Deep Learning Can be Applied to Natural Language Processing



By Carlos Perez, Intuition Machine.


Image credit

There is an article going around the rounds at LinkedIn that attempts to make an argument against the use of Deep Learning in the domain of NLP. The article written by Riza Berkan “Is Google Hyping it? Why Deep Learning cannot be Applied to Natural Languages Easily” has several arguments about DL cannot possibly work and that Google is exaggerating its claims. The latter argument is of course borderline conspiracy theory.

Yannick Vesley has written a rebuttal “Neural Networks are Quite Neat: a Reply to Riza Berkan” where he makes his arguments on each point that Berkan makes. Vesley’s points are on the mark, however one can not ignore the feeling that DL theory has a few unexplained parts in it.

However, before I do get into that, I think it is very important for readers to understand that DL currently is an experimental science. That is, DL capabilities are actually discovered by researchers by surprise. There are certainly a lot of engineering that goes into the optimization and improvement of these machines. However, its capabilities are ‘unreasonably effective’, in short, we don’t have very good theories to explain its capabilities.

It is clear that there are gaps in understanding are in at least 3 open questions:

  1. How is DL able to search high dimensional discrete spaces?
  2. How is DL able to perform generalization if it appears to be performing rote memorization?
  3. How does (1) and (2) arise from simple components?

Berkan’s arguments exploit our current lack of a solid explanation with his own alternative approach. He is arguing that a symbolicist approach is the road to salvation. Unfortunately, no where in his arguments does he reveal the brittleness of the symbolicist approach, the lack of generalization and the lack of scalability. Has anyone created a rule based system that is able to classify images based on low level features that rivals DL? I don’t think so.

DL practitioners, however, aren’t stopping their work just because they don’t have air tight theoretical foundations. DL works and works surprisingly well. DL at is present state is an experimental science and it is absolutely clear that there is something going on underneath the covers that we don’t fully understand. A lack of understanding however does not invalidate the approach.

To understand the issues better, I wrote in an earlier article about “Architecture Ilities found in Deep Learning Systems”. I basically spell out the 3 capabilities in DL:

  • Expressibility — This quality describes how well a machine can approximate universal functions.
  • Trainability — How well and quickly a DL system can learn its problem.
  • Generalizability — How well machine can perform predictions on data that it has not been trained on.

There are of course other capabilities that also need to be considered in DL: Interpretability, modularity, transferability, latency, adversarial stability and security. But these are the main ones.

To get our bearing right about explaining all of these, we have to consider the latest experimental evidences. I’ve written about this here “Rethinking Generalization” which I summarize again:

The ICLR 2017 submission “Understanding Deep Learning required Rethinking Generalization“ is certainly going to disrupt our understanding of Deep Learning . Here is a summary of what the had discovered through experiments:

1. The effective capacity of neural networks is large enough for a brute-force memorization of the entire data set.

2. Even optimization on random labels remains easy. In fact, training time increases only by a small constant factor compared with training on the true labels.

3. Randomizing labels is solely a data transformation, leaving all other properties of the learning problem unchanged.

The point here that surprises most Machine Learning practitioners is the ‘brute-force memorization’. See, ML has always been about curve fitting. In curve fitting you find a sparse set of parameters that describe your curve and you use that to fit the data. The generalization that comes into play relates to the ability to interpolate between points. The major disconnect here is that DL have exhibited impressive generalization, yet it cannot possibly work if we consider them as just memory stores.

However, if we consider them as holographic memory stores, then that problem of generalization has a decent explanation. In “Deep Learning are Holographic Memories” I point out the experimental evidence that:

The Swapout learning procedure which tells us that if you sample any subnetwork of the entire network the resulting prediction will be the similar to any other subnetwork you look sample. Just like holographic memory where you can slice of pieces and still recreate the whole.

As it turns out, the universe itself is driven by a similar theory called the Holographic Principle. In fact, this serves as a very good base camp to begin a more solid explanation of the capabilities of Deep Learning. I introduce the “The Holographic Principle: Why Deep Learning Works” where I introduce a technical approach of using Tensor Networks that performs a reduction of the high dimensional problem space into a space that is computable within acceptable response times.

So going back again to the question about wether NLP can be handled by Deep Learning approaches. We certainly know that it can work, afterall, are you not reading and comprehending this text?

