The Royal Society Corpus: From Uncharted Data to Corpus

The Royal Society Corpus: From Uncharted Data to Corpus Hannah Kermes, Stefania Degaetano-Ortlieb, Ashraf Khamis, Jörg Knappen and Elke Teich Universität des Saarlandes Universität Campus A2.2, 66123 Saarbrücken, Germany {s.degaetano, h.kermes, j.knappen, e.teich}@mx.uni-saarland.de, ashraf.khamis@uni-saarland.de Abstract We present the Royal Society Corpus (RSC) built from the Philosophical Transactions and Proceedings of the Royal Society of London. At present, the corpus contains articles from the first two centuries of the journal (1665–1869) and amounts to around 35 million tokens. The motivation for building the RSC is to investigate the diachronic linguistic development of scientific English. Specifically, we assume that due to specialization, linguistic encodings become more compact over time (Halliday, 1988; Halliday and Martin, 1993), thus creating a specific discourse type characterized by high information density that is functional for expert communication. When building corpora from uncharted material, typically not all relevant meta-data (e.g. author, time, genre) or linguistic data (e.g. sentence/word boundaries, words, parts of speech) is readily available. We present an approach to obtain good quality meta-data and base text data adopting the concept of Agile Software Development. Keywords: corpus creation, corpus annotation, metadata 1. Introduction As science developed to become an established sociocul- tural domain from the early modern times, it underwent a process of specialization. We assume that due to specializa- tion, scientific texts exhibit greater encoding density over time, i.e. more compact, shorter linguistic forms are in- creasingly used, in order to maximize efficiency in commu- nication. Examples of linguistic densification can be found at all linguistic levels, e.g. reductions at the syntactic level (e.g. relativizer omission), nominalizations at the morpho- logical level or contractions at the word level. Our assumption is that such densification effects are mea- surable in the linguistic signal in terms of information den- sity, i.e. the number of bits needed to encode a given mes- sage (Shannon information), which is conventionally rep- resented as the (log) probability of a linguistic unit given some context (Crocker et al., 2015). The more predictive a given context, the shorter the linguistic encoding (cf. e.g. variation in word length in Mahowald et al. (2013)) and the fewer the bits needed for encoding will be. To test this assumption, we need an appropriate data set. There are a number of diachronic corpora of scientific En- glish, but these are typically discipline-specific, cover a cer- tain time period only (e.g. the corpus of Early Modern En- glish Medical Texts (EMEMT) (Taavitsainen et al., 2011)) and are fairly small (e.g. the Coruña Corpus (Moskowich and Crespo, 2007) with c. 10,000 words for each discipline in the 18th and 19th centuries). Given that the Royal Soci- ety of London played a major role in shaping science from the mid-17th century (cf. Atkinson (1998)), we obtained a digitized version of the first two centuries of its publica- tions. When building new corpora from uncharted material, typi- cally not all relevant meta-data or linguistic data is readily available. We describe the procedures applied to enrich the base text we use for the …

Content truncated to 500 words