128.2 ]

the changing profession

The Meaning of the

WHAT IS THE MEANING OF THE DIGITAL HUMANITIES TO THE HU- alan liu MANITIES? his question of disciplinary meaning—which I ask from the viewpoint of the humanities generally—is larger than the question of disciplinary identity now preoccupying “DH” itself, as insiders call it. Having reached a critical mass of participants, publications, conferences, grant competitions, institutionalization (centers, pro- grams, and advertised jobs), and general visibility, the ield is vigor- ously forming an identity.1 Recent debates about whether the digital humanities are a “big tent” (Jockers and Worthey), “who’s in and who’s out?” (Ramsay), whether “you have to know how to code [or be a builder]” (Ramsay, “On Building”), the need for “more hack, less yack” (Cecire, “When Digital Humanities”; Koh), and “who you calling untheoretical?” (Bauer) witness a dialectics of inclusion and exclusion not unlike that of past emergent ields.2 An ethnographer of the ield, indeed, might take a page from Claude Lévi-Strauss and chart the current digital humanities as something like a grid of aili- ations and diferences between neighboring tribes. Exaggerating the diferences somewhat, as when a tribe boasts its uniqueness, we can thus say that the digital humanities—much of which ailiates with ALAN LIU , professor in the English depart- older humanities disciplines such as literature, , classics, and ment at the University of California, Santa the languages; with the remediation of older media such as books and Barbara, is the author of Wordsworth: libraries; and ultimately with the value of the old itself (history, ar- The Sense of History (Stanford UP, 1989), The Laws of Cool: Knowledge Work and chives, the curatorial mission)—are not the tribe of “ stud- the Culture of Information (U of Chicago P, ies,” under the sway of the design, visual, and media arts; Continental 2004), and Local Transcendence: Essays on theory; cultural criticism; and the avant- garde new.3 Similarly, despite Postmodern Historicism and the Database signiicant trends toward networked and multimodal work spanning (U of Chicago P, 2008). He started Voice of social, visual, aural, and haptic media, much of the digital humanities the Shuttle, a Web site for humanities re- focuses on documents and texts in a way that distinguishes the ield’s search, in 1994. Recent projects he has di- rected include the University of California work from digital research in media studies, communication studies, Transliteracies Project, on online reading, information studies, and sociology. And the digital humanities are and RoSE (Research-Oriented Social Environ- exploring new repertoires of interpretive or expressive “algorithmic ment), a software project. Liu is a coleader criticism” (the “second wave” of the digital humanities proclaimed of the 4Humanities advocacy initiative.

© 2013 alan liu PMLA 128.2 (2013), published by the Modern Language Association of America 409 410 The Meaning of the Digital Humanities [ PMLA

in “The Digital Humanities Manifesto 2.0” creasingly use social- network analysis to study [3]) in a way that makes the ield not even its plays, novels, literary reception, history, and earlier self, “humanities computing,” alleged so on.7 And of the communities of people who to have had narrower technical and service- overlap, trickster-like, between the digital hu- oriented aims.4 Recently, the digital humani- manities and new media studies, the HASTAC ties’ limited engagement with identity and collaboratory—which has had enormous suc- social-justice issues has also been seen to be cess tapping into the energies especially of a differentiating trait—for example, by the graduate students (“HASTAC Scholars”)—is vibrant #transformDH collective, which wor- perhaps the most coyote. ries that the digital humanities (unlike some Yet even if we were to complete our hy- the changing profession changing the areas of new media studies) are dominantly pothetical ethnographer’s chart, it would not not concerned with race, gender, alternative adequately explain the digital humanities. We sexualities, or disability.5 would be leaving unexplained the relation of Of course, there are overlaps of people the digital humanities to the humanities gen- and methods between tribes—to the point erally. My thesis is that an understanding of that, taking another page from Lévi- Strauss, the digital humanities can only rise to the we would want to mention trickster figures level of an explanation if we see that the un- embodying both sides of digital humanities’ derlying issue is the disciplinary identity not diferences. One such trickster is the ield of of the digital humanities but of the humani- book history. Proliic in its revisionary studies ties themselves. For the humanities, the digi- of matrial texts, ephemera, marginalia, social tal humanities exceed (though they include) reading, publishing history, and so on, book the functional role of instrument or service, history now partly parallels post- McLuhan the pioneer role of innovator, the ensemble media studies. Peter Stallybrass argues, for in- role of an “additional field,” and even such stance, that “the codex and the printed book faux- political roles assigned to new ields as were the indexical computers that Christianity challenger, reformer, and (less positively) ith adopted” as its “technology of discontinuity” column. his is because the digital humani- (74, 73). Book history therefore overlaps on its ties also have a symbolic role. In both their new- media- studies side with media archaeol- promise and their threat, the digital humani- ogy (especially in the German lineage well rep- ties serve as a shadow play for a future form resented in this context by Cornelia Vismann’s of the humanities that wishes to include what Files: Law and Media Technology) and on its contemporary society values about the digi- digital humanities side with approaches to tal without losing its soul to other domains digital texts rooted in revisionary textual edit- of knowledge work that have gone digital to ing, bibliography as the sociology of texts, and stake their claim to that society. Or, precisely the materiality of the digital.6 A similar trick- because the digital humanities are both func- ster is science and technology studies (STS), tional and symbolic, a better metaphor would which are curiously underrepresented in both be something like the register in a computer’s new media studies and the digital humanities central processor unit, where values stored even as they are clearly relevant in a way repre- in memory are loaded for rapid shuffling, sented by such scholars as N. Katherine Hayles manipulation, and testing—in this case, to and Tim Lenoir. So, too, historical sociology— try out new humanistic disciplinary identi- which now applies social-network analysis to ties evolved for today’s broader contention of document corpora—is a trickster splitting the knowledges and knowledge workers. diference between the social sciences and the he question of the meaning of the digi- digital humanities, the latter of which also in- tal humanities best opens such an argument to 128.2 ] Alan Liu 411 view because it registers both a speciic problem object to qualitative arguments and insights the changing profession in the digital humanities and the larger crisis of about humanistic subjects—culture, literature, the meaningfulness of today’s humanities. art, etc.—is not clear. (46; my emphasis)