There certainly is a lot of confusion in the ranks of expert data scientists and ML practitioners. I was aware of the existence of this “push back” when I wrote: “11 Arguments that Experts get Wrong about Deep Learning”. However, Deep Learning likely can be best explained by a simple intuition that can be explained to a five year old:


DE3p Larenn1g wrok smliair to hOw biarns wrok.

Tehse mahcnies wrok by s33nig f22Uy pa773rns and cnonc3t1ng t3Hm t0 fU22y cnoc3tps. T3hy wRok l4y3r by ly43r, j5ut lK1e A f1l73r, t4k1NG cmopl3x sc3n3s aNd br3k41ng tH3m dwon itno s1pmLe iD34s.

A symbolic system cannot read this, however a human can.

In 2015, Chris Manning, an NLP practitioner wrote about the concerns of the field regarding Deep Learning (see: Computational Linguistics and Deep Learning). It is very important to take note of his arguments since his arguments are not in conflict with the capabilities of Deep Learning. His two arguments why NLP experts need not worry are as follows:

(1) It just has to be wonderful for our field for the smartest and most influential people in machine learning to be saying that NLP is the problem area to focus on; and (2) Our field is the domain science of language technology; it’s not about the best method of machine learning — the central issue remains the domain problems.

The first argument isn’t a criticism of Deep Learning. The second argument explains that he doesn’t believe in one-size-fits-all generic machine learning that works for all domains. That is not in conflict with the above Holographic Principle approach that indicates the importance of the network structure.

To conclude, I hope this article puts an end to the discussion that DL is not applicable to NLP.

If perhaps you still aren’t convinced, then maybe Chris Manning himself should convince you himself:

Bio: Carlos Perez is a software developer presently writing a book on “Design Patterns for Deep Learning”. This is where he sources his ideas for his blog posts.

Original. Reposted with permission.

Related:



from KDnuggets http://www.kdnuggets.com/2017/01/deep-learning-applied-natural-language-processing.html

Fraugster, a startup that uses AI to detect payment fraud, raises $5M

Fraugster, a German and Israeli startup that has developed Artificial Intelligence (AI) technology to help eliminate payment fraud, has raised $5 million in funding.

Earlybird led the round, alongside existing investors Speedinvest, Seedcamp and an unnamed large Swiss family office. The new capital will be used to add to Fraugster’s headcount as it expands internationally.

Founded in 2014 by Max Laemmle, who previously co-founded payment gateway company Better Payment, and Chen Zamir, who I’m told has spent more than a decade in different analytics and risk management roles including five years at PayPal, Fraugster says it’s already handling almost $15 billion in transaction volume for “several thousand” international merchants and payment service providers, including (and most notably) Visa.

Its AI-powered fraud detection technology learns from each transaction in real-time and claims to be able to anticipate fraudulent attacks even before they happen. The result is that Fraugster can reduce fraud by 70 per cent while increasing conversion rates by as much as 35 per cent. The point of any fraud detection technology, AI-driven or otherwise, is to stop fraudulent transactions whilst eliminating false positives.

“We founded Fraugster because the entire payment risk market is based on outdated technology,” the startup’s CEO and co-founder Max Laemmle tells me. “Existing rule-based systems as well as classical machine learning solutions are expensive and too slow to adapt to new fraud patterns in real-time. We have invented a self-learning algorithm that mimics the thought process of a human analyst, but with the scalability of a machine, and gives decisions in as little as 15 milliseconds”.

Once integrated, Fraugster starts collecting transaction data points such as name, email address, and billing and shipping address. This is then enriched with around 2,000 extra data points, such as an IP latency check to measure the real distance from the user, IP connection type, distance between key strokes, and email name match. Then the enriched dataset is sent to the AI engine for analysis.

“At the heart of our AI engine is a very powerful algorithm which can mimic the thought process of a human analyst reviewing a transaction. As a result, we can analyze the story behind every transaction and say with precision which transactions are fraud and which aren’t,” explains Laemmle.

“You get a score or decision. Results are completely transparent (and not a black box), so you can understand exactly why a transaction was blocked or accepted. On top of this, our speeds are as low as 15ms. The reason why we’re so fast is because we’ve invented our own in-memory database technology”.

