Letters, Ideas and Information Technology: Using digital corpora of letters to disclose the circulation of knowledge in the 17th century

Availability of the Letters

CKCC limits itself to the ca. 20,000 letters written by scholars that were active in the Netherlands: René Descartes, Hugo Grotius, Constantijn Huygens, Christiaan Huygens, Caspar Barlaeus, Jan Swammerdam and Anthony van Leeuwenhoek. Modern editions of these correspondences—already published or in an advanced state of production by members of CKCC—form the basis of the digitised texts. The letters, once converted to a minimal TEI format, will then be made available through e-Laborate,² a web-based philological annotation tool that will be transformed into a collaboratory for the history of science and the humanities in general. It serves three purposes: (a) providing scholarly access to the letters; (b) allowing researchers to enrich existing datasets and annotate the letters; (c) using the letters and the input of researchers to visualise patterns meaningful for the circulation of knowledge.

Use of other datasets

We will incorporate a particular database of (meta)data, the Catalogus Epistularum Neerlandaricum (CEN), or the Catalogue of letters in Dutch repositories. It is a relatively old database, already available via Telnet in the early 1990s, before the world wide web came into being.

CEN is an exhaustive database of letters in the collections of five Dutch university libraries, the Royal Library, and four other important libraries. It contains more than 265,000 descriptions of approximately 1,000,000 letters, dating from 1600 until the present day (of which ca. 100,000 from the 17^th century). It supplies the following metadata: sender, recipient, place of sending, year, language, repository and shelf mark.

The format in which this database will be made available to the project is to be negotiated with the owner, OCLC.³

Usage of this database will enable us to make assertions about the fraction of the selected letters with respect to the total body of letters. Moreover, it allows us to increase the density of the networks we are interested in, leading to unprecedented research opportunities.

Language technology

In order to find meaningful patterns in social networks of scholars and in circulation of knowledge language technology is needed. For this, CKCC is cooperating with CLARIN.⁴ The mission of CLARIN is to make language technology interoperable and to make linguistic resources accessible on a European infrastructure, so that all the arts and humanities can make use of it. The Netherlands pillar of CLARIN, CLARIN-NL,⁵ has already obtained funding for constructing such infrastructure, and has issued a call for proposals for adding existing resources to this infrastructure and writing demonstrator services. Aided by expertise provided by CLARIN members, in particular by the University of Lancaster,⁶ CKCC is developing such a demonstrator. A proposal to this end has been accepted by CLARIN-NL. The demonstrator, comprising the correspondences of Grotius, Const. Huygens and Descartes (ca. 15,000 letters in all), is planned to be completed by October 2010. It will perform a time-sensitive keyword extraction, which can be visualised by means of a dynamic word cloud. As the source languages are 17^th century Dutch, French and Latin, one needs at least spelling normalisation and harmonisation of keywords across languages.

Interpretation and Enrichment

References to people, places and times are often implicit and can only be retrieved by studying contextual material or by using secondary sources. Named Entity Recognisers are helpful, but it is not possible to rely on technology alone. In order to get an accurate picture in sufficient resolution, interplay between manual work and automatic tools is needed. The collaboratory based on e-Laborate gives researchers the opportunity to collect their interpretations of the texts, compare them to others and to annotate them with their insights. Over time, the results of this hand/mind work might be automatically gathered and incorporated in enriched transcriptions of the texts.

Visualisation

By offering meaningful visualizations of the data, the CKCC will enable humanities researchers in a wider context to use the tools and the results yielded. Not only the relationships between corresponding authors will be made visible in time and space, but CKCC also aims at visualizing the dynamics of knowledge production by focusing on the emergence of themes in scholarly debates and social networks of 17^th century natural philosophers.

The dynamic word clouds based on keyword extraction is just a first step. CKCC will subsequently explore several approaches of gathering and visualising meaningful patterns, which are deliberately different in nature. The first approach (a) is a sophistication of keyword extraction, and the second one (b) is based on associations in text. Both methods can be used to evaluate the results of each other.

(a) Concept analysis. This requires considerable more analysis than keyword extraction. For example: the many surface expressions of a concept must be linked into one entity, preferably part of an ontology or thesaurus. Existing subject indices and reference corpora will be used. Visualising the behaviour of concepts over time will yield a good approximation of knowledge circulation.

(b) Associative neural network technology. This is an application of a recent effort, ANNH,⁷ to apply the idea of neural networks to the humanities. This approach enables the automatic comparison of texts, based on the degree of associative similarity. Concepts or themes occurring in the letters could thus be made visible, either by focussing on single terms (e.g. ‘observations’) or word pairs (e.g. ‘soul’ and ‘matter’). Moreover, it is possible to query for letters associated to a given one, and rank the results by the degree of association. The facility to track the circulation of knowledge will then be within reach.