Throughout, Heuser and Le-Khac give Meaning is clearly a metavalue and also meaning a gravity that indicates that in the metaproblem for the digital humanities. To digital humanities the meaning problem has unpack this meaning problem, I will spotlight roughly the same weight as the “saving the a recent work of digital literary scholarship by phenomena” problem in the philosophy of two beginning scholars that is state- of- the- science—the problem, that is, of relating em- art and representative of major trends in the pirically observed phenomena to explainable, digital humanities—a tactic that has the addi- theorizable, or predictable (“saved”) phenom- tional advantage of providing outsiders to the ena. In essence, Heuser and Le- Khac are out to ield with an end- to- end look at an example “save the data” by making it meaningful.9 he of research by digital humanists. he work is form of data they wish to save is quantitative. Ryan Heuser and Long Le- Khac’s A Quanti- Of course, no single example can typify tative Literary History of 2,958 Nineteenth- all aspects of the booming ield of digital hu- Century British Novels: he Semantic Cohort manities. So before we examine the speciics Method (2012), the fourth of the influential of the meaning problem in this instance, it digital pamphlets issued by the Stanford Liter- will be useful to make four observations that ary Lab.8 Heuser and Le-Khac report on their relate Heuser and Le- Khac’s research to other innovations in the methods of “distant read- work in the digital humanities. he meaning ing” and that are the signatures of problem may not be as central everywhere, the lab (where they worked with Matthew L. but its frame of analysis is convertible to Jockers, Franco Moretti, and others), and they other frames so that we can see that many do so with a methodological self- awareness parts of the ield link up to the same cluster of that puts the meaning problem front and cen- issues. he following is a kind of conversion ter. hey relect near their opening: table for relating Heuser and Le-Khac’s work to other digital humanities areas. [W]hat is the meaning of changes in word us- age frequencies? What do we do with such First, we note that Heuser and Le-Khac data? With much current research drawing select their research material from already on word frequencies and other quantiiable digitized texts (2,958 British novels from aspects of culture, these are big questions. We 1785 to 1895, all from The Internet Archive can see now that the greatest challenge of de- except for 250 from Chadwyck- Healey).10 veloping digital humanities methods may not his means that their work belongs not in the be how to cull data from humanistic objects, orbit of digitizing, text encoding, publish- but how to analyze that data in meaningfully ing, or archiving (activities characteristic of interpretable ways. (4; my emphases) many projects in digital editing, collecting, and curating) but in that of processing and And they add in their concluding section: analyzing already built digital repositories (in the manner of projects in text analysis, he general methodological problem of the dig- ital humanities can be bluntly stated: How do social- network analysis, visualization, spatial we get from numbers to meaning? he objects history, etc.). Still, the digital humanities are being tracked, the evidence collected, the ways young enough that these two broad modes they’re analyzed—all of these are quantitative. of work are not fully specialized. On the one How to move from this kind of evidence and hand, leading text- encoding and digital- 412 The Meaning of the Digital Humanities [ PMLA

archiving projects ind it necessary to create tion through diferent routes and at diferent their own analytic, processing, and visualiza- stages. Digital history emerged in a discipline tion tools to present materials. And, on the that had already forcefully experienced nu- other hand, text- analysis, visualization, and merical method—for example, the quantita- other processing projects oten have to go to tive side of the Annales school and cliometrics. great lengths to select, clean, and prepare pre- Perhaps as a result, digital history is farther existent digital materials as a usable corpus. along than most digital humanities branches Second, while Heuser and Le- Khac for- in molding quantitative work to a related mulate their meaning problem in terms of heritage of Annales method: spatial and car- quantification, that is not the only possible tographic history. he state of the art in digital the changing profession changing the formulation. Quantification is indeed a key history, as it were, is Fernand Braudel plus sat- digital humanities issue, especially when ellite mapping (Seefeldt and homas; Owens). linked to sheer quantity in the wake of recent Fourth, Heuser and Le-Khac’s research “” funding competitions in the ield. required a combination of skills in program- But digital humanists also have nonquanti- ming and interpretation, thus addressing the tative ways of putting the meaning problem. “do you have to be a builder?” question that One is the idea of models (and modeling pro- has recently bedeviled the digital humanities cesses) as richly developed, for example, by by answering, in efect, “you have to be both Willard McCarty. Models reveal meaning a builder and an interpreter.” Less interesting (recognized in patterns, trends, forms) only than this question itself, which is based on an by reducing the dimensions and features of increasingly obsolete notion of solo work in meaning. Diagrammatic models, especially which one is either a builder or interpreter, is the visualizations proliferating in the digital the way Heuser and Le-Khac are both builders humanities (including Heuser and Le-Khac’s and interpreters: through rich collaboration. essay), are a case in point, since large dis- Here we reach the outer limits of the frame of course networks (visualized through such the meaning problem, where it converts into tools as Gephi) are comprehensible when their a coextensive frame. Just as meaning is both a scope or detail is kept low but otherwise grow metavalue and a metaproblem, so is collabo- into beautifully mystifying galaxies of nodes ration as it bears on such urgent issues in the and links. Another case in point is the kind digital humanities as coauthorship, collec- of textual model (instantiated in document- tive project building, multigraph books, open type deinitions [DTDs] and XML schemas) peer review, social media, crowdsourcing, that enables text encoding. A well- known de- and the hiring and promotion implications bate in the digital humanities thus concerns of all these. Rather than explore the collabo- whether the principle of an ordered hierarchy ration problem in its own frame here, I note of content objects (OHCO) underlying such only that it is fundamentally convertible to models makes texts machine- readable only by the meaning problem. For example, the ques- disallowing the full range of what Jerome Mc- tion of what kind of knowledge is produced Gann calls human- readable “overlapping” and by “the wisdom of the crowd,” “collective in- “recursive structures” (“Position Statement”). telligence,” “the long tail,” “the hive mind,” hird, even if we concentrate on quantii- “folksonomy,” and so on (dominant memes cation as the key meaning problem, it makes of Web 2.0) is essentially a question about the a diference which disciplinary branch of the meaning of the social version of big data, the digital humanities we are dealing with, since big crowd. he mind, or mindlessness, of that digital history and digital literary studies, to crowd has been a core problem of modernity take two major branches, arrive at quantiica- since at least the French Revolution. 128.2 ] Alan Liu 413

