top | item 7223969

Hemingway makes your writing bold and clear

1040 points| jmngomes | 12 years ago |hemingwayapp.com

Great for testing your copy

327 comments

order
[+] jawns|12 years ago|reply
You know what's fun? Pasting in text from Ernest Hemingway and seeing what he did wrong.

But seriously, this is a nice, simple way to point out some general rules of thumb for improving writing, although I would love for it to be less proscriptive. Not every long sentence is a bad sentence, not every passive-voice sentence is a bad sentence, and not every adverb is a bad adverb.

Oh, and by the way, the copy editor in me can't help but notice that an app that's intended to help you improve your writing tells you to "Aim for 2 or less" adverbs, rather than "Aim for 2 or fewer."

[+] pmichaud|12 years ago|reply
The thing about pedantic snark (less/fewer), is that it loses its teeth when you're wrong.

Even if you go in for prescriptive grammar, less and fewer were never strictly divided. The distinction actually came to us through one Mr. Baker's expressed preference:

http://en.wikipedia.org/wiki/Fewer_vs._less#Historical_usage

[+] mkrecny|12 years ago|reply
I'd love there to be a proliferation of plugins for different writing styles. Hemmingway, Austin, Asimov, Wilde etc. Not so interested in "improving writing", more interested in achieving different aesthetics for different goals.
[+] hissworks|12 years ago|reply
>> Aim for 2 or fewer.

This also violates current usage conventions for handling numbers in text.[1] Figures are only meant to be used in greater numbers (a rule of thumb is anything over 30, but I think I read that in a David Foster Wallace text, fwiw) with more complex verbal / textual representations - that is, not "two".

[1] https://owl.english.purdue.edu/owl/resource/593/01/

[+] gabemart|12 years ago|reply
That is neat. I pasted in the first part of Big Two-Hearted River [1] and it scored a little "worse" than the first part of the novel I'm working on, suggesting it's not a perfect metric for how "Hemingway-like" a text is. I've always thought that the cultural meme of Hemingway's terseness is not a very good representation of his actual writing style. In the past when I've been prompted to try parodying Hemmngway [2], the result usually tends to a style that's a little more verbose.

[1] http://xroads.virginia.edu/~DRBR/hem_river.html

[2] http://www.reddit.com/r/WritingPrompts/comments/1t9571/wp_so...

[+] ojbyrne|12 years ago|reply
Not to mention that the Chicago Manual of Style recommends spelling out the number "two."
[+] deeviant|12 years ago|reply
It's not like there some magic linguistic genie that they are going to for their analysis. A programmatic engine to determine which long sentence or use of passive-voice are bad or not is not trivial, if even possible with their resources. Just identifying the hopefully few uses of complicated sentences and passive-voice would allow one to quickly sort through them and make a decision on whether the usage was proper or not.

They did a great job, it's a nice app.

[+] acqq|12 years ago|reply
Exactly. Strunk and White actually devised their "rules" without the exact analysis of the real texts. Likewise, I can imagine HemingwayApp giving the real Hemingway bad notes.
[+] jimmcslim|12 years ago|reply
> You know what's fun? Pasting in text from Ernest Hemingway and seeing what he did wrong.

I was thinking about giving some Joyce a pass through this...

[+] toxik|12 years ago|reply
That's funny, because I tried your comment in Hemingway and it came out with a much better score after I erased every letter of it.
[+] midgetjones|12 years ago|reply
I agree with your second paragraph, although this is following Hemingway's specific rules, so it's good to be strong with them. And of course, you have to know the rules before you can break them.

Aside from Hemingway, it's fun to paste in other classic literature and see what the tool makes of it. It hated Lolita.

[+] andrewliebchen|12 years ago|reply
You're right. Though we can tell you didn't compose your comment in Hemingway. It characterizes one of your sentences as "hard to read," and another as "very hard to read."
[+] wyclif|12 years ago|reply
It should say "aim for two or fewer", since it's better style to spell out any number < 10.
[+] coldtea|12 years ago|reply
Or "Aim for two or less". It's not a good style to write quantities below ten with digits.
[+] bgilroy26|12 years ago|reply
Writing a lot of short sentences develops your sense for when a long one is appropriate.
[+] enduser|12 years ago|reply
Your last sentence was "very hard to read"
[+] acqq|12 years ago|reply
I believe the logic behind HemingwayApp is misguided:

Hemingway the writer actually wrote long sentences and they were actually important in his writing.

Passive is also important in good writing.

You can't use machine metrics to force "good writing" you can only enforce mediocrity and the following some random rules "because the rules have to be followed."

Likewise, I as a writer of the software would absolutely hate to run some program to tell me "this function has more than 10 lines" or whatever. If I wrote 500 lines function it doesn't mean it shouldn't be that long: there are examples where exactly such functions are still necessary and good. Such automatic evaluations are for managers who probably don't understand what they enforce. Pointy-haired bosses, if you will.