Fraugster cites competitors as incumbent enterprise level companies like FICO or SAS, which, it claims are based on outdated technology.

Adds Laemmle: “At Fraugster, we do not use any rules, models or pre-defined segments. We don’t use a single fixed algorithm to analyze transactions either. Our engine reinvents itself with every new transaction. This lets us understand transactions individually and therefore decide which one is fraudulent and which one isn’t. As a result, we can offer unprecedented accuracy and the ability to foresee fraudulent transactions before they happen”.

from TechCrunch https://techcrunch.com/2017/01/16/fraugster/?ncid=rss

The 3 tenets of applied ethnography

User experience design is an intriguing field.


It’s relatively new, and relatively subjective. When designing a user experience, there’s a lot of judgement involved. For every piece of quantitative data we can use, there’s a piece of qualitative data that must be interpreted.


And even when there’s a wealth of quantitative data, we still must apply design logic—a practice that varies from designer to designer.


At first, this requirement may seem like a weakness, but it’s actually one of the design field’s greatest strengths: the interpretation and application of data in varying ways is why we see innovation in UX.


The industry giants in particular continue to set a rapid pace of innovation. Meanwhile, the day-to-day design process may have become a little stale for many “regular” UX designers.



“UX is an extension of psychology.”




So how can we designers jumpstart our innovative engines? Interestingly enough, many UX designers often overlook one of the richest sources of virtually free (and objective) innovation and inspiration: psychology.


Related: Design and the psychology of time


This shouldn’t come as much of a surprise—UX is essentially an extension of psychology. Indeed, UX is an application of psychological principles in a context that traditionally has little to do with psychology.


The practice of ethnography has many applications in UX. Let’s discuss some ways to apply ethnographic research to UX design.


Ethnography in UX




You’ve probably already encountered ethnography during the course of your UX career thus far, either in practice or in theory.
Unfortunately, most, if not all, of the articles on ethnography are about the same thing: why you should conduct ethnographic research.


What’s not often talked about is why you should embrace it. So here’s a UX designer’s a definitive guide to applying ethnography.




The true power of ethnography is in its exploration of the social, team, political, and organizational influences that guide the views and decisions made by humans.


It’s rooted in the principle that individual views and decisions are guided by culture as much as they are by, well, anything else. This is an intuitive stance, and it’s heavily applicable to UX design.


Let’s dive in.


1. Data collection


It all starts with data. How do you collect it, analyze it, and apply it?


Let’s address the former first. Data collection in ethnographic research is a very qualitative affair. The typical setup involves a significant amount of observation and note taking, along with the occasional question.


In psychology, ethnographic research is typically conducted as though the researcher were a part of the group that is the subject of scientific study. That is, the researcher should conduct research through the eyes of the subject, so that he or she may paint a more complete picture of the group’s views and beliefs.
And this is true for UX as well.



“Ethnographic research gives you data from your user’s point of view.”


Whether you’re conducting ethnographic research for an app that is going to launch in a foreign country, or you are designing an experience that is specific to a particular region, you should embed yourself into this culture for the duration of your research. 
A good way to do this is to experience firsthand the ways your product may be used in the day-to-day lives of your target users.


For example, if you’re designing an app for users of a certain culture who wish to send money back and forth to their family, you would research the needs and beliefs of that group by engaging in the process within their cultural environment—essentially, living a few days in their shoes while using your app.




2. Turning qualitative data into quantitative data


Along with proper cultural immersion, perhaps the greatest challenge in ethnographic research is converting the qualitative data it generates into quantitative data that can be used to make design decisions.


There are many qualitative measures in UX, so you’re likely to be familiar with this process.


One good way to generate quantitative measures is to categorize the qualitative observations and concerns into relevant and consistent groups. Then you determine the frequency of those measures.


The frequency of each issue type can be plotted on a graph and used to inform countless design decisions. This is called the incidence of an issue.


Another good tool when converting qualitative data into quantitative measures is a confidence interval, which allows you to apply a “confidence percentage” to your data, based on the number of people sampled in a specific data set.


The higher the confidence interval, the more certain you are that the sampled population is fairly represented by the sample data. This can often be useful when justifying design decisions to the stakeholders of a project.


3. Using ethnography to validate or generate assumptions


What good is ethnographic data if you don’t use it to validate or generate assumptions? In fact, the best way to use ethnographic data is to compare it with existing assumptions and develop them into knowns.