With these observations as a guide for re- “moral valuation,” “sentiment,” and “partial- the changing profession lating Heuser and Le- Khac’s work to other em- ity” (11–19). hen, serendipitously (instead of phases in the digital humanities, we can now under the sway of preexisting scholarship), look closely at their pamphlet. he meaning they fed the unlikely seed word hard into Cor- problem comes to the fore there as a conse- relator to discover a strikingly distinct, very quence of Heuser and Le- Khac’s key research large cohort of “concrete description words innovation. hey report on their use of a tool of a direct, everyday kind” whose frequencies they created called Correlator, which, when rose dramatically over time. his cohort they fed “seed words” suggested in part by “exist- subcategorized into semantic ields for action ing literary scholarship” (11), processed their verbs, body parts, colors, numbers, locational long-nineteenth- century corpus of novels and directional adjectives and prepositions, (with the aid of a database designed by Jock- and physical adjectives (19–27). he strongly ers that tabulated the number of occurrences inverse correlation between these two large of each word [Heuser and Le-Khac 6]) to ind and diferentiated word cohorts, which they other words that were statistically correlated further statistically corroborated (28–29), with the seed words and whose frequency started them on the path of interpretation. trends, measured longitudinally across the he “abstract” cohort, they concluded, con- century in decade intervals, closely followed sists of words whose usage, while unanchored the frequency trends of the seed words. In in speciics, was “monitored and tightly con- other words, Correlator finds what Heuser strained” by the traditional smaller, rural and Le- Khac call “word cohorts” in the cor- communities represented more or less ear- pus, consisting of words that kept company lier in the century in evangelical, gothic, and with one another and behaved similarly over village novels and the novels of Jane Austen, time, waxing or waning in frequency together Walter Scott, and George Eliot (Eliot is a and standing out from other companies of chronological outlier), while the “hard” co- words. he inal result—which I jump to while hort is populated by words whose stand- alone skipping the details of Heuser and Le-Khac’s referentiality and alienation from larger con- algorithmic method as well as, temporarily, an texts correlated later in the century with the important adjustment step in their method— “wider, less constrained social spaces” of the is the identification in the word cohorts of urban centers represented in city, industrial, “rich, consistent semantic fields” that are adventure, fantasy, science iction, and chil- “both semantically and culturally legible” in dren’s novels (30–34). historical trends. An example is the 136 words This insight finally led Heuser and Le- such as gentle, sensible, vanity, elegant, deli- Khac at the highest level of interpretation cacy, reserve, mild, and restraint that they label to follow Raymond Williams’s Culture and the “social restraint ield” in the novels (8). Society and he Country and the City in sug- Using this method, Heuser and Le-Khac gesting that the inverse “abstract” and “hard” made two principal discoveries. First, they trends reveal something signiicant about the used the seed words integrity, modesty, sen- “social space of the novel.” hey argue that sibility, and reason (suited to scholarship the “values of conduct and social norms” in on novels of the period) to ind a strikingly “knowable communities” (a phrase from Wil- large cohort of “abstract, socially norma- liams) declined in the face of “urbanization, tive, evaluative, and highly polarized words” industrialization, and new stages of capital- whose frequencies declined dramatically over ism” (33, 35–36). he computational veriica- the century. his cohort they subcategorized tion of this previously known thesis, coupled into semantic ields labeled “social restraint,” with the discovery of precise word cohorts 414 The Meaning of the Digital Humanities [ PMLA

giving genuinely fresh insight into the thesis, interpretation is equally a goal of the more enables them in later sections of their pam- postmodern side of the digital humanities, phlet to ofer more recognizably normative which argues that critics should use algo- literary and cultural criticism, touching on rithmic methods to play with texts experi- action, setting, and character (37–45). Here mentally, generatively, or “deformatively” to they closely read texts (including, for the discover alternative ways of meaning that are irst time, block quotations), match aggregate not so much true to preexisting signals as rifs trends to “units understandable and famil- on those signals. he common goal is to ban- iar to us as readers and literary scholars[:] ish, or at least crucially delay, human ideation the actual novels, genres, and authors” (31), at the formative onset of interpretation. the changing profession changing the and generalize about sweeping changes in However, tabula rasa interpretation puts cultural history, but with the important dis- in question Heuser and Le-Khac’s ultimate tinction that their reading is based not on the goal, which is to get from numbers to human- usual anecdotal, faux-empirical, or unique- istic meaning (“qualitative arguments and case observations of literary criticism (e.g., insights about humanistic subjects—culture, noticing that a word appears “oten” or in an literature, art, etc.”). It is not clear epistemo- “important” location) but on lines of inter- logically, cognitively, or socially how human pretation generated by machine observation. beings can take a signal discovered by ma- Here we reach the crux of the meaning chine and develop an interpretation leading problem in the digital humanities. While to a humanly understandable concept unless Heuser and Le- Khac mix “supervised and that signal (in order to be recognized as a sig- unsupervised procedures” (28), the newest, nal at all) contains a coeval conceptual ori- boldest, and most interesting part of their gin that is knowable in principle because, at a methodology is unsupervised. They help minimum, the human interpreter has known advance an important, general digital hu- its form or position (the slot or approximate manities goal that might be called tabula rasa locus in the semantic system where its mean- interpretation—the initiation of interpreta- ing, or at least its membership in the system, tion through the hypothesis-free discovery is expected to come clear).12 If the machine of phenomena. Also shared by such digital can discover word cohorts triggered by seed humanities methods as topic modeling (a words, in other words, then what seed con- mathematics-based way to discern differ- cepts—which is to say seed semantics (using entiated clusters of words that Heuser and “semantics,” for the moment, as overlapping Le- Khac use to corroborate their findings), with “concepts” in a manner consistent with the ideal in its purest form is what Heuser Heuser and Le-Khac’s usage)—lurk in the and Le- Khac call “an unsupervised method background as a latent, classificatory form that generates topics without subjective in- of relational semantic positions able to make put from users” (28).11 That is, a computer word cohorts into “proto- semantic ields” (7)? should be able to read texts algorithmically hus the immense importance of the ad- and discover word cohorts or clusters lead- justment step in Heuser and Le-Khac’s method ing to themes without acting on an initial that I earlier elided. In fact, Heuser and Le- concept from an interpreter looking to con- Khac used Correlator by itself to produce only firm a particular theme. Of course, Heuser initial word cohorts and not inished semantic and Le-Khac assume that there are preexist- ields because they realized that they needed ing themes to be found in the word cohorts to ensure that their cohorts had a semantic of primary materials and also that the main consistency that quantitative correlation alone mission is to discover them. But tabula rasa could not ofer. Some word cohorts discovered 128.2 ] Alan Liu 415 by machine, for example, suggested only fuzzy concepts, “the external world,” “the mind,” the changing profession semantic ields that seemed blurred by extra- and “society,” and then, descending its clas- neous words or, inversely, to lack words that siicatory tree by stages, into myriad ramiica- should have been there. In other words, word tions. On the lower branches of the taxonomy, cohorts had to be iltered and illed out in ways synonyms appear in the chronological order that made sense. Heuser and Le-Khac thus re- in which they entered the language.14 At irst alized that they needed not just seed words but glance, therefore, the HTOED is the perfect seed semantic concepts suspended precisely in concept- hunting guide for Heuser and Le- what I above called a classiicatory form of re- Khac’s word-cohort- hunting machine. his lational semantic positions, endowing cohorts seems even more apparent when we realize with a sense of what is and is not proximate just how human- powered the semantic in- in meaning. terpretation involved in making the HTOED In short, Heuser and Le-Khac needed a was. As documented in reports by Christian thesaurus, and not just any thesaurus but one Kay and Irené Wotherspoon, among its chief ofering a historical semantics matched to the editors, and in a detailed e-mail to me from longitudinal dimension of their word cohorts. the current HTOED associate director, Marc Ater they had already begun using Correlator, Alexander, the editors’ sorting of words by they pulled a rabbit out of the hat. hey turned meaning and chronology recapitulated the to the remarkable Historical hesaurus of the famous use of paper slips in the compilation Oxford En glish Dictionary (2009; hereafter of the OED itself.15 Started before humani- HTOED), which had just been published, and ties computing was practical, the HTOED borrowed its historical semantic classiications required human beings over decades to write through what they call “a dialogic method down individual words from the OED on that drew on both quantitative historical data paper slips with meanings, usage dates, and and qualitative semantic rubrics to construct sparse metadata, then to sort, bundle, and ile semantic ields with precision and nuance”: the slips in conceptual groupings and hierar- chies. When computation entered the picture, Having moved through an empirically and his- it did so originally in a secondary capacity (to torically focused stage of semantic ield devel- drive the print run of the work). In its forma- opment, we needed to return to the semantic tive state, the HTOED was a human labor of focus in order to make such purely empirical semantic ordering. word cohorts interpretable and meaningful. By installing the HTOED as what Our initial approach was to ilter through these amounts to a plug- in for Correlator, Heuser words for groups that seemed semantically co- and Le-Khac sowed their hermeneutical pro- herent, but this proved too loose and subjective. cess with a coseed of human semantic inter- . . . Finally we turned to the [HT]OED. . . . It’s pretation. They thus “solved” the meaning nearly exhaustive, its categories are nuanced problem only by deftly turning the aporia and specific, and it’s truly organized around meaning. We used this powerful taxonomy to between tabula rasa quantitative interpreta- do two things: irst, be more speciic in identi- tion and humanly meaningful qualitative in- fying the semantic categories that constituted terpretation into its own apparent solution: a our word cohorts; second, to expand these “dialogic approach that oscillates between the word cohorts with many more words. (7)13 historical and the semantic, between empiri- cal word frequencies that reveal the histori- Created at the University of Glasgow be- cal trends of words and semantic taxonomies ginning in 1965, the HTOED taxonomizes the that help us identify the meaning and con- English language into three master semantic tent of those trends” (9). hey add, “Strictly 416 The Meaning of the Digital Humanities [ PMLA