So I see HemingwayApp as the pointy-haired-editor app.

(Edit: Improving the text based on the human input, thanks Agathos!)

[+] buzzcut|12 years ago|reply
This is built on so many bad assumptions. At best the "rules" it's trying to enforce are training-wheel rules, the sorts of rules given to novice writers to help them avoid flabby, purple writing.

But the assumption that short sentences are better than long sentences, or that simple sentences are better than complex sentences is just wrong. There are all kinds of reasons why you might use one type of sentence over the other or vary them for effect. You might be concerned about rhythm, or you might be attempting to establish a certain tone, distance, closeness, formality, or lack of.

We have this weird cultural obsession with the clarity, brevity, and simpleness of writing. Jacques Barzun even wrote a writing manual called Simple and Direct, as if these are the only virtues to be found in writing.

But I think you want as many tools as possible to achieve the effects you want. There is a huge rich tradition here, that we've largely lost, a tradition that teaches about hypotactic and paratactic sentences, that teaches about periodic and loose sentences, that teaches how to make left and right branching sentences, that teaches subordination, that teaches rhetorical devices, and that advocates (at times) longer, more complex sentences for richer and denser writing.

Thankfully there are a number of books out (some of them) recently that seem to be fighting back against the austerity view of writing.

They include, if you're interested: - Brooks Landon, Building Great Sentences - Stanley Fish, How to Write a Sentence - Virginia Tufte, Artful Sentences, Syntax as Style - Richard Lanham, Analyzing Prose

I'd just add, there is nothing wrong with being simple and clear. There is nothing wrong with cutting out needless or weak adverbs. But there is something wrong with worshiping the austerity style as, at all times, the best and the only way to go. There are lots and lots of reasons and occasions to deviate from it, but the style orthodoxy these days is the one assumed by that (admittedly cool) website.

[+] bo1024|12 years ago|reply
I thought it was ironic that the first two sentences were very well-written, but highlighted. As if to show an example where the app would be useful. The sentences are excellent and don't need changing.
[+] jonnathanson|12 years ago|reply
I'm going to love using this. I write for a living, so I write a maddening volume of output per week. While I don't absolve myself of the need to edit everything, I'm working against the law of large numbers. Some stupid errors, or bad stylistic habits, are going to slip through the net every week.

I've been jonesing for a real-time style editor for years. Autocorrect is fine and dandy (and often wrong, but that's another story). But most autocorrect systems limit themselves to spelling and grammar. Hemingway selects for readability. That's very cool and very useful.

That said, I'm probably not going to copy & paste everything I write into the Hemingway editing environment. I'd love plug-ins and APIs for Word, Google Docs, etc. If you make these, I will use them, and I will bug the living shit out of every writer I know to do the same.

[+] nswanberg|12 years ago|reply
Paul Graham's writing seems simple and direct to me, so I wondered how the website would treat one of his essays. Here are the suggestions from the third paragraph of http://www.paulgraham.com/essay.html:

"The most obvious difference between real essays and the things one has to write in school is that real essays are not exclusively (only) about English literature. (Sentence hard to read) Certainly (Adverb) schools should teach students how to write. But due to a series of historical accidents the teaching of writing has gotten mixed together with the study of literature. And so all over the country students are writing not about how a baseball team with a small budget might compete with the Yankees, or the role of color in fashion, or what constitutes (Forms, makes up) a good dessert, but about symbolism in Dickens. (Sentence very hard to read).

This is given a "readability" score of grade 14, which I suppose means it can only be deciphered by college sophomores or above.

I wondered how it would read after being rewritten to achieve a perfect score in the site, so I took a stab at it:

"In school students write essays about English literature. But real essays can be about many more things. Schools should teach students how to write. But due to a series of historical accidents the teaching of writing has gotten mixed together with the study of literature. All over the country students are not writing about how a baseball team with a small budget might compete with the Yankees. They are not writing about the role of color in fashion. They are not writing about what makes a good dessert. They are writing about symbolism in Dickens."

The result brings me straight back to my days of taking standardized tests, where the test had a snippet of some essay, and was followed by questions on the topic. There was information in those snippets, but very little tone. It could be a bad attempt at my part, but while the information remains in my version, the tone is gone--I can no longer smell the air of Cambridge in that writing.

[+] e12e|12 years ago|reply
Your version reminds me of "easy reader" editions of normal text -- or of an essay written by someone that's just learned how to write (a 6-7 year old?). It's fascinating how a few changes can change the "voice" of even such a short paragraph.

Anyway, great illustration both on how tweaking text can create contrasting changes, and how following "best practices" can be a bad thing. I suppose the original paragraph could actually be (slightly) improved by following some of the suggestions from the site, though. There's always room for improvement in any text.

