This site is part of a corpus-based computational linguistics dissertation project by Michael McGuire. Click on a link to the left see sentiment trendlines and tagged sentences for each individual. Please note that the project is still in development and this is not the final published version. I will continue to work on the user interface and data visualization.

The project involves automated sentiment tagging and analysis of an 18th century Moravian corpus, consisting of personal autobiographical and religious testimonial memoirs (Lebensläufe) from members of the Moravian Church (Herrnhuter Brüdergemeine) living in the settlement of Fulneck, England in West Yorkshire. The Fulneck corpus is part of the larger corpus being created as part of the Moravian Lives Project hosted at Bucknell University. The project currently consists of English, German, and Swedish memoirs. The Swedish memoirs are hosted at University of Gothenburg.

136 memoirs are included in the sentiment tagged Fulneck corpus, comprising most of the memoirs that have currently been scanned and digitized from the archive in Fulneck. The memoirs were originally scanned and digitized with a particular interest and focus on single women. The large number of memoirs from women and in particular single women is unique and offers a valuable perspective that is not often found in other documents from the time period. Moravians believed each individual's personal or inner journey through life was important and should be documented. As such, virtually every member of the church has a personal memoir, many of which were personally written or dictated.

Sentiment tagging uses an expanded lexicon based approach that also accounts for context, including negation, content modifiers, and frequent collocations and phrases. The custom lexicon and sentiment tagger is based on VADER sentiment module in Python. However, a different tagging and sentiment calculation methodology is used to calculate positive and negative scores as well as sentiment intensity. Narrative trendlines are created by smoothing raw scores and normalizing and rescaling data to a common narrative time index. More specific details about the methodology will be provided in the dissertation.