Engineering Historical Memory

Since 2007, Engineering Historical Memory (EHM) is studying and practising “by what means” traditional historical scholarship can supply machine-readable information sets to empower historical sciences with artificial intelligence and machine learning, thus enabling all users to read primary historical sources according to different levels of knowledge and expertise interactively. In the history domain, EHM makes a cross-disciplinary use of established research processes, such as mapping as understood in mathematics and linguistics (i.e., an operation that associates each element of a given set, the domain, with one or more items of a second set, the range) and parsing as understood in computing (i.e., analyse narratives into logical syntactic components) to kick-off the exploration of primary historical sources. Using these operations of mapping and parsing for individual primary historical sources, EHM associates each element of given sets of information provided by the domain of the traditional disciplines (e.g., history, art history, philology, palaeography, diplomatics, codicology, archaeology, epigraphy, sigillography) with one or more elements of the range of machine-readable content management systems (e.g., spreadsheets, computational notebooks). The level of accuracy of this preliminary human activity is directly proportional to that of the aggregations generated and visualised by the EHM algorithms from different sets of similar written or depicted elements in the EHM database (e.g., geographical names, people’s names, goods, ships, governments, events, architectures, drawings) and from potentially relevant publications, images, videos, and news retrieved in online repositories.

The EHM approach to history can be construed as a hybrid human-machine methodology because it relies on both human scholarly touch and machine computational power. The attributes of the EHM methodology (i.e., set of methods) can be named as follows.

Analytic, because of the scholarly mapping and parsing of information from primary historical sources.

Synthetic, in reason of the interactive visualisation of selected information.

Exploratory, because of the automatic search for online publications, images, videos, and news potentially relevant to the user’s choices.

Aggregative, as far as it allows interactive selections and visualisations of different sets of search results.

Non-narrative in principle, because the organisation of the materials into narratives is up to the user who generates gamut accordingly.

Methodological foci

  1. Philological accuracy
    Goal: correctly transfer into digital database information already published, without diluting its philological accuracy
    Plan of action: collaboration with authors and publishers of critical editions and translations to access the most accurate scholarship available on crucial primary historical sources
  2. Provenance and validation
    Goal: provide a clear understanding of the information’s provenance and a robust validation method to assess its accuracy
    Plan of action: link information from secondary literature to the primary source on which it is grounded
  3.  Public participation (i.e., citizen science, crowd-sourced science)
    Goal: generate volunteer monitoring on the information published on EHM
    Plan of action: create a community to constantly update the information available on the EHM database (curated by professionals), and connect the information published on EHM to the relevant Wikipedia pages (curated by a larger pool of scholars and amateurs)