speaking, the methods developed here are not ing. Seeing design in data is a method for looking at word cohorts, which have histori- knowing meaning in the digital humanities. cal consistency but may lack semantic coher- he other deiciency I refer to is science ence, or semantic ields, which have semantic and technology studies (STS), which digital coherence but may have an ahistorical rela- humanists oten occlude even as they speak tionship. he real object of study is a hybrid of “digital technology,” “media technology,” one that satisies both requirements” (9–10). and so on, as if technology were an indivis- Such hybridity is a prevalent feature of digital ible part of the digital and media without humanities method.16 its own history, philosophy, sociology, poli- How might such hybrid method be better tics, economy, and aesthetics all tangled up the changing profession changing the grounded theoretically? I return to my earlier with, yet also distinct from, science. I invoke diferentiation of the digital humanities from especially the postmodern branch of STS neighboring ields. Two of the deiciencies in (e.g., Feyerabend; Latour; Pickering), whose the digital humanities revealed by that ethno- “against method” view of science (especially graphic map of ields are relevant. in its weird relations with technology) is that One is design theory and practice, any quest for stable method in understanding which Anne Burdick, Johanna Drucker, Pe- how knowledge is generated by human beings ter Lunenfeld, Todd Presner, and Jeffrey using machines founders on the initial fallacy Schnapp in their recent Digital_ Humanities that there are immaculately separate human declare with manifesto- like boldness to be and machinic orders, each with an ontologi- central to the ield (esp. 12–16, 117–19). Cur- cal, epistemological, and pragmatic purity rently, it seems to me, the union between the that allows it to be brought into a knowable digital humanities and new- media- studies methodological relation with the other— communities that would be needed for full whether a relation of master and slave, cause realization of this vision is more a goal than and efect, agent and instrument, or another. a reality, existing in speciic projects and not What could we learn from STS if we took the programmatically. But the book’s coauthors, Stanford Literary Lab and other digital hu- expressing a West Coast view of the digital manities centers and programs at their word humanities rooted in media and design arts, and studied them as labs, much as Andrew are right to aim for design as a principle of Pickering studied the “hunting of the quark” knowledge discovery and generation rather in a physics lab (68–112)? he answer is likely than (more typical in the digital humanities) that digital humanities method—converg- as an ater-the- fact rendering of data in scat- ing with, but also sometimes diverging from, ter plots, social- network graphs, and other scientific method—consists in repeatedly stale visualizations or, equally tired, book- coadjusting human concepts and machine like or blog- like publication interfaces. As technologies until (as in Pickering’s the- the coauthors put it, when “used to pose and sis about “the mangle of practice”) the two frame questions about knowledge,” design is stabilize each other in temporary postures “an intellectual method,” an “embodiment of truth that neither by itself could sustain. of a project’s argument and methodology,” Knowledge is an ice- skater’s dance on a slip- “an act of thinking,” and a “new foundation pery epistemic surface, on which neither the for the conceptualization and production of human nor the machine—the dancer nor knowledge” (13, 14, 15, 117). Interactive, mul- the skates—alone can stand. STS, in other timodal, dynamic, and participatory design words, is another method for knowing mean- in the digital age is a method not just of pat- ing in the digital humanities. In fact, it can tern recognition but of pattern understand- be thought of as complementing the method 128.2 ] Alan Liu 417

