Skip to main content

Making a Modern Archive

Search This Blog

Contact

Get link
Facebook
Twitter
Pinterest
Email
Other Apps

Contact:

Questions regarding the Making a Modern Archive -project can be directed to Maria Kallio at Maria.Kallio(at)arkisto.fi

Get link
Facebook
Twitter
Pinterest
Email
Other Apps

Comments

Post a Comment

Labels

19th century2
court records4
demonstration2
digital history2
HTR5
internationality1
KWS2
MaMA2
meetings1
NAF8

READ8
READ-COOP1
records1
resources2
search engine2
search engines1
seminar2
students1
Transkribus3
University of Helsinki1
user interface2
war journals1

Show more Show less

Popular Posts

Resources to transcribing Old Swedish

As mentioned in the previous blog posts, one of our main goals at the moment is to develop an HTR (Handwritten Text Recognition) model that can fluently read 19th century Court Records. In order to achieve this we must produce a lot of “teaching material” which we call GT (Ground Truth). Basically we have to transcribe lots of documents so currently me and my co-workers spend most of our working hours transcribing 19th century Court Records. As explained in the blog post “Renovated District Court Records”, most of the Court Records are written in Swedish (sometimes in Finnish also). In this blog post I will go through all the resources, which have helped me as I have transcribed Old Swedish. Most of the tips I give will only apply to Swedish. However, I hope that this post can also give ideas to the readers that don’t use or need help with Swedish. About the Transcribing and making GT Transcribing text that is difficult to read is a time-consuming job. When you are doing a research

Keyword spotting – an effective search tool

The subject of making full text searches possible in district court records was touched on the blog about Turku book fair some time ago, and now it’s time to take a deeper dive into how exactly the full text searching works. The key to the search tool is keyword spotting (KWS). Keyword spotting technology uses confidence matrices to find matching images to your search term. The confidence matrices are based on the HTR model run on the text you want to search: The HTR model assigns each letter in the alphabet a value of how certain the model is that that letter corresponds with that part of the image (ie. a section of a word on a row) when you run the model on a page. The transcription the model outputs is essentially the letters and words the model is the most confident match the image. High confidence result for the search "äktenskap". Low confidence result for the search "äktenskap". The confidence is low because the found word is actually "

HTR Models

To get the massive amount of text within the 19 th century renovated court records into a computer readable form, we use Transkribus to train HTR models to transcribe automatically the hand written text in the court records. Talking about HTR models can turn into a sea of acronyms pretty quickly, so it would probably be best to start with some key terminology, although some of them have already come up in previous blogs. HTR, or Handwritten Text Recognition, is the technology used in generating the automated text recognition models in Transkribus. Unlike Optical Character Recognition (OCR), HTR processes the text by line of text and not by character, and it uses artificial neural network technology. Ground Truth or GT refers to the set of human transcribed pages used in training of the HTR models. CER and WER are used to measure the accuracy of the HTR model, CER meaning character error rate and WER meaning word error rate. These show the average percentage of character

READ project’s final meeting and the Transkribus seminar in Helsinki

A final group photo of the international READ project staff at the National Archives of Finland on June 27 th . The last couple of weeks have kept us especially busy at the National Archives. As you may already know, READ as an EU funded project has now reached the finish line, and this has meant both summarizing the achievements so far as well as turning the focus on the future. Right after midsummer, starting from Wed the 26 th of June, NAF gathered our international READ partners to Helsinki for a three-day meeting and seminar spree. This provided the opportunity to thoroughly discuss the project’s three-year journey among the partners, but also to present some of its most central results to the public. The meeting days kicked off with a Transkribus Seminar that was open for the public to attend. There were around 100 attendees total, and the seminar guests represented a diverse set of organizations ranging from libraries, museums and archives to universities, informat

Renovated District Court Records

One of our main goals at the National Archive of Finland as a participant in the READ project is to develop HTR (Handwritten Text Recognition) model that can read 19th century Renovated District Court Records (in Swedish: renovedare domböckerna; in Finnish: renovoidut tuomiokirjat). They are written mostly in Swedish but some parts are in Finnish. Currently the HTR models made with the most training data have a CER (Character Error Rate) of around 3 %, so we have almost reached our goal. However, we still need to improve our HTR model’s ability to read court records from the early 19th century. In this blog post, I will introduce the Renovated Court Records to our readers. How have they come into existence? What do they contain? Why did we choose them to be our pilot material in the READ project? Volumes of Renovated Court Records from the rural district of Heinola Court Records from the geographical area of Finland start as a continuous archival series from the year 1623, wh

Powered by Blogger

National Archives of Finland