Reviews by Allison Parrish (AP) and Everest Pipkin (EP).
Airsun. The Value of Advertisements
AP. This is very effective--a great use of procedure and visual design in the service of parody (though why didn't you use Apple's San Francisco font instead of Helvetica?). But I'm not clear on the text generation procedure or how it relates to sentiment analysis...? (Is it just replacing words in pre-written copy with synonyms of different sentiment?)
EP. I like your tech-font layouts and your buzzword-filled copy, but I was actually pretty lost on what was happening here until I dug into the code. I thought perhaps these generated descriptions were being ranked by a sentiment analysis bot, a process that was actually more interesting to me than the real narrative here, of inserting 'positive' words into ad copy. It feels pretty choppy, and I am not sure if your insertions are clean enough to get away with it. I would encourage a more thoughtful approach to replacement, because it doesn't read well any more. The best part is definitely your design- well done on that front.
Breep. The Hands of Gutenberg
AP. A well-constructed, evocative project. The idea to augment the rhyming corpus by replacing words with other rhyming words is very clever! I'm left wondering about a handful (no pun intended) of creative decisions that feel under-motivated to me--why iambic pentameter in particular, and why the sonnet form in particular? What do these have to do with hands (or pictures of hands, or whatever other themes are a part of the project)?
EP. I loved this idea! Its a nice unifying structure that both explains what Project Gutenberg is like and lets us see just a few spare examples of text all to a theme. I think your scope is smart here, and the Iambic pentameter technique you used to force the poems is clever. If you were ever to re-write it, I would love to see a little more variation in your rhymes- wordnik and related databases often have a 'sounds-like' or 'slant-rhymes' category that may make this feel a little more organic. Your imagery works, but doesn't feel up to the quality of your poems. I'd love to see what you could do with a database of historical portrait painting and machine vision looking for hands, for instance. Maybe beyond the scope of this project, but an idea for the future. A note (that is not at all effecting your score, because I am not sure if you put the headings together)- it seems like your description is inaccurately copy and pasted from another piece.
Casher. Lyric Poetry
AP. This produces interesting and amusing output that succeeds in demonstrating the somewhat uniform semantics of song lyrics. I'm not sure that I agree with the author that "lyric poetry" is a name that can only be implied ironically to these artifacts--it seems to me that there was plenty of intent behind them (235 lines of Java code, at the very least). I feel that there was a missed opportunity here to look into the structure of songs, and how meter and rhyme actually work in them (it's not just rhyming couplets of arbitrary lengths!). But the QuickDraw guitars are a nice touch!
EP. They do mostly work together, you are right- but it is often obvious that they don't make a song. I suppose I wish there had been more of substance here than randomly grabbing lines that rhyme. For me, this project would have been improved by subject throughlines- pick a main concept, get a bunch of related words, and pull only lines that match those related words. Even establishing a chorus that is returned to a few times over the course of the lyrics could have done this a world of good. The quickdraw dataset is an amazing resource and I am glad that it is being used, but similarly- guitars seem like a pretty basic pull. Your songs have titles- could you have pulled a 'World' (The World we Live In), or 'Shelter', or a 'Whistle' from the quick-draw set procedurally? Maybe, and maybe it'd give a bit of depth here. You're not wrong that lyrics are pretty universal and follow universal themes, but a vapid data set doesn't make for an interesting project without some intervention.
Chaine. Recipes for the Mad...
AP. This is great--the verbs and phrasings in the grammar that produce the instruction portion of the recipies are delightful, producing something that feels like a cross between Alton Brown and Yoko Ono. As standalone artifacts, they're effective, but in aggregate the patterns of the generator become very apparent and it becomes more difficult to pay attention--the project could use a bit more variation in structure and/or a bit more cohesion.
EP. There is a joy in the utter goofiness of generated text, but I think this might go a little too far. Part of humor is knowing when to stop, and I think a lot of these recipes would be funnier if more on the edge of believable. The titles alone could do a lot of the heavy lifting here- Crispy Sour Braid is incredibly evocative, but the grab-bag instructions don't actually add a lot for me. I'd scale back a little next time. Still, pretty clean layout and varied enough source material it doesn't repeat. Good for you on that. I actually liked your illustrations a good deal- they are universally unappetizing but do, somehow, gesture at food or a plate.