of design. he diference between a mangle of HTOED has always been that the classiica- the changing profession practice and a dance, ater all, is design. tion at whatever level should develop from Indeed, an STS approach opens a fasci- the data rather than be imposed upon it nating chapter in our reading of Heuser and using some predetermined schema” (“Clas- Le-Khac’s work that I can only briely relate.17 siication” 258). he underlying authority While I have not conducted STS- style eth- of the HTOED, it turns out, is “data” (the nographic and documentary research on the record of the En glish language observed through the instrument of the OED). hat Stanford Literary Lab (except through infor- is, the HTOED itself adhered to the princi- mal observations during a daylong visit at the ple of the tabula rasa discovery of phenom- invitation of the lab),18 I have looked from a ena and initiation of interpretation (in this distance into the “lab” of the making of the case, taxonomic interpretation). HTOED—that is, the human- semantics lab 4. It was thus ater surveying their lexical that Heuser and Le- Khac position as the part- data that the HTOED editors contravened ner to their computational-analysis lab. When the ordering scheme of their canonical we view the HTOED in this way, we realize that predecessor, Roget’s hesaurus, which had Heuser and Le- Khac pulled not just any rabbit put “abstract relations” irst. he HTOED out of a hat but a special rabbit much like the puts the “external world” irst in its tri- one that led Alice down the rabbit hole. he nary taxonomic structure. hat’s because HTOED is less a solution to the meaning prob- concrete and near- to- hand word senses lem than a recursive, looking-glass version of relating to the external world entered the the very same problem. At least ive aspects of language earlier (Kay, “Classiication” its making are relevant in this regard: 257–62). he implications for our read- ing of Heuser and Le- Khac are startling. 1. While the creation of the HTOED was he long historical trend identiied by the essentially precomputational, it was not HTOED, in which concrete words precede pretechnological. Requiring human be- abstract ones, is the reverse of the long- ings to write words, meanings, and meta- nineteenth- century novelistic trend that data on paper slips and then to sort the Heuser and Le- Khac identify, in which slips in drawers (and later in databases), “abstract” words dominate earlier and the HTOED originated as a thoroughly en- “hard” words later. Of course, this by no tangled human- technological, rather than means contradicts their thesis, since in simply human, semantic act. more fully stated form their argument may 2. Following Heuser and Le-Khac, I have well be that, because of urbanization, the so far treated “semantics” and “concepts” nineteenth century (and the novel form) as coincident. But in relecting on the was exceptional, or at least matched the HTOED’s method, Kay separates out lexical trend of only a few other centuries (and semantics (meaning relations among words) forms) in history. But a commitment to and conceptual semantics (meanings linked reading nineteenth- century novels in con- to external-world referents). A focus on junction with the HTOED’s larger corpus the former, she says, was “essentially the would require further testing to see how procedure adopted in HTOED, and is what exceptional the nineteenth- century trend we mean by saying that the classiication really was and thus whether additional should ‘emerge’ from the data,” meaning sociocultural or other phenomena must be the purely lexical data in their source text, factored in to explain its speciicity.19 the OED (“Classiication” 265–66). 5. Finally, as a grace note, I add that while 3. he phrase “‘emerge’ from the data” is the computation was an aterthought in mak- cue for an important tenet of the HTOED. ing the HTOED, today it is crucial for As Kay argues, “Our theoretical position on advanced uses of the work. he HTOED 418 The Meaning of the Digital Humanities [ PMLA

editors eventually migrated their content not how we must conclude. I thus climb out into a succession of relational databases be- of the speciic purview of my example, and cause they realized that computation might even of the digital humanities ield, to open support advanced, real- time querying my argument to its most general extent. Ap- (Wotherspoon, “Historical hesaurus” 218). ropos is the following insight from N. Kath- Even more dramatically, a recent essay by erine Hayles: Alexander shows that querying the HTOED for insight into the history of the language he further one goes along the spectrum that can itself be a form of the digital humanities ends with “machine reading,” the more one (“‘Various Forms’”). Alexander conducts implicitly accepts the belief that large- scale FIG. 1 quantitative computational analysis of the multicausal events are caused by conluences Marc Alexander’s HTOED to generate visualizations leading that include a multitude of forces interacting tree- map visualiza- to hypotheses about language changes (see simultaneously, many of which are nonhu- tion of present- day the igure below). Connecting the STS and man. . . . If events occur at a magnitude far ex- En glish in the design approaches I outlined earlier, we ceeding individual actors and far surpassing HTOED. Each dot might even say that at this point using the the ability of humans to absorb the relevant in- represents a word, HTOED becomes an experiment in digital formation, however, “machine reading” might and the shade of design. In the end, the HTOED is not the be a irst pass toward making visible patterns the dot corresponds “other” that Heuser and Le- Khac need to that human reading could then interpret. (29) to when the word help make the work of their Correlator entered the lan- meaningful; it is the precursor of Correlator. guage (darker dots It is not accidental, I can now reveal, that at show earlier words). the beginning of this essay I alluded to Lévi- Words are arranged But going down a rabbit hole, while nec- Strauss and structural anthropology. Structur- by semantic prox- essary in pursuing any case of the digital hu- alism is a midpoint on the long modern path imity as indicated manities to its methodological foundation, is toward understanding the world as system in the labels. 128.2 ] Alan Liu 419

(e.g., as modes of production; Weberian bu- them). Here opens a set of topics that I cannot the changing profession reaucracy; Saussurean language; mass, media, deal with in present limits but that more fully and corporate society; neoliberalism; and so demonstrate how the digital humanities regis- on) that has forced the progressive side of the ter the larger issues of the humanities. humanities to split of from earlier humanities One topic is the way digital pedagogy— of the human spirit (Geist) and human self to as witnessed in current controversies over adopt a worldview in which, as Hayles says, massive open online courses (MOOCs), all- “large-scale multicausal events are caused by digital “campuses” in public universities, and conluences that include a multitude of forces so on—registers the possibility of gigantic . . . many of which are nonhuman.” his is the changes in the aims, practices, audiences, in- backdrop against which we can see how the frastructure, and staf of humanities teaching. meaning problem in the digital humanities he bluf is now called on decades of defen- registers today’s general crisis of the mean- sively legitimating the humanities as thinking ingfulness of the humanities. The general and language “skills” added on top of tradi- crisis is that humanistic meaning, with its tionally meaningful humanistic knowledge. residual yearnings for spirit, humanity, and If the humanities it that mold in part, then self—or, as we now say, identity and subjectiv- maybe—some administrators and legislators ity—must compete in the world system with think—they should it it entirely so that their social, economic, science- engineering, work- content can be “delivered” modularly through place, and popular- culture knowledges that the Internet in the manner of the MOOCs or do not necessarily value meaning or, even Khan Academy courses in science, technol- more threatening, value meaning but frame ogy, engineering, and mathematics that have it systemically in ways that alienate or co- opt made the biggest public impression. humanistic meaning. Humanistic knowledge Another topic is the job market for digital today is thus increasingly assimilated to what humanists, especially in regard to tenurable humanists themselves call research, evidence, faculty lines versus “alt- ac” (alternative aca- analysis, method, productivity, and “impact” demic career) adjunct, staff, support, and (as this term is institutionalized in “research cultural- institution positions. he turmoil and assessment exercises” in British universities), uncertainty in the nature of digital humani- with no unfilled time and space left for any ties jobs register the larger uncertainty of em- old ghosts in the machine—unless, as I have ployment in the humanities as “meaningful” argued in Laws of Cool, there remains a yearn- jobs transition away from tenure and toward ing, nowhere keener than among our students, a corporatized ideal of reconigurable and re- caught in the educational mangle, to be cool. placeable professional- managerial knowledge Cool people say nix to today’s knowledge-work workers perpetually threatened with restruc- system even as they walk into the cubicles. turing layofs in favor of even more exploited Of course, if this were only a problem of “permatemp” and outsourced labor. research methodology, then I would be extrav- In all these ways, the digital humanities agant to call it a “crisis” in the meaningfulness register the crisis of the humanities. For that of the humanities. But “crisis” is appropriate reason, I and others started the 4Humanities when we realize that the meaning problem advocacy initiative, “powered by the digital also afects pedagogy and jobs in the wake of humanities community,” so that the digital economic recession, which brings the problem humanities can try to advocate for the hu- cruelly to bear on individual humanists in manities and not just register their crisis. I training or seeking jobs (not to mention on the do not know how much diference that ini- humanities programs that nurture and employ tiative and others like it will make in the 420 The Meaning of the Digital Humanities [ PMLA

