Machine reading

For your final projects, you will use one or more tools, platforms, or applications to perform a “machine-assisted” reading of either Henry James’ In the Cage or Herman Melville’s Bartleby the Scrivener, both of which you can download from Project Gutenberg.

Text analysis tools include Voyant, TAPoR, MONK, Crawdad (free trial); visualization tools include Many Eyes, SIMILE Widgets, and TextArc; simulation platforms include Scratch and Second Life; and social network graphing tools include Gephi and yED (Liz Shayne, a former NYU grad student, has a wonderful post on using yED to read Daniel Deronda). For a complete list of tools available for this assignment, see Alan Liu’s Toy Chest and/or The Rhetoric of Text Analysis: Tools. You might also try working through the list of TAPoR recipes for text analysis.

Throughout the term we will be considering the changes in everyday reading practices. For the final, we will turn our attention to what we might call professional reading practices – more specifically, academic literary criticism as it has been re-imagined by the digital humanities. Put another way, attention continues to shift from the solitary to the computer-aided reader. As you will see in Geoffrey Rockwell’s overview of the work of text analysis, Stephen Ramsay’s comments on algorithmic criticism, and Franco Moretti’s distant reading project, researchers are now able to run sophisticated syntactical queries and perform machine-assisted searches of large textual corpora. Instead of working with a database – e.g. the Internet Shakespeare Editions – you will work with a single text, in this case either Henry James’ or Herman Melville’s novella, the full text of which you can download from Project Gutenberg.

Our basic question is this: how does text analysis enhance human interpretation? We will also consider how and to what extent the traditional modes and methods of humanist inquiry can be supported by machine reading. How can they be mutually productive? In what sense does machine reading ask us to reconsider our governing assumptions about what a text is, what is involved in “proper” reading, and what knowledge production looks like? What are the advantages and disadvantages of adopting what we might call a computational perspective on a literary text?

Your assignment is to use the tool or tools of your choice to analyze the novella of either James or Melville and then to reflect on these questions in a 4-5 page critical commentary on the data that results from your machine-assisted text analysis.

Final projects–with at least one image of your data–due in hard copy December 20. At the same time you should post on this blog page a short reflective statement about the utility and value of machine-assisted reading for humanities scholarship. What in your view are the possibilities and limitations of using a text-analysis or visualization tool in your coursework? The questions are similar to those posed for your data analysis, but in that piece of writing you will focus on your individual or subjective reading practices, whereas your short blog post should consider the questions from an institutional perspective. We will discuss further in class.

18 Comments to “Machine reading”

Ian Hundiak
December 12, 2012 at 7:48 pm

Machine assisted reading certainly has a place in the classroom. While traditional readings of traditional texts are still essential to becoming a well rounded scholar, there are aspects of certain texts that become more apparent when analyzed with machine reading methods. The society that we live in today revolves around data and percentage points. News programs give the number of civilians killed in a war, or the percentage of people in a certain area that are susceptible to West Nile Virus, because the audience can easily grasp and analyze these figures. It makes perfect sense then, that this number based analysis would make its way into humanities coursework. While some people seem fearful that classical analysis will be lost with the introduction of machine reading, this seems to be based on the false assumption that only one type of reading can be used at a time, when in fact, it is perhaps most advantageous to use both classical, or traditional, reading and machine reading together. For instance, while my project “3 Paper, Crafting Table” (summarized in the video below) does give insight into Henry James’ In The Cage that one cannot get from simply reading the novel in a traditional way, it also cannot stand as a substitute for simply reading the novel. In order for the humanities field to move forward, it is absolutely essential that we do away the notion that we must either use machine reading or traditional reading to do text analysis, and accept that both methods provide unique interpretations of a text.