Chewie. High Stakes
AP. This is a thoughtful, sophisticated study of narrative and genre that doesn't quite come together--I get the underlying motivations, but in my opinion working with Calvin and Hobbes plot summaries instead of the strips themselves undermines the project's ability to make the point that the author wants to make, since the plot summaries retain so little of what makes C&H strips interesting structurally. It's difficult to recover any of that context from the generated text. I don't understand what motivates the visual design of the piece (large photos with superimposed text) and the text is kind of difficult to read. Would love to see further iterations on this idea, though!
EP. I won't lie- I lost the plot of this one a little. Obviously, its about poker- but the narratives felt, well, out there. Reading the blog post about Calvin and Hobbes gave me some context (both for the source material and the endpages) but didn't really explain the why. I'm not sure if the mashup between sources brought much to the table for either. The Calvin and Hobbes corpus you found is a pretty amazing resource- I'd suggest really thinking about where it could be poetic, fun, or moving in a more fundamental way.
Chromsan. The Extended Encyclopedia of Philosophy
AP. This is a great success! Makes use of multiple techniques (scraping, cut-ups, nlproc) to fashion interesting and insightful artifacts. The clean design aesthetic suits the text. I do agree with the author that the descriptions are too long--or if not too long, then lacking in the kind of coherence that would lend itself to verisimilitude at the current text length.
EP. I think this broadly works- the descriptions and titles are actually quite believable, and are unparsable in that specifically philosophical way. Occasionally they delve into real humor. I think your attention to priming the source material shows- they feel clean in a way that laziness doesn't allow. No points for a shiny layout or extras, but the text is really pretty good- well done.
Dinkolas. Bioinvasive Dingus
AP. This is really compelling--despite the author's assertion that the text is "difficult to read... impossible to parse," I actually found this to be among the most visually compelling and "readable" of the chapters. The nonsense words serve to draw attention into the surface form of the text--there's a great rhythm and feel to it that is enhanced by face-circles moving your eyes down the page. It's this attention to materiality that makes it especially effective as satire, I think!
EP. I'm not personally the biggest fan of this style of nonsense generation / cutups, but if you're gonna do it, this is the way. This has a very Burroughs-esque dystopian vibe that moves in and out of utter mishmash, as well as internet conspiracy theory. I think your strength and anchor here is your imagery, which lend an illustrative edge to something that otherwise is not rewarding to read. Perhaps there is extra life to be had in this while read aloud - a good poet struggling with some of these words could be a thing of beauty. I recommend paragraph breaks and a closer eye to the end, which stops mid-sentence.
Harsh. Build Limericks Not Walls
AP. Using pre-written lines to guarantee a bit of coherence (both in sound and meaning) is a great move here, I think, but it comes at the cost of variation--it doesn't feel like the expressive range of this particular generator is especially large. The Twitter logo "particle system" is an interesting idea, but in general I think the visual design could have been a bit more intentional (e.g., font choices, making sure that limerick lines don't break across physical lines, etc)
EP. I'm always happy to see political work, even easy/goofy shots like this. The first one really tickled me. And then the second one was the first one. And the third one was also kind of the first one. And so on. I've got nothing against hand-writing some external structure, but it would have been pretty easy to pull some various rhyme schemes in here for a little variation. I don't think you can use the same joke 4 times, even if it does show the breadth of generation. That is what multiple runs of the book is for. If you'd written a bunch more of these or programmed in more variation, this score would have been a lot higher! I think its really pretty funny, just only the once.
Lass. MEDICAL FACTS
AP. This is evocative and bold, with very thoughtfully chosen source texts. I'm not sure that I understand the relationship between the text of the project and the visual design (which isn't to say that I dislike the design, just that it's not clear to me how it's supposed to be supporting/complicating the text). It's not clear from the documentation whether or not the creation of the illustrations is included in the procedure that generates the book chapter?
EP. I liked a lot of this, particularly the corrupted models of the human bodies that form the background. Did you know that you can do this procedurally? Try opening a model in a text editor and editing the points, its a lot of fun. I did a whole generated book project about that last year, actually... Anyway, something to try if you ever get sick of blender. The text itself feels only fine to me, but it is so spare and bold that it doesn't have to be deeply readable. It feels like a series of posters or artworks, and thats good with me.
Nannon. REDDIT BIBLE?
