Wow! Adding an annotation deletes letters in the text!

I sympathise with the frustration over getting things to work on different systems. It's annoying in class but even worse for workshops. Outside of academia, time is allotted for it, and it's considered part of doing business.

But you do seem to have picked up the big for employing "screwmeneutics", which will allow you to get down and see what you can do. And I think the first step is believing that you really can learn to do what you want by googling how to do it!

I think this deserves a longer annotation. I suspect that the "let's get on with it" sentiment, as attractive as it is, may be unrealistic for a couple of reasons First, the tools being leveraged are not uniquely employed by traditional humanities disciplines. When the topic modelling community consists of social scientists and search engine creators, there will inevitably be talk about how they can be applied within the context of literary criticism. Second, the lit crit field itself has passed through a period in which those who do not examine their methods critically are stamped as committers of the most heinous intellectual (and political) crimes. When digital humanists engage in such introspection, however, the terms in which they do so are so foreign that those who follow more traditional paradigms tend to have knee-jerk reactions. So the problem will probably never go away. But you are right that we can get on with our own research and satisfy ourselves that we have learnt something. Ultimately, achieving a critical mass of people who can use DH methods to do so will quiet down some of the angst.

As an added note, much of the defensiveness and polemic dates back five years or so. It's worthwhile looking at Andrew Piper's Enumerations to see the difference in tone in 2018.

What happened in March to cause a huge explosion of tweets. NEH grant announcements?

It would be useful to have a link to this page.

One possible approach is to maintain several stopword lists and compare the results of analysis using each list. You can then evaluate which results seem to be the most stable.

This section could use some reference to existing efforts to collect or analyse the data sources you are working with.

This might be a good place to put the time frame. Perhaps suggest the past year, which should be substantial enough for topic modelling, and say that we would be happy to receive a larger historical archive if it is available.

I seem to recall your saying that you had come across information that transcripts might be available. It would be a good idea to mention this hear so that they know what possible sources you are thinking about.

This is a perennial problem. One way around it is to place milestone tags in your text and cut on the milestones. If you cut on the number of words, there are also ways to determine where segments begin and end. You can also download the cut segments and open them in an editor.

You would need to have a look at Beowulf Unlocked. Here's a review. Also this article as a starting point..

This probably refers specifically to tenure-track academic jobs. Typically, a research university will have three medievalists and a comprehensive university or liberal arts college will have one. If your the institutions in your region have these positions filled, there will be no jobs in your region, and you will have to look elsewhere.

It may be striking that the sense that "grievance studies" are empowering for those with "grievances" is either not reaching or not resonating with the public. Or else it is with a portion of the public but driving them into niches.

One of the challenges is in identifying how closely ethnic and gender studies should be linked to the term "humanities" since the institutional units in which they are often situated are frequently interdisciplinary. Of course, you could say the same thing about Digital Humanities.

Keep in mind that the branches can be swung around, so Sense and Sensibility could appear to the right of Emma, and the short vertical branch between Sense and Sensibility and the Emma/Pride and Prejudice clade indicates a close relationship. By this reading, I think that what you have is a pre-20th century clade whee the three novels by Brontü are closest together because of their common authorship. The Zelazny story shows some distant affinity to these works, and Lovecraft is even more stylistically distinct from the other texts than Austen is.

This is potentially important, but I'm not sure what it means. Can you say more?

An astute observation about DH, or at least screwmeneutic studies like this one. In the Humanities, we rarely go about things this way.

One common observation about authorship attribution is that the authorial "fingerprints" are actually to be found in patterns of usage for function words (prepositions, pronouns, articles), which are typically treated as stop words.

With your smaller set of stop words, the Sense/Northanger clade is flatter, suggesting that the differences aren't as great as when you did not apply a stop word list.

This is not necessarily bad, depending on what you are studying. However, a way to get around the problem is to cut the novels into arbitrary segments of a particular size. You may find, then, that the odd segment does not cluster with the rest of the novel (or chapter), but with text from elsewhere. When you encounter that, it's really interesting, and some explanation is required.

See the instructions under Code, Code Blocks, and Syntax Highlighting here for instructions on how to represent code in Markdown.

