Historic England

Unlocking England's past through AI

The Challenge

Historic England is the public body that champions and protects England’s historic places. Like many public bodies, Historic England has embraced digital transformation and has identified that it has large numbers of valuable, yet underutilised, text and image data assets. Historic England was interested in exploring the ways in which Artificial Intelligence (AI) may be able to boost public access and engagement for some of these assets.

Storm ID was appointed by Historic England to undertake a Discovery project to review some of the incredible data assets that Historic England holds, to identify how AI could be applied to enhance public engagement.

Our Consultancy and Data Science Teams rapidly evaluated and scoped the most significant opportunities to leverage AI, both to improve the quality of digital engagement with audiences, as well as support internal process efficiency.

We reviewed the Historic England Archive, which contains 12 million photographs, drawings and report records relating to the archaeology and architecture of England. We also reviewed the National Heritage List for England (NHLE), England's official database of protected heritage assets, containing 400,000 records as well as England's Places Collection and aerial photographs covering the whole of England, dating from the early 20th century to the present day.

From the review, an initial twenty AI use cases were identified. These were ranked using feasibility and value for public engagement variables. This enabled a short list containing the ten strongest AI candidates to be established. 

The Objectives

  • Identify Challenges and Opportunities

    To identify the most appropriate use cases for AI for public engagement in Historic England we first looked at the key challenges and opportunities within the organisation relevant to AI. To do this we undertook one-to-one stakeholder engagement sessions and workshops, analysed and experimented with several sample datasets and reviewed a number of reports and strategy documents to better understand organisational objectives.
  • Identify Value and Feasibility

    In such a data rich organisation, much of the challenge would be identifying what applications of AI could deliver greatest value to Historic England’s public engagement, with ease of implementing. In terms of value, consideration was given to understanding if a task is large and repetitive enough that it could be delivered more cost effectively by AI than humans. In terms of feasibility, consideration was given to the quality and quantity of the data available. 
  • Use Case Prioritisation and Proof of Concept

    With our assessment of the possible applications of AI, we then set about prioritising quick wins which could be implemented with relative ease with quick benefits realisation too. We identified ‘strong candidates’, these being priority areas for Historic England to focus on over the medium term. We also identified use cases where we could quickly develop proof of concepts to demonstrate AI value. 

The Insight

Historic England has very valuable text assets, both structured and unstructured, most of which are not being utilised currently. These assets can be intelligently indexed using Natural Language Processing (NLP) techniques to allow for easier searching, asset linking, and increased discoverability. Historic England also has a significant volume of image assets, which could be enriched using, for example, Azure Cognitive Services Computer Vision, allowing richer tags to be added to existing tags to improve discoverability.

We identified opportunities for multiple AI uses cases on the England's Places collection, an important, digitised collection of photography from 1850s to the 1990s from eminent and amateur photographers. The photos, containing handwritten information on the backs, were placed on boards. The boards also contain handwritten text. The current format of these digital images is not well-suited to viewing online and it is difficult to read the associated text.

We developed a proof of concept centred around the England’s Places collection to demonstrate image cropping of the photos, text extraction of handwritten text and generation of auto tags from the text extraction. The proof of concept has helped to illustrate how AI can be applied to improve content discoverability while also applying automation to the lengthy manual process of digitally recreating the handwritten text on the board and photos.  

Key Features

Custom cropping

We generated custom Python scripts to automatically detect the boundaries of photos mounted on boards, enabling automatic extraction of individual photos from boards.

Image tagging

We used Azure Cognitive Services Computer Vision to generate tags and captions from the images, as well as confidence scores for the captions.

Text extraction

We used Azure Cognitive Services Optical Character Recognition technology to read and extract the handwritten information on boards and backs of photos.    

The Process

Consulting and Data Science

  • Understand user needs and business goals.
  • Engagement with stakeholders through interviews and workshops.
  • Review of existing data sets for value and feasibility.

POC Development

  • Undertaking development of AI proof of concept levering combination of Azure Cognitive Services and Custom AI scripts.  

Report and Presentation

  • Discovery report, proof of concept and recommendations on next steps to operationalise AI through digital services. 

Get in Touch

If you're considering how Artificial Intelligence (AI) and Machine Learning (ML) could be applied to your digital services, get in touch and we can discuss your requirements.