AP. This is a great use of sentence embedding, and it produces some really elegant and poignant juxtapositions. I'm not sure if I really understand why the complementary source text is questions from the Bible in particular--there are a lot of questions in the Bible, but very few of them are asking for advice, so there isn't a clear mapping between the two corpora. Not that there needs to be a "clear" mapping, but I was left wondering what the project would be like if the selection of questions was a bit more intentional. The visual layout (with questions flipped and opposite each other on the page) is inspired, though I don't understand the choice of the pixely typeface for the Bible questions...?
EP. A cute idea I was expecting to have more issues with than I did. I think it broadly works for you, despite how hard reddit data often is to work with (there are so many slurs/etc across a lot of the boards that I have had trouble with it in the past.) Using vector-space difference between the two sources points the humor-finger at machine vision more than advice-seekers or religion, which is a generous gesture. Because I didn't get my hands on a physical book, the rotated page format- while clever- actually made it quote difficult to read. This was made worse by the 8 bit-aesthetic font which is doubly hard to parse upside down. This may be something to think about in the future, especially if this is primarily going to be interacted with as a pdf. Part of this could be fixed simply be switching up which of your sources is upside down every once in a while- being able to read a few bible examples without straining would likely prove the point you are trying to make.
Nerual. A-Z, Or Something Like That
AP. The visual design accomplishes the goals that the author set out to accomplish (bold, angsty)! I'm a bit less sure of how well the artifacts produced by this program operate as satire, though that's maybe just because I'm not familiar with the genre (2012 Facebook posts by angsty teenagers). Given the relative straightforwardness of the textual generation procedure, I might expect a bit more polish in the particulars (e.g., selecting appropriate parts of speech, not producing duplicates, etc.).
EP. This is spot-on hot topic teen (is that still a thing? idk i'm old), and that alone is pretty admirable. That said, it is spare so it really needs to be perfect. And there are moments where this simply breaks down- 'C is for CONCIOU', 'they are ___'s' that repeat, 'They are an.', etc. For such a simple gesture, I wish you'd nailed it. Clean up your code or source, add in a bunch more content so you get less repeats, and I think it could really work. As is, it feels a little broken.
Nixel. Fake Love
AP. I think the concept behind this is solid, and the visual design is perfect. From a procedural standpoint, though, it's just a search-and-replace operation on a particular token, which feels like a missed opportunity.
EP. Using the compelling imagery of Kpop videos is clever, and your layouts are slick. The combination of images and captions are slightly unsettling, which is not a bad thing- they have an ominous undertone. That said, I'm not sure if applying unicode combining characters to lyrics really counts as generative literature- despite being put together with code. If you were interested in ideas of corrupting text, I feel like there might be some more aggressive ways to go about eroding sentence structure and language than diacritical marks. !00% for design and style, but I feel like it wasn't taken nearly far enough.
Ocannoli. John Mulaney's Comedy Hour
AP. The method of producing rhymes with RiTa and string replacement is clever, and the output is at times amusing. The author shows awareness of some of the shortcomings of the project in their documentation, which is a good thing. What I'm missing from this project is an awareness of the specifics of Dr. Seuss' distinctive style (Seuss used of rhyme, alliteration, and assonance goes far beyond end rhymes) and an answer to the question--why make a "dirty" version of Dr. Seuss? What are you hoping to show about Dr. Seuss (or about dirtiness)? (Also "Seuss" is misspelled throughout the project and documentation! I'm not ordinarily a stickler for spelling, but in a project that is supposed to be specifically about the writing of a particular author, it seems important to correctly write the name of that author...)
EP. In can see where you're headed with this, and sorting Markov text to fit with a rhyme-scheme is a good tried and true method. I didn't get a lot out of this personally, however. Maybe it is because neither source is all that compelling to me to start, or maybe it is because they don't add a ton to one another (Seuss is brilliant as a wordsound person- does comedy need that? Does Seuss need contemporary humor? Idk, maybe not.) Maybe something could be done with bringing in another source of imagery to stand against these stanzas as captions, as opposed to reproducing Seuss again. But also maybe this one just didn't work out so well- I think they aren't a good match.
Paukparl. Generated Self-Help Books
AP. This is a funny and enjoyable text generator! (I really like "Carelessly cool" as a book review, ha!) As satire, though, I think there are some missed opportunities to dig a bit more into how the rhetoric of the covers of these books actually work. As it stands, the project is fairly minimal.
