5-Texts 1

pdf version

Workset 5

This workset is about working with texts on the small scale. We’ll engage with larger scale analysis and topic modeling text week.

Choose a small-ish set of texts to work with. This may be an individual work, several smaller ones, or something in between; it should be between 10,000 and 1 million words, but you could go down to 3,000 if there’s something you really want to look at. Choose them based on the first questions below.

Voyant

  1. Upload your texts as a single document to Voyant-tools.org. Just screw around for a little while with the results. Try to obey Steven Ramsay’s suggestion to think about what the results will be before you look. What are the most common words going to be? Will there be trends?

  2. Upload your texts as multiple documents to Voyant tools using some level of address. You might lump together a few of the corrected versions vs uncorrected of scans; compare two authors whose works you can find online; etc. What are some of the differences between the levels of text? The statistics will come out easily.

  3. What is one obvious functionality that an online portal like Voyant should have but doesn’t?

Our course as a corpus

  1. I’ve uploaded a copy of all of our reading responses into an instance of my own text analysis platform somewhat similar to Voyant, Bookworm. In class we (should have) looked at your blog post usage of individual terms. What did you find? Here is a link that tells you which individual words distinguish your writing: are they explicable? Do you believe them? What’s present here that isn’t elsewhere?

  2. Finally, I’ve pasted below a topic model trained on the paragraphs of texts. Give it a reading. Where does that model work? Where does it fail? How might it be useful on a much larger corpus of posts? How would you want to see it plotted?

Topic label Words
media McLuhan Manovich medium visual ways encourage 325.0
text analysis texts topic modeling Rhody Blevins 1544.0
slaves impact potential writers cliometric figures free 290.0
form cultural fully reproductions objects aesthetics freedom 256.0
contributors menus Menu dishes menu NYPL transcription 190.0
project process crowdsourcing transcribe transcription documents transcribing 1165.0
Marathon archive Boston Bombing story contribute contributed 227.0
data Fogel Cross Engerman Haskell slavery evidence 1847.0
initiative Illinois Bentham site Omeka workers Bracero 270.0
museum Presidential living Foundation museums galleries Obama 317.0
History history http com org year University 1255.0
buildings large windows building prominent built th 484.0
maps map modern location needed historical control 221.0
tags project media images music video frames 322.0
interest old city Cape came Caesar Classics 242.0
data logs captain’s website record role helping 231.0
Ruggles quantitative conclusions Engerman’s United American studies 566.0
access OCR copy scanned 1930 ease hurdles 247.0
value reproduction film piece original Benjamin’s lost 380.0
Ramsay evidence computing House humanistic relationship Dream 327.0
recorded ages person data African age ship 261.0
Greek fragments papyrus Ancient identifying Coptic Lives 277.0
digital humanities Cecire field Digital Humanities Liu 1224.0
book level books easy ultimate presented outside 401.0
galaxy galaxies Bush Web scientists transport memex 154.0
year floggings whippings watch math 0.7 days 209.0
women agriculture efficiency Southern ibid Northern 1975 360.0
data analysis Gibbs scholars Owens methods qualitative 549.0
video Trove translated crowd name articles Nazis 294.0
work use historical history digital sources works 3496.0
text highlight website created multiple idea different 213.0