A Parsed Historical Corpus of Modern English

University Of Pennsylvania, Philadelphia PA

Investigators

Abstract

With National Science Foundation support, Dr. Anthony Kroch and Dr. Beatrice Santorini will create a one million word syntactically-annotated electronic corpus of Modern English texts and text samples, covering the years from 1700 to 1900 C. E. This corpus is the fourth in a series created over the past decade by researchers at Penn and at the University of York, England. The three existing corpora cover 900 years of the history of English (ca. 800 C. E. to 1700 C. E.) and comprise more than 4.5 million words of running text, tagged for part of speech and annotated for syntactic structure. The corpora support studies that combine grammatical analysis with the statistical tracking of changes across time, at a level of detail and precision never before possible. As a result, they have been used in the study of many changes that the English language has undergone over the centuries. However, the linguistic changes that distinguish Modern English from the language of earlier periods only go to completion in the 18th and 19th centuries, so that the new corpus is needed for temporally complete investigations of the rise of Modern English. The new corpus will reinforce in several ways the growing influence of electronic corpora on the language sciences. First, the study of language change is being revolutionized, because corpora provide, in manageable form, the data needed for large-scale and precise investigations. Because the corpora are publicly available, common data resources, different analyses can be compared with a confidence and precision not previously possible. Second, the corpora permit historical work to interact more effectively with other areas of linguistics. Thus, recent discussions of the long-assumed connection between language change and language acquisition have been stimulated by corpus-based studies of the time course of language change. In addition, applied mathematicians have begun to use the detailed data that the parsed historical corpora provide to test dynamical systems models of how language change diffuses through a population. Third, in computational linguistics, the corpora are beginning to prove useful as test beds for automatic processors, especially syntactic parsers. Crucially, they include a wide range of genres and levels of syntactic complexity, unlike the present the standard test bed, the Penn Treebank. Because its syntax differs very little from Present-Day English, the proposed new corpus of Modern English will be particularly useful in this regard. Finally, parsed corpora are beginning to have an impact on undergraduate education in linguistics. They are used in undergraduate research courses, where learning to use them reinforces students' understanding of language structure while, at the same time, the corpora provide datasets for research projects. Because its language poses no linguistic barriers to understanding, the present corpus will be particularly useful for this purpose.