# Digital Humanities and Big Data History I · Historical Sources · https://tryals.app/learn/history-i/digital-humanities-and-big-data ## Distant Reading and Computation Nobody has read the nineteenth-century novel. Any claim about it is based on **a canon**, a few hundred survivor books selected by a century of filtering. **Distant reading** is the response: don't read the books, count them. You lose the nuance of individual texts to see **the shape of the whole**. | Task | Why the machine wins | |---|---| | **Counting** | Word frequencies across 50,000 texts; no human can | | **Linking** | Record linkage across censuses, registers, shipping lists | | **Mapping** | Plotting thousands of events; patterns emerge spatially | | **Networks** | Who corresponded with whom, at scale | Distant reading and close reading are not rivals. Close reading answers what a text means; distant reading answers what was typical, when a pattern began, and what the canon left out. ## What Computation Cannot See **Digitization is a new filter** stacked on the old ones. The digital corpus is just the well-funded, out-of-copyright, cleanly-printed corner of history. Watch out for specific traps: - **The OCR floor:** Optical character recognition fails on damaged print, gothic type, and handwriting. - **Searchability bias:** Concepts without a keyword are invisible to full-text search. - **Metadata is an argument:** Database categories impose a classification the sources may not share. - **False precision:** A number from an unexamined corpus is not evidence; it is a rumor with a decimal point. **The productive loop:** distant reading finds the anomaly, close reading explains it, and the explanation suggests the next thing to count. The machine finds where to look, but it does not know what it found. ## Practice questions 8 of this lesson's 13 practice questions, with answers. The full set is in the app. ### 1. Why is "metadata is an argument" more than a slogan? A. Every field imposes a classification the sources may not share B. Metadata takes longer to produce than scanning C. Metadata standards change frequently between institutions D. Metadata is usually copied from the source’s own categories **Answer:** A. Every field imposes a classification the sources may not share **Why:** The schema is an interpretation made before any analysis begins, and it disappears into the data. Once entered, it stops looking like a decision and starts looking like a finding you can plot. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/why-is-metadata-is-an-argument-more-than-a-slogan ### 2. Complete the principle. **Answer:** Digital methods change the **scale**, not the epistemology. The digital corpus is the archive after two extra filters — **scanned** and OCR-legible, neither of which is random. So every question you would ask a document, you must ask the **corpus**: who made it, why, and what is missing. **Why:** Scale, scanned, corpus. The machine finds where to look and has no idea what it found, which is fine, as long as you never mistake its output for a population. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/complete-the-principle ### 3. Sort each task by whether **computation** or **a human reader** does it better. **Answer:** - Computation wins: Word frequencies across 50,000 texts, Linking names across censuses and registers, Finding anomalies in a mass no one could eyeball - Human reader wins: Explaining why an anomaly exists, Judging what a passage means, Deciding which pattern is worth pursuing **Why:** The machine finds where to look; it does not know what it found. The productive loop runs both ways: distant reading finds the anomaly, close reading explains it, the explanation suggests the next thing to count. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/sort-each-task-by-whether-computation-or-a-human-reader-does-it ### 4. Why is the OCR floor more dangerous than a simple gap in coverage? A. It slows searches on large corpora B. It fails on damaged and handwritten text, silently C. It introduces spelling errors that are easy to detect D. It only affects texts published before 1700 **Answer:** B. It fails on damaged and handwritten text, silently **Why:** It returns nothing, and says nothing about having returned nothing. Your "complete search" skips the damaged pages, and damage correlates with poverty, provinciality, and cheap printing, so the silence has a politics. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/why-is-the-ocr-floor-more-dangerous-than-a-simple-gap-in-coverage ### 5. An n-gram chart shows 'liberty' declining sharply after 1800. Why is that not yet evidence that liberty mattered less? A. Because OCR errors make nineteenth-century counts unreliable B. Because the corpus is too small before 1800 for reliable comparison C. Because word frequency tracks vocabulary, not concepts D. Because frequency charts cannot be normalised across differently sized corpora **Answer:** C. Because word frequency tracks vocabulary, not concepts **Why:** Counting a token is not counting an idea, and concepts migrate between words as registers shift, so a curve is a hypothesis about vocabulary that needs reading to test. Options 2 to 4 name real technical problems that better data would fix; this one does not go away with a bigger corpus. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/an-n-gram-chart-shows-liberty-declining-sharply-after-1800-why-is ### 6. Why does a full-text archive quietly narrow what historians end up writing about? A. You find what you can search for B. Full-text search is slower on handwritten sources C. Search engines rank popular results higher D. Databases prioritize recently added records **Answer:** A. You find what you can search for **Why:** You cannot type a keyword for a thing nobody named. Full-text search quietly reshapes research toward the nameable, which is a strange principle on which to organize a discipline. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/why-does-a-full-text-archive-quietly-narrow-what-historians-end-up ### 7. Which are genuine achievements of computational history? A. Name-linked censuses making nineteenth-century mobility measurable B. Determining what a novel’s central metaphor means C. Mapping the correspondence networks of the Republic of Letters D. Turning scattered manifests into a countable slave-voyage database **Answer:** A. Name-linked censuses making nineteenth-century mobility measurable; C. Mapping the correspondence networks of the Republic of Letters; D. Turning scattered manifests into a countable slave-voyage database **Why:** Three real reorganizations of knowledge, each one a scale no human reader could hold. The fourth is close reading’s job, and the machine has no opinion about metaphors. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/which-are-genuine-achievements-of-computational-history ### 8. A large digitized corpus can be treated as a representative sample of the archive. **Answer:** False **Why:** It is the well-funded, out-of-copyright, OCR-friendly corner of the archive, sampled by nobody according to no design. It resembles everything only from the inside of the search box. Page: https://tryals.app/practice/history-i/digital-humanities-and-big-data/a-large-digitized-corpus-can-be-treated-as-a-representative-sample-of