Video: https://www.youtube.com/watch?v=dblNaK2cieA

Reply
Benjamin Newgard
December 13, 2012 at 6:25 pm

Perhaps this is due to my limited experience with machine readers, but I think their scholarly utility is somewhat limited, at least in their current form. I used basic tools like the TAPoR frequency table and concordance analyzer specifically because I had trouble seeing how they could be tremendously useful going into the project. I wanted to see if actually using these tools would reveal their true utility. What I actually found, however, was highly specific data that I struggled to connect to my prior traditional reading of the text, data that could not stand meaningfully on its own. As Ian has already said, a computational reading is, in no way, an adequate substitute for a traditional reading; in fact a computational reading depends on an organic interpretation in order to have any significant meaning whatsoever. Machine readers like the TAPoR frequency and concordance analyzers are capable of augmenting our understanding of a text, but only through highly specific / niche channels. This limitedness will preclude them from reaching widespread academic utility for the time being, but an expansion of their interpretative capabilities might change this.

Reply
Michelle Piekarus
December 18, 2012 at 4:40 am

To completely deny the utility of text analysis and visualization tools would be to ignore the method of literary analysis itself. First a reader recognizes a pattern, and then the reader collects pieces of the text that apply to the pattern and which might come to a larger idea. In the end, the reader incorporates these pieces into a fuller, whole analysis. There is no literary analysis that does not start with an abstract idea that must be further explored, and there’s no other way to further explore it than to read the text with this idea in mind, and to collect the pieces of the text that apply. Text analysis provides tools to extract the information that a reader needs to form such an analysis of a literary text. Of course, it’s impossible to make proper use of text analysis tools without an initial interpretation of the text, which is why machine-assisted reading is limited in its functions (and why the machine is only an assistant in analysis, not a replacement for human interpretation). As they exist currently, however, text analysis tools have limited utility. The greatest limitation of these tools is their narrow scope, and the way in which they might act to narrow a reader’s own interpretation. Tools like concordance analyzers and Link visualizations provide very specific data sets that don’t always apply to a reader’s interpretation or analysis of a text. This potential (and possibly frequent) discord between a reader’s analysis and the machine’s results could easily frustrate the reader, as well as pull him even further away from actual analysis. So while I understand the function of these tools, and the benefits of machine-assisted analysis, I still think that these tools currently serve to restrict analysis more than anything.

Reply
nikkicunninghamn
December 18, 2012 at 8:35 pm

It has become clear over the course of the semester that machine-assisted reading will become more valuable for humanities scholarship as it develops. Machine-assisted reading offers opportunities for illuminated literary analysis that reaches far beyond the limits of human interpretation. While I continue to believe that human interpretation is paramount, to ignore or neglect the utility for developing reading technologies is to continue the fear of change that I believe plagues humanities scholarship. I believe there are endless possibilities for using text analysis or visualization tools in my coursework, and that the limitations lie within our current conception of literary study, not with the tools themselves. For example, I believe that survey courses such as American Literature and British Literature could be enhanced with graphs similar to those used by Franco Moretti to chart British Novelistic Genres. It would be beneficial to understand and to read literature in a non-linear way. Additional possibilities include finding trends in author’s corpora, comparing shifts in genres to shifts in the socio-political climate, etc. Machine-assisted reading has the power to infinitely enhance literary study, so long as we don’t completely abandon human interpretation. After all, it is human interpretation that has made literature into something worth studying in the first place. In future literature classrooms, I hope that text analysis tools will be considered beneficial supplements to standard close reading practices. With the integration of text analysis tools into close-reading based curricula, “proper” reading practices will adapt to new definitions of reading and knowledge production will become more individualistic, creative, and insightful. Only through embracing technological developments will the humanities continue to be an influential field of study.

