Obtaining high-definition photographs of papers in archives is difficult. Text blurring lowers the accuracy of any transcribing system powered by artificial intelligence. Most people could benefit from upgraded equipment and simple tweaks to camera settings. We provide free advice and can provide affordable archival photography kits. Southampton University used our kit to photograph the Port Books archive (E190) at the National Archives in Kew, London. 60k high-quality images were taken here – and we lived to tell the tale!

In 2021, the London School of Economics provided our first significant data collection challenge. Our objective was to create AI models to extract text from 35k photos of papers detailing historical British trade and format it in Excel. No small feat!

Cust-3 (TNA) is an astonishing dataset that comprehensively captures England's global trade from 1697 to 1780. The list also includes government taxation, which is an important resource for understanding the rise of the British Empire amongst and other significant events.

On the right we see our models identifying and linking tabulated data. Jamaican imports included 'Elephants Teeth’, Tortoise Shells, and Snuff.

We can process much more than early handwritten manuscripts. Historical newspapers offer a wealth of information for scholars. El Debate was a Spanish Catholic-leaning journal with millions of typeset words offering a unique perspective on European modern history from 1911 to 1936.

Charles III University Madrid commissioned us to conduct advanced keyword searches across a quarter of a century of printed material and to deliver results in a comprehensible format. One difficulty was to pick up thousands of line breaks accurately. Many OCR tools do not detect column breakers and so the text jumps across columns.

We are glad to work with a variety of organisations to improving our understanding of archives dating back before the digital age.

The NIC asked us to create new statistics that would show the development and timing of the UK's communications infrastructure.

Our data, shown on the right, shows telegrams transmitted since the beginning, encompassing a period when the world was first connected by electric cables to quickly transmit messages across cities and continents in seconds, rather than days, weeks, or months.

We worked with researchers to extract and match all the individuals listed in over 31000 pages of wounded and deceased German soldiers.

We built segmentation models to capture the layout of these documents and matched the output of the OCR’ed text to exisiting data captured by genealogists from the Genwiki project.

We can extract almost flawless content from challenging sources spanning millennia, in any language, and with any typeface.

Such as handwritten Vatican reports of discussions about important theological issues dating from the fifteenth century. We transformed idiosyncratic Latin into digestible, fluid prose.

On challenge was to eliminate background watermarks to enhance machine reading accuracy. Identifying archaic abbreviations in the text was another.

We built an interface for Aviva to help map entries from historical policy registers. It provides extracts of digitised policy entries with automatic transcriptions, helping users identify, and then map, policy details. The original documents date back to the Hand in Hand Fire and Life Insurance Society, the oldest of Aviva’s heritage companies, with policy registers dating back to 1696. Mapping these historical policies will help Aviva’s specialist archive team learn more about their earliest customers, as well as providing information for family historians and adding to our understanding of how London looked hundreds of years ago. Aviva hopes to make the completed maps publicly available in the future. The tool works by randomly allocating users one of the first 3,240 policies, which relate to addresses in London. The first postcodes were not introduced until 1959 and many houses in the registers did not have numbers, so using details of the parish, road names and other geographical markers (such as ‘west of the Tower’) extracted from the policy entry, users can navigate a historical map of London and mark the exact or approximate location of the policy, as well as amending an automated transcription of the policy wording. Once entries are validated, Aviva’s archive team extracts the corrected transcriptions together with the coordinates for the location of the property, providing a picture of the inhabitants and insurance of historical London. This follows more than two years of work to digitise 150 volumes of historical Hand in Hand policies, covering around 550,000 entries. The details required for working out the insurance premium and identifying the individual property provide a unique window onto London in this period. The tool can be found here: Amicable Contributors