meaningfulness of the humanities to the (given nuance below) by which I separate them from the world. But I do know that such an effort— digital humanities. For helping me think about these inclusion issues, I am dedicating the digital humanities to the soul grateful to members of the audiences at talks where I de­ of the humanities—is what is meaningful for livered versions of this essay who pointed out that they do a humanist, digital or otherwise, now.20 not entirely recognize their ield or work in my description of the digital humanities. Ultimately, of course, the “who is in the digital humanities?” issue will be adjusted on the ground through normal professional processes of adjudica­ tion—where one gets a job or places one’s students, where one publishes in print or online, which forums or blogs one NOTES posts on, which Twitter hashtags one is associated with,

the changing profession changing the which conferences one goes to, which grants one gets, etc. 1. Among recent surveys and overviews of the digi­ tal humanities are Svensson’s articles; Hayles; Kir­ 4. On algorithmic criticism, see Ramsay’s essay by schenbaum, “Digital Humanities” and “What Is Digital that title. Humanities”; Liu, “State” and “Where Is Cultural Criti­ 5. On #transformDH, see Phillips. he collective now cism”; Burdick, Drucker, Lunenfeld, Presner, and appears through its hashtag on Twitter, has a Tumblr Schnapp; and “Digital Humanities Manifesto 2.0.” Re­ page, and is a HASTAC group. cent or forthcoming essay collections about the digital 6. Examples of the textual­editing, bibliography­as­ humanities include Schreibman, Siemens, and Unsworth; sociology­of­ texts, and materiality­of­ the­ digital ap­ Gold; Price and Siemens; and Goldberg and Svensson. proaches in the digital humanities include, respectively, On whether the digital humanities are a ield and on the McGann, Radiant Textuality; the Text Encoding Initia­ linguistic usage of the phrase digital humanities, see Liu, tive’s TEI: P5 Guidelines, with its attention to prosopog­ “Is Digital Humanities a Field?” raphy and social relations (“Names, Dates, People, and 2. These expressions in recent conferences, blogs, Places”); and Kirschenbaum, Mechanisms. tweets, discussion threads, papers, etc., have become im­ 7. Examples of the historical sociology I refer to in­ portant “memes” shaping the digital humanities commu­ clude Franzosi; Bearman and Stovel; and Mohr and nity. On the issues of theory, building, and hack versus Duquenne. An example of social­ network analysis in dig­ yack, see also Cecire, “Introduction.” ital literary studies is Moretti, Network heory. (On such analysis in general, see my “From Reading.”) An example 3. Since the question of who is included in the digi­ of social­network analysis in digital history is Lemercier tal humanities can be a sensitive one, I should clarify my and Rosental. (For other historians using the methods understanding of the scope of the ield. While I focus in and tools of social­ network analysis, see “Bibliography.”) this essay on digital literary studies and one concrete re­ search example, the digital humanities are much broader. 8. See also Heuser and Le­Khac’s “Learning to Read I thus take care to relate my example to shared methods Data,” which summarizes the research reported in their and problems across the ield (including, e.g., digital his­ pamphlet. tory). However, while my goal is to address a fairly broad 9. For a discussion of “saving the phenomena” that bears notion of the digital humanities as they are commonly centrally on the issue of data, see Bogen and Woodward. practiced and discussed, I do not try to make the tent so 10. he full list of the novels in Heuser and Le­ Khac’s big—by, for instance, covering what should or could be corpus can be found in their “Online Companion.” In­ part of it but so far is not commonly recognized to be so formation on the source of the digitized texts is from an (e.g., the design ield)—that the term digital humanities e­mail from Heuser to the author. becomes formless or aspirational. 11. Goldstone and Underwood offer an explana­ Also, I do not specifically discuss digital work in tion and example of topic modeling of special interest such near humanities as the branches of the social sci­ to readers of PMLA. Separately, Underwood more fully ences, archaeology, and anthropology that have joined explains the methodology. For a succinct discussion of the humanities in the so­called linguistic and cultural the diference between supervised and unsupervised data turns and related trends. Whether such work is consid­ mining—a distinction originating in the ield of machine ered digital humanities depends on the prior issue of learning—see “Analytics.” whether, digital methods aside, it is humanistic (itting, 12. Sculley and Pasanek consider the problem of “cir­ e.g., the loose deinition in the 1965 National Foundation cularity” (and other issues) in humanities data mining. on the Arts and Humanities Act: “those aspects of social 13. For a fuller description of the way Heuser and Le­ sciences which have humanistic content and employ hu­ Khac used the HTOED, see appendix C in their Quantita- manistic methods” [National Endowment 1]). tive Literary History. Caveats also apply to my generalizations about 14. For this essay, I have consulted the online HTOED, new media studies and to the gross simpliication here which now appears as part of the online OED. On the 128.2 ] Alan Liu 421