[+] hsshah|12 years ago|reply
Interestingly, when I pasted the entire article, it gave me the score of Grade 8. Complex sentences can still be interspersed between simple ones and get a good readability score.
[+] TheSOB888|12 years ago|reply
These tiny sentences (in your version, and in Hemingway's stories) remind me of little children's books (and the alt-comedy storyline from the Martian steampunk shoot-'em-up Jamestown). IMO they appear a little condescending themselves.
[+] JasonFruit|12 years ago|reply
I don't think Hemingway could breathe at Cambridge.
[+] grandalf|12 years ago|reply
Paul is not writing for a 6th grade audience, so it's reasonable that it would not be written at a 6th grade level.
[+] normloman|12 years ago|reply
Professional writer here.

Writing well takes years of practice. If you already write well, you won't need this program. You'll know the rules and the right times to break them.

But if you can't put in the time and effort to become a great writer, just using this program can improve your writing a lot.

[+] pistle|12 years ago|reply
For presenting utilize as a wasteful term, I want to tearfully hug everyone involved in this.

Please kill 'utilize.' We should reach out to stakeholders and incentivize the sunsetting of the leveraging of the word 'utilize' from all slide decks.

Slide decks - the (not) new version of the tri-fold foam presentation board. It's the clear binder of our age.

[+] adam_b_long|12 years ago|reply
Hey guys, my name is Adam Long and my brother and I created Hemingway a few months ago!

Loving the comments here. As many of you pointed out, rules are meant to be broken. Our goal was to fix a simple problem: when you're looking at your own writing for too long, you start missing the simple, obvious errors.

You can follow me on Twitter @Adam_B_Long if you're interested in chatting about Hemingway with me.

[+] vkb|12 years ago|reply
This is a very beautiful and logical interface, but it's the wrong approach, because it's a very developer-centric approach to writing.

The problem with writing is you can't loop through it and find whether each sentence passes or throws an exception. A written work needs to be evaluated as a cohesive whole. That's what "bold and clear" writing means to me: a written piece of work that stands on its own and says what it means.

Computers are not smart enough yet to understand why "Lolita, light of my life, fire of my loins. My sin, my soul. Lo-lee-ta: the tip of the tongue taking a trip of three steps down the palate to tap, at three, on the teeth. Lo. Lee. Ta." is a complete, perfect paragraph. It doesn't have a verb and it looks like Lola is misspelled multiple times, so it doesn't pass the subset of grammar rules set up in the backend. But the meaning, the essence of the paragraph is clear.

Writing may someday be able to be governed by algorithms, but not yet. I ran the second paragraph of David Copperfield through Hemingway [1], and it gave me too many adverbs, a misspelling of the British neighbourhood, and the use of passive voice. This is understandable, as Charles Dickens was a verbose writer who got paid by the word. And yet, it doesn't detract from the fact that he is one of the most-loved in the English cannon.

We can't measure good literature yet, because there is no straightforward formula, and although this is an interesting attempt, it can't teach good writing better than a human.

For a better, and still technical, approach to understanding how and why sentences and paragraphs work with us or against us, it's better to read Strunk and White, and even better to read "How Fiction Works" by James Wood.

If there is a way to incorporate at least those two books into conditional statements, I would be excited to see it.

[1] http://imgur.com/k9hsHfj

[+] Duhck|12 years ago|reply
I love this, and I love your test for the desktop version.

One suggestion, make the price a slider from $0->$100 and instead of asking "Would you pay $5 for a desktop version of Hemingway? It would add the ability to save and open text files." ask "Please suggest a price for the desktop version"

This will give you a better idea of the true value of the application to people without being suggestive.

Awesome idea and implementation!

[+] jaysonelliot|12 years ago|reply
Also a big fan of the "desktop version" test, because it is triggered by the action of a user trying to get the desktop version as opposed to thinking they are about to answer a poll.

To that end, I wouldn't trust the data you would get if you asked people to state what they would be willing to pay. Self-reported data out of context is notoriously unreliable. Instead, really put it to the test. Tell people that you'll develop a desktop version if enough pledges are received, and ask people to make a binding pledge in advance. Give them a discount for jumping in early, of course, but use the average pledge as a guideline for final pricing.

If the median pledge amount is, say, $10, then you could probably assume that people would be willing to pay $15 once the product was actually available for download. (Those are made-up numbers, but you get the idea.)

[+] munificent|12 years ago|reply
I think people are reading a bit too much into the name, and into the feedback the app gives. I don't think it's "get rid of all the pastels and your writing will be like Heminway's", nor do I even think "get rid of all the warnings" is what it's trying to do.