EP. Oh Daisy Keneally, what a prolific author... - I like these. They do what they say and have a kind of real dark humor to them. I think you did have a chance to do something a little more impressive with your layouts than simply centering and generating a bar code. That said, they do just enough that you know what they are. I could wish for a little more variation, but all and all I think you did just enough to get away with it.
Rigatoni. Punk Rock Album Generator
AP. This entire project expresses a punk rock aesthetic--the lyrics, of course, but also the minimalist dense abrasiveness of the visual design (bold helvetica! full justification!), and the lack of regard for whether or not the notes actually create "music." It's just a really energetic visual artifact that demonstrates sharp design instincts. Plus the songs actually seem good?? (though there's a bit too much repetition maybe when you start reading closely)
EP. I think in general you just get away with unplayable music, by dint of good lyrics and clever placing. It is certainly punk. I'd love to see what you did with actually playable scores, which you mention in your intro. You might be interested in some of Emma Winson's work with generated scores, which are 'playable' but in an abstract way. She generates shapes, patterns, and word-descriptions which are then interpreted into music. Just another way to think about this.
Sapeck. Antisemitic Absurdities
AP. The rhetorical design of this project is strong (with the facing black/white pages and opposing right/left justification), as is the decision to provide a structural backbone to the generative procedure (the progression of generality, logic, etc. in the first word). In the end, I'm not sure what the piece is arguing, as many of the ethnic groups substituted for "Jew" in the "absurd" pages are themselves subject to violent oppression, stereotyping, etc. Is the point to draw attention to formulations of xenophobia that are specific to antisemitism, or to show that the language of xenophobia potentially shares many characteristics, regardless of the group it's directed at? (or a superposition of the two?) It's not necessary to make a clear and specific "argument," but I'm left wondering whether this lack of clarity was unintentional.
EP. Oof- this one was hard for me to read. I had mixed feelings about the project conceit, but know what you were trying to do. I think the trouble for me (also a Jewish person) is that although we are at times a marginalized group, we are not subject to nearly as much daily and constant racism as many of the other groups scraped for your list. So, although yes 'All Australian are heretics' might come across to a western-centric, white, worldview as goofy and baseless (the point you were trying to make) American Roma, or Libyans, or Mexicans, or Iroquois (other groups you scraped) face this type of thing constantly. It is difficult to read and might not carry the point through. I think technically it is proficient and I applaud you for making political work, I just wonder if this was the best way to approach this issue. I hope that as you continue to make political work, especially about ideological difference and hate speech, you can tune this sense to be generous and careful.
AP. I know that the font was chosen for the sake of verisimilitude to the idea of the written diary, but it is kinda impossible to read. Font choice notwithstanding, the generated text is sometimes very evocative! The documentation doesn't explain the generation method, but it seems like a Markov chain on one or two source texts, plus random number generation for the lat/long, speed, etc.?
EP. You went above and beyond to deliver a thing of substance- well done. The various moving parts to this project (Lat/Long pairs, weather descriptors, and the actual text, the imagery) do a lot to set it apart. Occasional moments of grammatical failure (a period followed by a comma, for example, or phrases that cut-off mid sentence) are moments of breakage that I'm sure you can figure out how to fix. When you are moving forward with this text and idea, I hope you think about histories of colonization and exploration, 'the natives' (a phrase that comes up in this text regularly), and where you as an artist stand on this narratives. I don't think you can have a neutral stance when working with material like this, and your introductory text (or lack thereof) will influence the read.
Shuann. A Guide to Absurd Movies
AP. Some of the language produced from this is very evocative ("The Tree explains the monkeys serve the Wicked Witch of the other; the T-rex, slashing the dinosaur's skull with its talons")--the work of massaging the output of the Markov chain definitely paid off. The visual design is minimal but effective in organizing each page. Some of the language in the reviews is a little repetitive; they're less effective overall than the plot summaries in my opinion.
EP. This is an ambitious project that must have taken a good deal of scraping and cleaning data. It mostly works; I actually don't know how you'd fix this, but the highly disparate reviews (and changing options sometimes within the reviews) leave this feeling pretty disjointed. Maybe you could bring in some sort of last-step sentiment analysis that sorts things into good and bad camps... Something for the future. A quick fix; Reviewer1, 2, and 3 should really have generated names, and the movies would be served well by titles. Everyone loves a goofy name generator, and they'd be easy compared to the rest of this.