HTOED’s classiication system, see “Structure of the Histor- Alexander, Marc. Message to the author. 6 Oct. 2012. the changing profession ical hesaurus.” For additional general information about E‑mail. the work, see Kay, “What Is the Historical hesaurus.” ———. “‘he Various Forms of Civilization Arranged in 15. For information about the making of the HTOED Chronological Strata’: Manipulating the HTOED.” referred to below, I have consulted Kay, “Classiication” Adams 309–23. and “What Is the Historical hesaurus”; Kay and Chase; “Analytics: Supervised vs. Unsupervised Learning.” and Wotherspoon, “Historical hesaurus” and “Making”; MONK: Metadata Offer New Knowledge. N.p., n.d. as well as Alexander’s e‑mail. For photos of the paper Web. 1 Dec. 2012. slips, iling drawers, and computers used, see Historical Bauer, Jean. “Who You Calling Untheoretical?” Journal of hesaurus Photo Gallery. My special thanks to Alexander Digital Humanities 1.1 (2011): 68–74. Web. 23 Nov. 2012. for helping me gather resources and for his e‑mail illed Bearman, Peter S., and Katherine Stovel. “Becoming a with details and relections on the manual and techno‑ Nazi: A Model for Narrative Networks.” Poetics 27 logical processes used in making the HTOED. (2000): 69–90. Print. 16. See, e.g., Gibbs and Cohen’s discussion of their “Bibliography.” Ed. Marten Düring. Historical Network hybrid method (70, 76). Research. N.p., 18 Sept. 2012. Web. 26 Nov. 2012. 17. I abbreviate here a longer discussion of the Bogen, James, and James Woodward. “Saving the Phenom‑ HTOED to be included in a version of this essay for my ena.” Philosophical Review 97.3 (1988): 303–52. Print. book in progress on the digital humanities. My thanks Burdick, Anne, Johanna Drucker, Peter Lunenfeld, Todd to Clare Birchall, who, in a conversation with me ater Presner, and Jefrey Schnapp. Digital_ Humanities. I presented an early version of this essay at King’s Col‑ Cambridge: MIT P, 2012. Print. lege, asked a question that made me start looking into the Cecire, Natalia. “Introduction: heory and the Virtues technological dimensions of the HTOED. of Digital Humanities.” Journal of Digital Humanities 18. My thanks to the Stanford Literary Lab for invit‑ 1.1 (2011): 44–53. Web. 23 Nov. 2012. ing me to visit on 21 May 2012. On the genesis and prin‑ ———. “When Digital Humanities Was in Vogue.” Jour- ciples of the lab, see Jockers. nal of Digital Humanities 1.1 (2011): 54–58. Web. 19. In their statistical analysis of diction in En glish lit‑ 23 Nov. 2012. erary works over a longer period (the eighteenth and nine‑ “The Digital Humanities Manifesto 2.0.” Humanities teenth centuries), Underwood and Sellers make a related Blast. N.p., 2009. Web. 9 Sept. 2012. point when comparing their indings with Heuser and Le‑ Feyerabend, Paul. Against Method: Outline of an Anarchis- Khac’s. hey track a rise in the nineteenth century in the tic heory of Knowledge. London: Verso, 1978. Print. proportional incidence of earlier‑ vintage, oten Anglo‑ Saxon words in literary genres (including prose iction) 4Humanities: Advocating for the Humanities. Ed. Chris‑ that is “largely consubstantial” with the rising incidence tine Henseler, Alan Liu, Geoffrey Rockwell, Stéfan of “hard” words in novels of that century found by Heuser Sinclair, and Melissa Terras, with Lindsay homas. and Le‑ Khac. However, they add that their longer histori‑ 4Humanities, n.d. Web. 16 Jan. 2013. cal baseline shows that the trajectory of such words “had Franzosi, Roberto. Quantitative Narrative Analysis. Los recently reversed direction” as part of a new way of be‑ Angeles: Sage, 2010. Print. ing “literary.” “he relative scarcity of simple action verbs Gibbs, Frederick W., and Daniel J. Cohen. “A Conver‑ in early‑ nineteenth‑ century writing, for instance,” they sation with Data: Prospecting Victorian Words and note, “was a recent development” in literary language Ideas.” Victorian Studies 54.1 (2011): 69–77. Print. that reveals social transformations only through the me‑ Gold, Matthew K., ed. Debates in the Digital Humanities. diation of “competing ideals of literary reinement.” My Minneapolis: U of Minnesota P, 2012. Print. thanks to Le‑ Khac for his e‑mail calling my attention to Goldberg, David heo, and Patrik Svensson, ed. Humani- Underwood and Sellers’s article, which I had not seen. ties and the Digital. Cambridge: MIT P, forthcoming. 20. Advocating for the humanities does not necessar‑ Goldstone, Andrew, and Ted Underwood. “What Can ily only mean defending older or disciplinary notions of Topic Models of PMLA Teach Us about the History of the humanities. My “Humanities and Tomorrow’s Dis‑ Literary Scholarship?” he Stone and the Shell: His- coveries” attempts to reframe the humanities in common torical Questions Raised by a Quantitative Approach cause with other disciplines and with public needs. to Language. N.p., 14 Dec. 2012. Web. 9 Dec. 2012. Hayles, N. Katherine. How We hink: and Contemporary Technogenesis. Chicago: U of Chi‑ WORKS CITED cago P, 2012. Print. Heuser, Ryan. Message to the author. 13 Sept. 2012. E‑mail. Adams, Michael, ed. “Cunning Passages, Contrived Cor- Heuser, Ryan, and Long Le‑ Khac. “Learning to Read ridors”: Unexpected Essays in the History of Lexicogra- Data: Bringing Out the Humanistic in the Digital Hu‑ phy. Electronic ed. Monza: Polimetrica, 2010. PDF ile. manities.” Victorian Novels 54.1 (2011): 79–86. Print. 422 The Meaning of the Digital Humanities [ PMLA