Reply
Winne Chen
December 19, 2012 at 3:07 am

Text analysis certainly enhances human interpretation in that it can provide statistics and visualize trends faster that pure traditional reading practices can. For example, it can provide useful data such as word frequency and word trees. This can help people make connections that they may not have otherwise noticed. However, that being said, machine-assisted reading tools also have their faults and limitations. The software itself is restricted to what it is programmed to do, which is problematic because it cannot go beyond its original functions. In addition, it is more difficult to gain access to a computer and machine-assisted reading software than it is to possess a book. For instance, I have both a Macbook (from 2011) and a PC (from 2009), and the PC could not support Gephi. This was surprising to me because I thought my HP Pavilion PC was relatively new, and I have never had trouble running any other programs on it, as it is still just as fast as it was when first purchased. Therefore, completely adopting a computational perspective on literary works is challenging because not everyone has access to a computer, let alone a computer good enough to support whatever software is needed to perform machine-assisted reading. Nevertheless, if this technology is readily available and the data derived from the software supports a given argument, then the use of machine-assisted reading tools is certainly helpful, although it cannot be solely relied upon as one must still examine the original text. This is because analytical tools can take words and other information out of context, and display connections, patterns, and other data that may not be indicative of what the text actually suggests. Also, the data cannot come up with an argument by itself, so human analysis is still necessary in order to interpret that data. Personally, I found it difficult to derive an argument purely from the data, although I eventually managed to do so after struggling to make meaningful connections from the various tools I was experimenting with. Hence, machine-assisted reading tools can supplement human interpretation, but cannot entirely replace it.

Reply
Rachel
December 19, 2012 at 9:17 pm

As of right now, I think machine-assisted reading is a good supplement to a more traditional, closer read. What I’ve been able to tell from the visual and analytical programs has been interesting, surprising at the best of times—but it hasn’t quite replaced what I get from sitting down with a book and reading it for myself. Instead, I like the compound knowledge of a firsthand reading experience and a machine-assisted analysis. I find this problem a bit like the question we’ve asked throughout the semester, about where to take education from here now that we have a wealth of information at our fingertips. We know that things have to change—but how? I definitely think that the integration of machine-assisted reading into a school curriculum is a good first step. For now, though, it feels like many possibilities remain unexplored (which makes sense, since the technology is still fairly new). For me, and for now, machine-assisted analysis is like an extra set of eyes, catching things you might miss or would otherwise be unable to comprehend. Would I have known that “prefer” is the 51st most used word in Bartleby the Scrivener without Voyant Corpus Term Frequencies? Seems unlikely. Still, without having read Bartleby for myself, this word list would be virtually meaningless. I could make some inferences as to the content of the story, but beyond that, what would I have gained?

I do think that, sometime in the future, we will have a machine-assisted reader that will be able to enhance our knowledge of a text in ways unimagined. Perhaps it will be more capable of “reading” (in a mining for information/themes/characterizations sense of the word) then even we are. But, much like education reform, although I feel certain that it will happen, I can’t know how or when. Or what form it will take.

Reply
Kerry
December 20, 2012 at 3:09 am

I think that the steep learning curve associated with machine-assisted reading practices will inhibit its widespread use in the classroom. It is necessary to first become acquainted with the tools and their manner of representation before making informed interpretative statements about the data set. Because each tool is different, time must be devoted to first learning how to input the data correctly, then appropriately analyze the data that is produced. Much of my time spent while completing this assignment was devoted to learning and understanding the language of these online tools. One could speculate that in a classroom setting, students (and teachers as well) would have to attend training sessions on how to use the tools, which could eat up valuable time and funds. However, one advantage would be that certain students might be more receptive to the technology-based form of interpretation, such as those who are good at math and science or devote time to video games or other virtual environments.

Additionally, while traditional close reading only requires a book, pen, and paper, machine-assisted reading practices requires a computer and Internet access. The introduction of the Internet as a necessary third component makes this practice less accessible to the greater reading public. Although I discuss the negative aspects of machine-assisted reading, I realize that I approach this debate from the perspective of standard literary analyst who has been conditioned to close read from an early age. Ultimately, a lot of the fear and hesitation surrounding machine-assisted reading stems from its classification as a “new technology” that is making waves in a community of scholars who have survived off of pen and paper for hundreds of years. As scholars, we cannot let this fear inhibit us – we must strive to learn new disciplines and challenge the way that we think as individuals.