AP. Produces some amusing juxtapositions and is rooted in a good formal analysis of the underlying genre (the "Placing" essay). As noted in the documentation, though, there's nothing in the work that draws attention or invites closer reading; it feels a bit too "random" to me.
EP. I suppose you can't argue with results. I was ready to write this off, as plugging in some essays to a Markov chain generator is generally not particularly interesting, nor technically challenging, nor compelling. That said, the combination of images and captions you've served are really quite interesting and downright poetic at times. Your advantage here, of course, is your source material- concise writing that can apply equally to the many beautiful images available to you. Its not cheating to pick a good corpus. Well done. The Markov chain begins to show by the end of the chapter, as phrases have repeated multiple times. As always with Markov chains, I believe that they can be truly interesting if constrained by other variables- have it remove full chunks of text every time it uses them, have it increase its level of choppiness over the text, etc. I do hope you can explore more next time.
AP. This is strong, well-crafted work--I think the tone/style of the generated text is perfect (laconic, utilitarian in a faux science-y way). Ultimately, though, I think the generator lacks expressive range--all of the outputs look more or less the same, and there isn't enough visible variation to draw you into a sustained reading of the images or the text.
EP. All in all, I think you did a good job- you found a prompt that let you iterate in small ways on a theme (limiting intense technical overhead). Because taxonomic classification is always a little repetitive (generated or not) it doesn't end up sounding too stale, despite being limited in scope. Good generative projects are more often than not their corpus or their scope- congrats on picking an effective one. The illustrations are simple but very effective. They have a flat, clean look that complements the spareness of the text. My one suggestion is that rather than remix already extant planetary body names, you instead look at the themes on which they are named- mythology, etc- and pull, say, lists of thousands of mythological gods. This may give you a little more space to work in with your naming conventions. I also think you could use an additional few adjectives in your 'puffy' slot.
Yalbert. Gender Bended Classics
AP. A very thoughtful study--the idea of using Levenshtein distance to find the most similar name is a great idea (leading to wonderful transformations like "Marc Poppins"). The implementation is a bit of a let down--it's not able to cope with different parts of speech ("they saw their Father coming out of the drawing room with a visitor following his") and gets tripped up on some punctuation. I'm not sure if the abrupt method of "excerpting" (cutting off mid-word) is intentional or not?
EP. I like the conceit (and it could certainly scale up, which is always a plus) but the implementation is messier than I had hoped it would be with such a targeted concept. In your intro, you mentioned needing some sort of machine learning to do this well, but your examples are all well-known books. I feel like you could probably scrape a fan wiki that lists the gender of your characters, which would solve the 'Harry Potter' problem of Hermione, Hagrid, or Dumbledore not being in the names database you were referencing. I also wish you had started the excerpts at a chapter, paragraph, or even sentence break. I feel that the full-tilt launch into the action (mid-word!) removes some of the context of the story, which reduces the power of the thing you are trying to do. That all said, it does mostly work- I think you just need to be tidier next time.
Yuvian. Fortune Cookies
AP. I appreciate the work that went into researching the onomastics of this kind of restaurant, and for me the restaurant name generator is the most successful part of the piece! Unfortunately, the code to generate fortune cookie and take-out box images doesn't produce artifacts with noticeable variation (I didn't even notice that they were randomly generated until I read the blog post...) and the visual design of the output doesn't read as "advertisement" to me.
EP. This is a smart, scaled-down idea that I wish had been taken just a little bit further. I agree with you that image-caption relationships can do a lot of the heavy lifting in generated text, and your breakdown of a flier into small component parts- name of restaurant, phone number, fortune, lucky numbers, image- is clever. The fortunes are succinct but all quite different from one another. That said, the actual image generation leaves a lot to be desired. I'm not sure that making a drawing of takeout and pasting two phrases on it in slightly variable places deserves a caption. Did you consider making a few fortune cookies in various states of being eaten/broken that could be placed around the box? Did you consider placing the text actually on a strip of paper? Or doing a little more design to your layout to give it that red-ink printed vibe of a takeout fortune? Of course this would be a little more work- but not actually appreciably more. I think it would make this shine in ways it does not right now.