question and only use it was a guide.

Secondly and more importantly to date you have to put this up on the web which means anyone can access it.

Herein you have to pay for the hosting and you also have to pay the retraction, which leaves a hole in your budget.

How do you plug the hole?

So far Ai can't give us an answer. However, further research reveals a reconstruction of the database into a different (more numbers and less text) format can be done on the museum’s own computer (space considerations pending and processing power).

This means you can access the data and charge for the research. However, you need younger minds and most of the older working in country research centres have neither the ability or the inclination let alone the computer knowledge.

Looks as if Ai will be in the clouds for a few more years yet.