Reply
Dan Chen
December 20, 2012 at 3:20 am

In interpretation of all situations, and not only written text, we have a tendency to focus on things we can personalize. For example, people upon hearing of some massive natural disaster claiming tens of thousands of lives will not have as strong a reaction as they would have to a narrative about the tragic life of one child suffering an agonizing hereditary disease who has no chance to live past ten years of age. Some things simply have a greater resonance with us. When we read a text we will show partiality to certain things; characters we like or hate, certain settings or tropes, and specific authors. The mind’s level of attention and focus to the text waxes and wanes in the reading. It would be difficult to read anything with a completely neutral mindset, and having the aid of a machine can get us past our natural biases towards specific moments to see the whole text in wider perspective. I think the primary symbiotic benefit of coordinating human mind and machine processor is the machine can re-package the text for us so that we will not overlook what we normally would.

Although there are still limitations to what machine aided text analysis can tell us, it is not as difficult a field of study as certain natural phenomena. Gathering data and mathematically analyzing it are the easy parts, and it is not as if we need something the level of a large hadron collider to run our tests. So it comes to asking the right questions that are answerable by data mining. A lot of the questions we can think of are not going to produce much actionable information, and I think that is still because of a limitation on how people regard text. We recognize that not every author is specific about things like say, place and time, so we could not really analyze a text for a map or timestamps without the aid of regular human reading. We can analyze writing styles, and as a craft, writing will have certain consistent elements that we can key in on. If you believe that there is structure and intent behind the words, and that will not always be the case, then a computer should be able to make some observation about that structure. But since we are studying readily available and non-perishable subjects, just try any test you can think of. It is a rare opportunity in any field of study where you can basically do whatever you want with the evidence and have no significant consequences, so have at it.

Reply
Brittany Mania
December 20, 2012 at 3:42 am

The future of machine-assisted reading in an institutional setting is definitely possible, especially because the tools it relies on can be accessed on most computers, tablets, and smartphones. As a result, a school would not have to purchase iPods or any other special device for each one of its students in order for each student to engage in machine-assisted reading practices. Seeking funding for computers that run the programs would be easy because a computer, a device that can be used for many other purposes in other disciplines, is a justifiable purchase in today’s educational setting. The recent trend of personalized education is supported by machine-assisted reading because students do not need to be in a class room to partake in computational exercises; they can perform such exercises from home, the library, or even the nearest coffee shop. However, I do think teachers may find a decline in the quality of their students’ work as a result of machine-assisted reading due to a disconnect between traditional and digital reading practices. I do not think that the academic community at large has formed the necessary link between machine-assisted reading and human interpretation yet, and therefore machine-assisted reading is not understood to be a tool to be used as a secondary source of information rather than primary evidence. Nothing, not even a computational re-visioning of a text, can replace how a reader interacts with that text in its original form. For now, I see traditional close reading practices remaining the standard of literary interpretation, enlightened by machines, but not substituted.

Reply
Chelsa
December 20, 2012 at 8:37 am

From looking over what other people have said, it seems like most of us share the same opinion about the place of machine-assisted reading as currently being merely supplementary to traditional forms of analysis. While I agree with Nikki that visuals such as graphs could have seriously enhanced my “big-picture” understanding of literary history in some of my survey classes, I still think that the best analysis of a single text comes from the close reading that only the human mind is currently capable of. It seems like many of the tools we have looked at in this class are best suited for analyzing large corpuses of either printed or web-based writing—huge amounts of text that would take forever to analyze without a computer. Many of the tools, especially the gaming related ones, also provide new and interesting visual ways to creatively interpret a text; but this is not the same thing as closely examining a text and then critically writing about it. The process of spending time analyzing a text and formally writing about it also has value in that it helps develop a person’s writing skills, a desirable quality no matter what field a person aspires to. And so, while I think that humanities scholars and professors should seriously consider using things like graphs and other machine-assisted readings to supplement their work—especially to appeal to their more visual readers/students—I don’t think that it can or should ever replace how the human brain critically analyzes texts. I really hope that traditional forms of literary interpretation will still be relevant in the future.

