Data Science
and Digital Humanities

(HVEE.00.016)

Maciej Eder

2025-09-08

topics covered today

  • Digital Humanities
  • Text analysis
  • Data
  • (Open) Science

Digital Humanities

what is Digital Humanities?

get a randomly drawn definition: https://whatisdigitalhumanities.com/

Digital Humanities

  • library catalogs
  • virtual collections
  • text analysis
  • vision
  • folklore studies
  • linguistics
  • computational literary studies
  • …

DH emerges

  • early journals
    • Computers and the Humanities (1966–2004)
    • Literary and Linguistic Computing (1986–2014)
    • cont. as Digital Scholarship in the Humanities (2015–)
  • conferences
    • ALLC/ACH conference (1989–2005)
    • cont. as “Digital Humanities” conference (2006–)
  • summer schools
    • Humanities Computing Summer Institute …
    • … cont. as Digital Humanities Summer Institute
  • textbook Companion to Digital Humanities (2004)

DH as a big tent

  • collecting the data
    • libraries
    • text collections
    • Text Encoding Initiative
  • data analysis
    • statistical modeling
    • distant reading
  • data curation
  • data visualization
  • programming
  • critical DH

a few examples

The following is just a small selection of DH projects:

Text Analysis

concordances, attributions, …

  • abbot Hugo de Saint-Cher: first concordance (ca 1230)
  • Lorenzo Valla: authorship attribution of The Donation of Constantine
  • Erasmus Roterodamus: attribution of St. Paul’s and Seneca’s letters
  • Roberto Busa: Index Tomisticus (1949–1970)

concordance to King James Bible (1737)

Biblical concordance (1526)

computers and the humanities?

  • the Enigma and code-breaking (Rejewski, Zygalski, Różycki)
  • Alan Turing
  • first computers suitable to count numbers only
  • IBM and a Jesuit monk Roberto Busa (late 1940s)

father Roberto Busa (1913–2011)

the internet vs. text collections

  • page images (no access to the text)
  • raw text (no formatting)
  • enriched corpora (containing information about structure and/or syntax)

WBC and similar sources

small & tidy vs. big & dirty

  • Optical character recognition (OCR): an easy way to convert image into text
  • However, a considerable number of mistakes might occur
  • Vast majority of corpora are OCR-ed and (sometimes) corrected
  • Gutenberg Project, Wikisources, etc. etc. etc.

PoS tagging: grammatical labels

My father was a clergyman.

👇

👇

PRP$ NN VBD DT NN

(poss. pron.) (noun) (verb) (determiner) (noun)

Part-of-Speech tagging

My father was a clergyman .
PRP$ NN VBD DT NN .
(poss. pron.) (noun) (verb) (determiner) (noun) (punct.)

Stanford Tagger

This     DT
is       VBZ
a        DT
short    JJ
sentence NN
.        .

So       RB
is       VBZ
this     DT
.        .

Structured output

<sentence id="0">
  <word wid="0" pos="DT">This</word>
  <word wid="1" pos="VBZ">is</word>
  <word wid="2" pos="DT">a</word>
  <word wid="3" pos="JJ">short</word>
  <word wid="4" pos="NN">sentence</word>
  <word wid="5" pos=".">.</word>
</sentence>
<sentence id="1">
  <word wid="0" pos="RB">So</word>
  <word wid="1" pos="VBZ">is</word>
  <word wid="2" pos="DT">this</word>
  <word wid="3" pos=".">.</word>
</sentence>

Data

data as representation

  • sample vs. population
  • operationalizing and acquisition
  • the biggest corpus is still small
    • under-resourced languages

data: biases

  • opportunistic sampling
  • cherry picking
  • biases (gender, race, culture traits, …)
    • e.g. COHA, COCA
  • uneven coverage
    • diachronic corpora

data: quality

  • small & tidy vs. big & dirty
  • poor OCR might affect the quality
  • other factors are also there

data: a model

Data is never a “neutral” material to conduct further modeling. Data is already a model, and it hightly depends on (tacit) assumptions and arbitrary decisions.

