Mobius digitally transcribes 25 million Scottish heritage records with 99.56% accuracy

The Business Need

The Digitising Scotland project was commenced with the objective of digitizing up to 25.1 million Scottish record images that carried in it critical event records of births, marriages, and deaths dating way back to 1856. Funded by the Economic and Social Research Council (ESRC) and National Records of Scotland (NRS), the Digitising Scotland project is seen as a means to ease the cumbersome efforts taken by a researcher to extract information from the records in the images.

One has to pore through the record images of about 18 million individuals, to locate an individual’s specific record and then manually copy down the information they need, which makes it an increasingly painstaking process for a researcher embarking on a large-scale research.

The Digitising Scotland project is a collaboration between 3 institutions - the Universities of Edinburgh, St.Andrews and Queen's University Belfast, with the University of Edinburgh creating a multidisciplinary research database from the 14 million birth records, 4 million marriage records and 11 million death records between the years 1856 to 1973. Mobius being a major contender in digitization was approached with this high-volume transcription of heritage material that held historical significance.

Challenges we faced

A cost-effective solution aiming 95% accuracy had to be deployed to digitize the images that held handwritten records of varying levels of legibility. Owing to the veracity of data, we realized that complete digitization could not be accomplished by a single tool and sought to our adept human expertise. Comprehending the style and nuances of the written language which goes back to almost a century, proved to be an interesting hurdle for the team.

How we solved the problem

A 115-strong team of data entry specialists headed by 2 able project managers took up the novel digitization project of 25.1 million handwritten registration records. Attributes of family members such as name, age, gender, occupation, place of birth, the cause of death and more had to be captured from the Census records.

Though the records were of questionable clarity, the team assessed 98% of the images to be readable and aimed for high accuracy in their transcription. Hence the digitally transcribed output was run through various levels of rekeying, spot checks, and external validation. A Process Excellence Group (PEG) was exclusively set up to check the consistency and precision of every digitized text output. The record images were not contained in any on-site storage areas, thus adhering to meticulous data security measures.

"After a competitive tender process, Mobius was awarded the contract to transcribe the 25.1 million handwritten Vital Registration Records for Scotland as part of the ESRC-funded Digitising Scotland project. The contract includes the transcription of all births, deaths, and marriages from 1856 to 1973. Most of the images from which transcription was based were taken from pre-scanned image copies of variable quality. Nonetheless, accuracy records to date are 99.56%, well above the 97% contractual requirement and an industry best for this type of content. Each transcription project is carried out by the skilled transcription team at Mobius. Upon completion, a variety of approaches are used by Mobius to verify the transcription including some re-keying, spot checks, and external validation. This highly viable transcription model has proved to be both reliable and cost-effective for large-volume digitization"
Elaine Reid, Project Manager - The Centre for Data Digitization and Analysis,
School of Natural Built Environments, Queen's University Belfast

Results

Exemplary digitized outputs of consistently high accuracy levels of up to 99.56%, way above the contractual standard of 97% were achieved in a fast and reliable manner. While 39 digitized records per hour were expected on an average, we delivered 47 records pushing the productivity levels by about 21% per hour. Achieving top-notch quality in digitization with an optimally numbered proficient team, earned us accolades from our client.

25.1
million heritage records

Dated from 1856 AD to 1973

Handwritten and illegible image files

Digitized output delivered with high accuracy levels of 99.56% in record time

Looking out for a similar solution?