Reply
Giulia Modenese
December 20, 2012 at 2:25 pm

Prior to this class or project, my experience with machine-assisted reading was, in essence, non existent. I wasn’t even aware such tools existed. Admittedly, when we started talking about them in class and began experimenting with them at home I was skeptical. How is this helpful? I would ask myself. Why can’t I just analyze Melville’s novella with my own mind and resource? However, after an initial objection and lack of appreciation for this new reading practice, I started to see how it could be helpful. Experimenting with different applications and learning what they could do was actually very interesting. Once I found a tool (Word Tree) that actually revealed aspects of the text that I had never noticed before, I was sold: machine-assisted reading definitely has its place in the classroom and research. It shouldn’t and won’t replace deep reading, but it can most definitely enhance it. Deep reading allows us to make our own interpretations, truly dig into the text and really understand what we are reading and analyzing. Machine-assisted reading, instead, can give us precise and accurate results, show us word frequency and help us perform research that would be too time consuming to perform ‘by hand’. Therefore, by combining traditional reading practices with a computational reading we can reach an even deeper level of textual understanding, making our literary analysis more original, persuasive and profound.

Yet, there is something that I fear. With machine-assisted reading technology becoming better by the day, and websites like Sparknotes and Gradesaver so readily available to students of every age, will deep reading actually still be performed, or will laziness and convenience prevail? Combining knowledge from an online summary with a computational reading of a full text could result in an empty reading; a mere exercise done to satisfy a requirement. Nothing truly valuable will be learnt and nobody’s mind will be nurtured or inspired. In my opinion–and judging from the blog posts above, most of my peers’–deep reading is and will always be relevant. Nonetheless, most students will undoubtedly choose to take easy way out and scrap this traditional mode of reading altogether, making machine-assisted reading tools, in a way, quite problematic.

Reply
Aaron Marks
December 20, 2012 at 3:01 pm

The machine reading techniques we’ve explored in class and through the final projects have illuminated for me myriad lenses through which to view a text that I might not have otherwise considered. And while observing these charts, datasets, graphs, etc. can shine a light on a text, I found these techniques to be mostly useful only after conducting a “typical” close reading. On their own, data and statistics do not seem to offer the human-focused perspective which we consider necessary to read into a work’s characters, in the case of fiction, or author’s ideas, in the case of non-fiction. Where machine assisted readings CAN perhaps assist on the level of individual texts would be after the fact, once one has conducted a primary close reading, the machine readings could be used as support for a thesis. Much in the way we use quotation as evidence, statistical properties of the book are equally valid and deserving of attention.

But if machine assisted reading’s place is only as support for close reading, it’s hard to argue for its implementation, since it won’t be creating any new theses. Where it CAN have value without necessarily reading and conducting close readings of individual texts, is when analyzing large bodies or corpora of texts. While a certain book’s word frequency list means little on its own, comparing it to similar books of the same era (Or of the preceding / following eras) seems like it could illuminate generic and periodic differences, like in the Moretti piece. Perhaps machine assisted reading’s institutional home is at the university and graduate level of study, rather than in all English classes. At the university level scholars are concerned with making claims about certain periods and genres in general, and machine reading techniques could help generate such ideas. Placing machine assisted reading at this institutional level also avoids the problem of machine-assisted reading necessarily needing computers and technology skill. Because while if younger students at primary schools, for example, needed to use machine assisted reading, poorer schools would be disadvantaged by not being able to participate. But at the collegiate and university level, funding seems more available for digital humanities projects so few to none would be put at a disadvantage due to class / economic reasons.