  • POS-tagging inherits the grammar on which the tagger was trained
  • TEI is always modeled as a tree
  • annotation as an arbitrary decision:
    • should a general sctructure be annotated?
    • should the annotation focus on minor elements?
    • etc.

example: structural annotation

<body>
    <div type="act" n="1">
    <head rend="caps">Primera Jornada</head>
        <div type="scene" resp="#rojas" n="1">
        <stage type="withText">Salen por una puerta, habiendo tocado cajas y trompetas, Jonatás, Simeón y Judas, y por otra Matatías, viejo, Zarés y músicos.</stage>
            <sp who="#música">
            <speaker rend="caps">Músicos</speaker>
                <l>Cuando alegre viene</l>
                <l>Judas vencedor,</l>
                <l>su frente coronan</l>
                <l>los rayos del sol.</l>
            </sp>
            <sp who="#matatías">
            <speaker rend="caps">Matatías</speaker>
                <l>Valerosos Macabeos,</l>
                <l>legítima sucesión</l>
                <l>de Palestina; hebreos,</l>
                <l>cuya gloriosa opinión</l>
                <l>vence al tiempo en los trofeos,</l>
                <l>triunfad dichosos; y vos,</l>
                <l>Judas valiente, a quien Dios</l>

example: linguistic annotation

            <seg xml:id="morph_1.2-seg" corresp="ann_segmentation.xml#segm_1.2-seg">
              <fs type="morph">
                <f name="orth">
                  <string>tej</string>
                </f>
                <f name="translit">
                  <string>tey</string>
                </f>
                <f name="interps">
                  <fs xml:id="morph_1.2.1-lex" type="lex">
                    <f name="base">
                      <string>ten</string>
                    </f>
                    <f name="ctag">
                      <symbol value="adj"/>
                    </f>
                    <f name="msd">
                      <vAlt>
                        <symbol xml:id="morph_1.2.1.1-msd" value="sg:dat:f:pos"/>
                        <symbol xml:id="morph_1.2.1.2-msd" value="sg:gen:f:pos"/>
                        <symbol xml:id="morph_1.2.1.3-msd" value="sg:loc:f:pos"/>
                      </vAlt>
                    </f>
                  </fs>
                </f>
                <f name="disamb">
                  <fs type="tool_report">
                    <f fVal="#morph_1.2.1.3-msd" name="choice"/>
                    <f name="interpretation">
                      <string>ten:adj:sg:loc:f:pos</string>
                    </f>
                  </fs>
                </f>
              </fs>
            </seg>

data: capta?

Capta is “taken” actively while data is assumed to be a “given” able to be recorded and observed. From this distinction, a world of differences arises. Humanistic inquiry acknowledges the situated, partial, and constitutive character of knowledge production, the recognition that knowledge is constructed, taken, not simply given as a natural representation of pre-existing fact.

(Drucker, 2011: 11)

further reading

  • Arnold, T. and Tilton, L. (2015). Humanities Data in R: Exploring Networks, Geospatial Data, Images, and Text. New York: Springer.
  • Crawford, K. (2021). Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University Press.
  • Drucker, J. (2011). Humanities Approaches to Graphical Display, Digital Humanities Quarterly, 5(1). url: https://dhq.digitalhumanities.org/vol/5/1/000091/000091.html.
  • Flanders, J. and Jannidis, F. (2019). The Shape of Data in Digital Humanities: Modeling Texts and Text-based Resources. Routledge.

further reading (cont.)

  • Karsdorp, F., Kestemont, M., & Riddell, A. (2021). Humanities Data Analysis: Case Studies with Python. Princeton University Press.
  • Schriebman, S., Siemens, R., and Unsworth, J. (2004) A Companion to Digital Humanities. Oxford: Blackwell. url: https://companions.digitalhumanities.org/DH/.