It is useful for a writer to throw some text at it and see what you can learn. More feedback is almost always better for writers. The trick, is always, is having the judgement to incorporate intelligently.

For example, I stumbled onto a book about procedural content generation in games[1]. As a writer, game programmer, and dedicated fan of roguelikes, if this book were any farther up my alley, it would be banging against the back fence.

But, ugh, when I tried to read it, I just gave up after a few paragraphs. It's not gibberish, but it's almost physically painful to wring the actual information out of it.

And, indeed, when I throw some of those paragraphs at this app, I see:

    Paragraphs: 1
    Sentences: 31
    Words: 833
    Characters: 4196
    11 of 31 sentences are hard to read.
    11 of 31 sentences are very hard to read.
    10 adverbs. Aim for 0 or less.
    10 words or phrases can be simpler.
    13 uses of passive voice. Aim for 6 or less.
If the authors took a bit of advice from this app, they'd end up with a better book. That sounds like a win to me.

[1]: http://pcgbook.com/

[+] agentultra|12 years ago|reply
For the hackers, I use write-good-mode in emacs to catch passive voice (which is based on some simple shell scripts[0]). I've heard good things about diction-mode, grammar-mode and artbollocks-mode. And of course flyspell.

I find emacs very pleasing for writing text. I also use org-mode and it's LaTeX exporter extensively for publishing.

Now if only it could integrate with text-fields in my browser...

[0] http://matt.might.net/articles/shell-scripts-for-passive-voi...

[+] Flenser|12 years ago|reply
I use org-mode for a to-do list and reference/archive. How can it help with publishing?
[+] lhnz|12 years ago|reply
Well, this is great.

But it's missing something.

If I'm to learn how to write clearer, I will need to use this more often.

Could they create an API and a chrome addon?

Ubiquity is the killer feature of any communication tool.

[+] adam_b_long|12 years ago|reply
We're definitely planning on some form of Chrome add-on. That would be awesome. Thanks for using Hemingway!
[+] splitbrain|12 years ago|reply
The source of this would be useful to integrate such features in other apps (like editors). An API might do as well at least for online tools like blogs. As a standalone site it's too much hassle to integrate it into your daily workflow I think.
[+] jaimebuelta|12 years ago|reply
I like the idea of the app, but I'm not totally sure this kind of "review" will be very useful.

The problem is that, in some cases, you need complex sentences , passive voice or adverbs. And that means that a perfectly fine article won't be pristine. I had a similar problem when facing syntax correctors that show a lot of warnings. Yes, they help you make less mistakes, but they also give falso positives, which can be distracting. I want to clean up and get to zero errors, after all.

So, this can reduce your writing to be "too conformant".

Man, writing is hard :_(

[+] chadwickthebold|12 years ago|reply
I thought the same thing on first glance. However, maybe the use case for this is rather to filter entire blocks of solid color text. One complex (red) sentence every now and then is perfectly fine, but I would rather see a whole paragraph of them, say 4 or so in a row, as a problem.
[+] aneisf|12 years ago|reply
I'm reminded of a story[1] I heard on NPR the other day. Researchers have drawn correlations between writing style and the eventual onset of Alzheimer's. Apparently nuns who had a habit of writing verbose, idea-dense sentences were less likely to develop Alzheimer's later on.

[1]: http://www.npr.org/templates/story/story.php?storyId=1272118...

[+] CalRobert|12 years ago|reply
This saddens me. I appreciate clear writing as much as anyone, but are we not denuding our language if we attempt to describe everything using short sentences and a small lexicon? I hate reading James Fenimore Cooper as much as the next person (including, notably, Twain), but surely there is a place for complex ideas expressed with a rich vocabulary and nuanced structure.
[+] CalRobert|12 years ago|reply
And no adverbs?!???? That's just ludicrous. I modified an adjective with one right there. I regret nothing!
[+] k-mcgrady|12 years ago|reply
I like this. I pasted in some text from a blog post I'm working on. All the edits it suggested made the post much better. I was worried that through the suggestions it might take the personality out of a persons writing, but because not all the suggestions are explicit (e..g change this word to this word) that might be avoided.
[+] bergie|12 years ago|reply
Awesome! I think it would be great if editors could help users to not only format their contents, but also to write better.

Another idea in sort of similar direction is doing automated link suggestions: http://bergie.iki.fi/blog/automated-linking/

[+] pathdependent|12 years ago|reply
I like this app. However, most of my work Latex based. If you had something like this for Sublime Text, I would buy it.
[+] serverascode|12 years ago|reply
I think I would too. I like the idea. Not sure about calling it Hemingway, because I am a huge fan of the writer, but I would like to have something like this available for Sublime.

It might have to deal with markdown.

[+] buster|12 years ago|reply
Does lots of written work in restructed text and would buy as well (a sublime text addon, for example)!