Reply
Cynthia Caraturo
December 20, 2012 at 5:21 pm

After going through different ways one can “close read” using technology, I must say that I am still slightly torn about the whole thing. It is true that I like the idea of new ways of interpreting a text, and the possibility of using machines to make that interpretation seem more interactive sounds appealing. I also like the idea of students being able to come up with so many more different ways of looking at a text. Papers, I imagine, can begin to sound quite repetitive to professors and teachers, so the idea of possibly saying the same thing but through a different medium seems like a positive occurrence of machine reading. Additionally, these “same idea, different approach” methods would make the arguments sound stronger, since different data could still be used to support the same idea behind the interpretation of a text. That being said, I too see problems in the idea of implementation within the school setting. While it is most likely true that some students will use these sources in a lazy sort of way, I also think that students may be intimidated to use these programs. When I was trying to figure out what to do for the final assignment, I was a little intimidated by the programs at my disposal, mostly because I didn’t know what to do with them nor how to work with them. Students might use the easier tools more not because they are lazy, but because they don’t know what to do with what they have. So, if the use of machine reading were to be brought into the general humanities classrooms, it would probably be a slow implementation. Of course, even after that, technology will have likely progressed even further, therefore making it a kind of catch-up situation. Therefore, I think it would be best if using machine assisted reading techniques were done in some sort of introductory course, or perhaps it could branch off into a separate field of the humanities, like the digital humanities.

As for what use these machine reading techniques have, I would say it’s still up in the air. The use of such technology is still in its early stages, so only a certain amount has been done with them. Most of what I’ve seen seems to be nice and interesting, but I think it is still too early to tell whether or not heavy scholastic use of these programs could be used. The Moretti piece we read seems to point in a convincing direction, but whether or not the majority of scholars find such reading methods beneficial and with advantageous purpose is yet to be known. I think that, if the right kind of ideas and programs come along, machine reading could serve a useful function scholastically. What will be even more interesting is whether any authors try to outwit this kind of reading in their works as the use of technology increases.

Reply
Allegra Mondillo
December 20, 2012 at 6:37 pm

It has become quite clear that machine-assisted reading is of great value to humanities scholarship. As it forces us to let go of the way we might read ordinarily, it is also sharpening our abilities to analyze and interpret. Machine reading produces multi-faceted or two-way thinkers while simultaneously offering a way into the text that the reader wouldn’t have otherwise. These tools offer a meaningful way to play, which could very well be the next best idea for a cirriculum in the middle of a rapidly-changing digital world. I still do firmly believe that machine reading and close reading must be used together in order to properly analyze a text. Cathy Davidson proposed collaboration in the classroom or workplace as the solution to attention blindness or selective attention, and the concept of machine reading is quite similar. Machine reading is the most effective when used in conjunction with traditional methods of reading. Computational analysis is very one-sided and rigid in its abilities, so it is the responsibility of the reader to use his or her own ideas and observations picked up from close reading practices. Based on my own experience, it was extremely difficult to “unlearn” the lessons I picked up from close reading. It took hours of experimentation before I finally discovered patterns or fascinating moments in the text. However, when I did, I found an overwhelming amount. Machine-assisted reading will undoubtedly enhance literary interpretation and humanities scholarship simply because it has the power to collect data and display patterns or images in seconds. Nevertheless, it must also be noted that such tools change and expand the reader’s mode of thinking—another valueable skill in the classroom. As long as we do not become dependent on the “machine” to do all of the work for us and as long as we can add the richness and depth of a close read to our machine-assisted interpretations, this new method of reading is quite beneficial to humanities scholarship.