———. “Online Companion to ‘Learning to Read Data: ———. “Is Digital Humanities a Field? An Answer from Bringing Out the Humanistic in the Digital Humani- the Point of View of Language.” Alan Liu. N.p., 6 Mar. ties,’ Victorian Studies 54.1; and Stanford Literary 2013. Web. 6 Mar. 2013. Lab Pamphlet 4.” Stanford Literary Lab. Stanford Lit. ——— . he Laws of Cool: Knowledge Work and the Culture Lab, May 2012. Web. 13 Sept. 2012. of Information. Chicago: U of Chicago P, 2004. Print. ——— . A Quantitative Literary History of 2,958 ———. “The State of the Digital Humanities: A Report Nineteenth-Century British Novels: he Semantic Co- and a Critique.” Arts and Humanities in Higher Edu- hort Method. Stanford Literary Lab. Stanford Lit. Lab, cation 11.1–2 (2012): 8–41. Print. May 2012. Web. 1 Sept. 2012. Pamphlet 4. ———. “Where Is Cultural Criticism in the Digital Hu- Historical hesaurus. Oxford English Dictionary. Oxford manities?” Gold 490–509. UP, 2012. Web. 25 Nov. 2012. McCarty, Willard. Humanities Computing. Basingstoke: Historical hesaurus of the Oxford En glish Dictionary, with Palgrave, 2005. Print. the changing profession changing the Additional Material from “A hesaurus of Old En glish.” McGann, Jerome J. “Position Statement from McGann.” Ed. Christian Kay, Jane Roberts, Michael Samuels, and Panel: What Is Text? A Debate on the Philosophical Irené Wotherspoon. Oxford: Oxford UP, 2009. Print. and Epistemological Nature of Text in the Light of Hu- Historical hesaurus Photo Gallery. U of Glasgow, n. d. manities Computing Research. By Susan Hockey, Allen Web. 25 Sept. 2012. Renear, and McGann. N.p., 1999. Web. 25 Nov. 2012. Jockers, Matthew L. “houghts on a Literary Lab.” Mat- ——— . Radiant Textuality: Literature after the World thew L. Jockers. N.p., 4 Jan. 2013. Web. 10 Jan. 2013. Wide Web. New York: Palgrave, 2001. Print. Jockers, Matthew, and Glen Worthey. “Introduction: Mohr, John W., and Vincent Duquenne. “he Duality of Cul- Welcome to the Big Tent.” Digital Humanities 2011. ture and Practice: Poverty Relief in New York City, 1888– N.p., 2011. Web. 23 Nov. 2012. 1917.” heory and Society 26.2–3 (1997): 305–56. Print. Kay, Christian. “Classiication: Principles and Practice.” Moretti, Franco. Network heory, Plot Analysis. Stanford Adams 255–70. Literary Lab. Stanford Lit. Lab, 1 May 2011. Web. ———. “What Is the Historical hesaurus of the OED.” 6 July 2012. Pamphlet 2. Oxford En glish Dictionary. Oxford UP, 2012. Web. “Names, Dates, People, and Places.” P5: Guidelines for 25 Nov. 2012. Electronic Text Encoding and Interchange. Vers. 2.2.0. Kay, Christian, and T[homas] J. P. Chase. “Constructing TEI, 25 Oct. 2012. Web. 14 Jan. 2013. a hesaurus Database.” Literary and Linguistic Com- National Endowment for the Humanities, Oice of Digital puting 2.3 (1987): 161–63. Print. Humanities. Digital Humanities Start- Up Grants: Fre- Kirschenbaum, Matthew G. “Digital Humanities As/ Is a quently Asked Questions. NEH, n.d. Web. 17 Nov. 2012. Tactical Term.” Gold 415–28. Owens, J. B. “Jack.” “What Historians Want from GIS.” ——— . Mechanisms: New Media and the Forensic Imagi- ArcNews. Esri, 2007. Web. 15 Jan. 2013. nation. Cambridge: MIT P, 2008. Print. Phillips, Amanda. “#transformDH—a Call to Action Fol- ———. “What Is Digital Humanities and What’s It Doing lowing ASA 2011.” HASTAC. N.p., 26 Oct. 2011. Web. in En glish Departments?” ADE Bulletin 150 (2010): 24 Nov. 2012. 55–61. Print. Pickering, Andrew. he Mangle of Practice: Time, Agency, Koh, Adeline. “More Hack, Less Yack? Modularity, he- and Science. Chicago: U of Chicago P, 1995. Print. ory and Habitus in the Digital Humanities.” Adeline Price, Kenneth M., and Ray Siemens, eds. Literary Studies Koh. N.p., 21 May 2012. Web. 23 Nov. 2012. in the Digital Age: An Evolving Anthology. MLA Com- Latour, Bruno. Reassembling the Social: An Introduction mons. MLA, 2013. Web. 12 Jan. 2013. to Actor- Network- heory. Oxford: Oxford UP, 2005. Ramsay, Stephen. “Algorithmic Criticism.” A Companion Print. to Digital Literary Studies. Ed. Susan Schreibman and Le- Khac, Long. Message to the author. 4 Feb. 2013. E-mail. Ray Siemens. Oxford: Blackwell, 2008. Alliance of Dig- Lemercier, Claire, and Paul- André Rosental. “The ital Humanities Organizations, n.d. Web. 25 Nov. 2012. Structure and Dynamics of Migration Patterns in ———. “On Building.” Stephen Ramsay. N.p., 11 Jan. 2011. Nineteenth- Century Northern France.” Vers. 3. HAL: Web. 20 June 2011. Sciences de l’homme et de la société. Centre pour la ———. “Who’s In and Who’s Out.” Stephen Ramsay. N.p., Communication Scientifique Directe, 5 May 2010. 8 Jan. 2011. Web. 23 Nov. 2012. Web. 24 Nov. 2012. Schreibman, Susan, Ray Siemens, and John Unsworth, Liu, Alan. “From Reading to Social Computing.” Price eds. A Companion to Digital Humanities. Oxford: and Siemens. Blackwell, 2004. Alliance of Digital Humanities Orga- ———. “he Humanities and Tomorrow’s Discoveries.” nizations, n.d. Web. 27 Nov. 2012. 4Humanities: Advocating for the Humanities. 4Hu- Sculley, D., and Bradley M. Pasanek. “Meaning and Min- manities, 25 July 2012. Web. 14 Mar. 2013. ing: The Impact of Implicit Assumptions in Data 128.2 ] Alan Liu 423

Mining for the Humanities.” Literary and Linguistic ———. “he Landscape of Digital Humanities.” Digital the changing profession Computing 23.4 (2008): 409–24. Print. Humanities Quarterly 4.1 (2010): n. pag. Web. 23 Nov. Seefeldt, Douglas, and William G. homas III. “What Is 2012. Digital History? A Look at Some Exemplar Projects.” Underwood, Ted. “Topic Modeling Made Just Simple Faculty Publications. Dept. of History, U of Nebraska, Enough.” The Stone and the Shell: Historical Ques- Lincoln, 1 May 2009. Web. 12 Oct. 2012. Paper 98. tions Raised by a Quantitative Approach to Language. Stallybrass, Peter. “Books and Scrolls: Navigating the Bible.” N.p., 7 Apr. 2012. Web. 14 Jan. 2013. Books and Readers in Early Modern En gland: Material Underwood, Ted, and Jordan Sellers. “he Emergence of Studies. Ed. Jennifer Andersen and Elizabeth Sauer. Literary Diction.” Journal of Digital Humanities 1.2 Philadelphia: U of Pennsylvania P, 2002. 42–79. Print. (2012): n. pag. Web. 4 Feb. 2013. “he Structure of the Historical hesaurus of the OED.” Ox- Vismann, Cornelia. Files: Law and Media Technology. ford University Press. Oxford UP, 2009. Web. 25 Nov. 2012. Trans. Geofrey Winthrop- Young. Stanford: Stanford Svensson, Patrik. “Envisioning the Digital Humanities.” UP, 2008. Print. Digital Humanities Quarterly 6.1 (2012): n. pag. Web. Wotherspoon, Irené. “Historical Thesaurus Database 23 Nov. 2012. Using Ingres.” Literary and Linguistic Computing 7.4 ———. “Humanities Computing as Digital Humanities.” (1992): 218–25. Print. Digital Humanities Quarterly 3.3 (2009): n. pag. Web. ———. “he Making of he Historical hesaurus of the 30 Nov. 2010. Oxford English Dictionary.” Adams 271–87.