Yes, that is exactly what happened. You can set the random-seed to an arbitrary number so that the model can be duplicated. However, even when you get the same topic, it may not come out with the same topic number. I haven't tested that.

I think you may be putting your finger on something important here: that DH works through a combination of screwmeneutics and serendipity. You might get some insight from a talk I gave in Lausanne a few years ago on Play as Process and Product.

You've perhaps inadvertently hit on an interesting convergence of literary text analysis and biology. Some of the algorithms we use are borrowed from bioinformatics and might, in that field, be seen to use probability as a way to overcome the shortcomings of Linnean (and similar) classifications. I would rephrase as "And yet the platypus, somewhat improbably, continues to exist".

I think that a common criticism of DH is still that it demonstrates what is apparent or obvious with the implication (from its critics) that that is somehow unnecessary. That might be worth some class discussion.

You are not the only one to observe this. I may have to change the order of readings in the future. The major reason I set the article for this week is to give you an idea of what it might mean to read at scale.

Amongst digital humanists, no. But we might think a little bit about whether quantitative methods are seen by others as somehow violating the sacred qualities of our canonical literature.

You're right to highlight this phrase. Others may have comments to make, or we can discuss it in class.

Is the distinction that you're trying to make that The Lord of the Rings is not "canonical"?

I don't think you quite go in this direction, but it follows nicely from your point. Hyperspecialisation (and related) factors have prevented students from feeling comfortable with the territory of, say, computer science, making them afraid to learn computational skills that they might usefully apply in their work. But for students willing to learn coding or other skills typically considered outside the domain of English studies there are real opportunities for insight. For further information on this, try Googling "C.P. Snow Two Cultures", and you'll get a lot of material about how we came to this pass.

Historically, scholarship in literary studies was much more focused on details than it is today. Are DH methods doomed to be ignored for being detail-oriented by those who might be potential audiences for the results of those methods? Is not being detail oriented a good thing? It is a skill I try to teach my students, especially in their writing, by I admit that my successes have been few...

I think that there might be a fear that this is inappropriate for Humanities. Traditional Humanities scholarship privileges the completed, polished written argument. Lacking empirical notions of "truth", work in the Humanities tends to rely on persuasiveness, which is dependent on rhetoric — and the rise of poststructuralist theory has exacerbated (or exaggerated?) this phenomenon. Is Ramsay arguing for a more "scientific" way of doing literary scholarship (which is at the same time not science)?

I would add that the rise of these new technologies has coincided with a growing interest in book history in English departments.

This is sort of an anecdote to reflect on. I recently met a woman who was involved in training many of the designers of algorithms for companies like Facebook. She pointed out that the designers are often fully aware of the shortcomings of their algorithms, but that their reservations don't get voiced properly because the business model requires releasing products on a deadline. The limitations are lost in flurry of marketing cycles and hype.

Is this true? Are there ways in which we might measure "influence"?

I'm not sure I would target Google specifically, but there has been a general observation that people no longer go to the library and "serendipitously" discover something interesting sitting on the shelf next to the book they were looking for. This was the inspiration behind Serendip-o-matic, which (full disclosure), I worked on. Take a look at that project to see how the issue was addressed.

I think what you are saying here is that the algorithm is a black box. Marche's essay does not reveal its inner workings or point us to a code repository where we can explore it.

However, you might also be referring to a kind of epistemological/hermeneutic question. Even if we understand that algorithmic process, what do we understand about the nature of the results of that process?

One of the issues that typically accompanies the "subjectivity" question is the "representativeness" question. How do you assemble a representative corpus, and what makes it representative.

This is a really important observation. DH is often accused by critics of not being interpretive, but it says something that one such critic dramatically demonstrates the role of human interpretation when he actually tries to implement a DH method himself.

Here are some things to think about:

• The Zine Scenes website is hosted on Tumblr. What are the consequences of choosing that technology. What are the advantages and disadvantages for the stated purposes of the project. What different choices could have been made?
• What intellectual property issues arise in acquiring and distributing zine source materials?
• How would you want to deploy computational resources to analyse, as opposed to disseminate, zine materials?

How do you pronounce this?