Reply
Michela Marsh
December 20, 2012 at 7:29 pm

I think machine assisted reading is an interesting concept – I don’t know if it can be called a practice yet, since it seems to be a relatively rare/unusual method that is not mainstream and widely used. Machine assisted reading can help us as scholars of literature to better understand entire periods of writing; it can help us to understand language used, popular themes of the time, it can differentiate between authors and time periods and help to separate and identify differences in writing style and content. Machine assisted reading, as pointed out in one of our readings, is a great tool to help literary scholars reap all that they can from literature (the example given in our reading was, despite having read the 200 canonical books of Victorian literature, one really has no sense of the 15,000 some-odd books written during the period, unless one uses machine-assisted reading to “read” all of these books for their vocabulary, content, themes, etc.). So in this sense, I see great things that may come from machine assisted reading. I myself was very interest by Google’s tool that allowed one to compare multiple words over a given period of time. I used this tool to make certain points about the treatment and inclusion of women in literature between the 1700s and today, comparing keywords like “slut, whore, prostitute” (heightened usage in the 17-1800s (penny novel smut), the 1920s (suffrage), and then in the 1970s (rise of radical feminism)) and “he, she, him, her” (literature is masculine dominated throughout the time period).

But I also think machine assisted reading is a dangerous concept, semantically speaking. “Reading?” I wouldn’t call what is occurring thanks to these machine tools “reading.” It isn’t reading, except insofar as one reads what is presented on the page from an analyzed text. I can read Bartleby’s wordlist, but am I reading Bartleby? No. I am not discovering the intricacies of the text, the beauty of the language used, the meta story; machine assisted reading is relatively useful in a brief, surface, literal-content reading of a book, but not in the actual close reading of a book, not in how one understands the ‘true’ meaning of a text, not in identifying symbolism and other vital aspects of understanding what an author actually meant in his writing. Machine assisted reading – which I think is much more aptly named by ‘distant reading’ – is no replacement for readers – as opposed to scholars – of literature, and I hope that this never changes. I don’t know how it could replace a good afternoon with a book in one’s hands – if there’s one thing I think we’ve all learned and agreed upon in this course, it’s that our relationships with the books we read aren’t replaceable by or replicable through anything else, and I hope that the field of literature continues to act this way, as well….

One note, however, relating to Cynthia’s final point – The infographic you showed us of Danielewski’s book was AWESOME. Seeing an author that wrote a book (reading a book that an author wrote?) with machine assisted reading techniques in mind is veryyy fascinating. And I wonder, when Danielewski kept machine assisted reading in mind, DID he outwit this kind of reading?? Perhaps not in this circumstance, as this machine assisted reading was a beautiful act of symmetry – but I wonder, if he had prepared for MA reading in “House of Leaves” what tricks and troubles his readers might find…

Reply
Phillip Ma
December 20, 2012 at 10:09 pm

Machine-assisted reading works best when handling large sets of data or texts written with machine-assistance in mind, like Moretti’s charting of genres or the Only Revolutions infographic. Machine-oriented texts will only continue to grow, securing machine-assisted reading as a means of approaching texts. Given the wealth of text now available through archives, I am inclined to say that large-scale readings like Moretti’s will grow as well. This is the kind of reading I find most interesting and believe has the most area to grow, since it has been so far been limited by human capacity.
As tools available become more sophisticated and algorithms more comprehensive, the nuance and the extrapolation of traditional close reading will be feasible for the machine.
Machine-assisted reading should have a place in the curriculum simply because it expands the repertoire of methods at our disposal for approaching a text. Like any guided inquiry, machine-assistance must derive from human focus, with the resulting data always oriented towards a certain interpretation. This would be an integral point to be taught and helps to distinguish it from the aimless collection of data that sometimes occurs in science with the trial-and-error approach to experimentation.

Reply
Ashleigh Smith
December 20, 2012 at 10:20 pm

Machine assisted reading is a necessary tool for humanities scholarship, not because it is a superior form of literary analysis, but because it encouraged the integration of digital media in the humanities curriculum. As a literature major, I am challenged by the idea of adopting mechanical textual analysis unilaterally to every literary explication I write. However, I do believe that distant reading has a meaningful role within a humanities curriculum as a supplementary tool to close reading. During our machine reading assignment, it became apparent to me that the tools were providing me with insight into the text that I otherwise would not have been able to gather through my own analysis of the text. I was excited by the new connections the tools allowed me to discover, but I was unsure whether the process was worth the product.

During the process of the machine reading, and subsequent reevaluation of my primary thesis, I realized that the beneficial information I gathered from my distant reading was primarily a result of analyzing word frequencies. I believe this is the most relevant component of machine-assisted reading. Outside of the information I collected from the word frequencies, I found the tools to be cumbersome, challenging, time consuming, and not worth the effort for the results they produced. Perhaps this challenge was in part due to my rudimentary understanding of these tools and their practical applications.

It seems as though machine assisted reading has the potential to become an essential component of literary analysis, if it is taught as comprehensively as close reading. If students have a prolific understanding of the distant reading practices, they will be better prepared to apply them to literary studies, and the humanities more generally. Text-analysis provides students with an unique contextualization and connectivity of a text which can enhance and further develop academic papers, allowing for students to postulate upon the significance of their findings in a similar way to a close reading of a text. While I believe text-analysis has a place within the academic community, I don’t believe that visualizations would be able to have the same impact. The information that can be collected from the visualizations is highly relevant, however the visualizations may not have a place within academic writing in more than a supplementary role.

The academic humanities need to be willing to integrate various forms of digital media into their curriculum to keep up with the changing media ecology, machine assisted reading is simply one of the many digital mediums they should adopt.

Reply
Andrew
December 20, 2012 at 10:33 pm

Though I began this project assuming that text-analysis would ‘be’ to a book as a list of ingredients ‘is’ to a wedding cake, I was wrong. I guess a more apt and still confectionary analogy would be text analysis is like watching flour and eggs and frosting become a wedding cake. I was surprised at how mechanical some of my results were, that I seemed to be dealing with a close up, kind of slow motion sequencing of reading. It was the ultimate breath of fresh air to involve those folks who haven’t been discussing algorithmic criticism, ergotic literature and literary mapping & graphing for a term, and see how that unseen yet often invoked hypothetical public actually took to some of the methods we’ve discussed over the term. There were certain ironies that surfaced during my project– participants would get frustrated with the machine-assisted reading, claiming and believing they could outperform the programs (whatever that means…), but then would quickly turn and succumb to their cell phones to make a plan with a friend or tell their mom what time the flight leaves. Initially I suspected this to be a crash course in normalizing, give machine assisted reading time to grow, develop and normalize as a practice and these students would manipulate it like a text message– but then I reconsidered my thoughts. Perhaps my participants seeming aversion or reluctance to embrace machine assisted reading has to do with personal/human-kind pride. It’s a cliched and sappy word-trick, but maybe this safeguarding of the HUMANitites has to do with the way we use interoperation, empathy, critical speculation and observation to define and establish our human-ness. Reading is not a way of communicating with fictional characters. It’s not a way of locating them or defining them or ‘gaining’ as a kind of knowledge. In fact many of our most fiery debates this term had to do with what (if any) is the functional value of reading. Of course reading’s symbolic value is profound– but its much more difficult to empirically measure, state or even point to the functional value of reading. One might idealize, sensationalize, or moralize that value, but whatever explanation it may be should be muddy but intrinsically human. The suggestion or notion that machines may ‘invade’ the space that has for so long been considered inexplicably human may send a kind of primordial chill up the academies spine, but this view of M.A.R is a kind of hyperbolic war-cry, humanistic extremism if you will. The machine assisted reading I witnessed didn’t seem to corrupt or infest the students’ critical faculties as much as it did force each participant to make a sturdy and serious choice as to how and what they would read. In some ways, it demystified reading as a kind of unspeakable human exercise and instead barely scratched the surface of a long and potential exploration into the possible structures and orders of what has been branded as “anything goes hermeneutics.”

Reply