So what can we validate through ethnographic data?


First, let’s consider the basics: a group’s feelings and presumptions.



“Ethnographic data helps you generate, challenge, and validate your design assumptions.”


By immersing yourself in a group or culture for a period of time, collecting observational data, and determining the frequency of those observations, the diligent UX researcher can easily determine a group’s feelings and presumptions on a subject.


In turn, knowledge of a group’s feelings and presumptions can be used to confidently inform design decisions.


Another interesting data set that can used to validate assumptions is the incidence of experience roadblocks. That is, how often—and in what form—do roadblocks occur in the experience of your target user base.


A roadblock can be classified as anything that deters a particular group from accomplishing a given task, or interacting with a given system. Knowledge of roadblocks, and specifically their incidence, can be used to help address whatever inefficiencies may be causing them.


Another phenomenal dataset to generate through ethnographic research is motivation and reward within a group. Specifically, what thoughts, feelings, and environmental factors might be causing members of a group to make particular choices?


This information can be critical when designing an experience that must inherently motivate a particular group to interact with the product.


Conclusion: Use ethnographic data as often as possible




As challenging as it can be to research a target audience from its own perspective, the rewards sure do seem to be worth it. The notion of a truly validated assumption, albeit through qualitatively generated data, is a profound one.




By immersing yourself in the culture and lifestyle of a group, you can generate data that can be extremely difficult to collect through surveys and other traditional data collection techniques.

Often, the challenges facing a group are not easily voiced, nor easily communicated. Sometimes the only way to mitigate the challenges facing these traditional methods is to take on the perspective of the user, for as long as it takes to generate enough usable data.


Ethnographic research can be used to generate and validate assumptions that can otherwise be very difficult to prove.


Should you succeed in your mission to collect ethnographic data, you will be rewarded with deep insight into the problems, thoughts, and concerns of your target population to an extent that is rarely attainable in the world of UX.



More posts like this

Yona Gidalevitz
Yona is Codal’s technical researcher. At Codal, he is responsible for content strategy, documentation, blogging, and editing. He works closely with Codal’s UX, development, marketing, and administrative teams to produce all manner of written content. You can check out his work on Codal’s blog, Medgadget’s blog, and Usability Geek. In his free time, Yona is an avid guitarist, cook, and traveler.

from InVision Blog http://blog.invisionapp.com/ux-applied-ethnography/

Unpacking Atlassian’s Acquisition Of Trello

Wrapping up my first CES/Vegas retreat, I boarded the plane to check Twitter to see — lo and behold — that Trello had been acquired by Atlassian for $425M in a great, quick early-stage venture outcome. There’s quite a bit to unpack here, so I’ll just leave a few thoughts here but would love to hear more from the crowd about the implications of this move:

1/ Accidental Happenings and Side Projects: I do not mean to suggest Trello’s success and outcome is accidental, but rather that it doesn’t appear (from afar) that Trello had a normal birth or childhood. Trello was created inside Fog Creek Software, co-founded by Joel Spolsky, and then spun out in 2014 and funded by a mix of seed investors and early-stage VCs. Spolsky became CEO of Stack Exchange and was Chairman of Trello, and I believe another Fog Creek founder ran Trello. As it started to grow, someone else ran Fog Creek. This may be fodder for another post at a later date, as the genesis of this outcome seems both accidental and also a bit looser, more creative than the traditional business rigidity with which we read about in countless startup “how-to” blogs. (Fun update: Per my friend, Sean Rose: “when Trello was still part of Fog Creek, it was funded via Fog Creek employees opting to have their bonuses go to the project.”)

2/ Cross-Platform Architecture, Mobile Card Format, and Business Integrations: Slack launched cross-platform from day one, on web and mobile. I am not exactly sure of Trello’s history — it seems if they were web-first, mobile responsive, and then launched for iOS. Additionally, the interaction model of Trello featured boards (like Pinterest), which displayed nicely as cards in a mobile app. Finally, the Trello team had quietly built many storage and business process integrations into their offering, giving some of them away as a hook and charging larger teams for the privilege to stack them up. (Trello also didn’t have thousands of integrations, but enough to make customers happy — more integrations likely doesn’t mean they’re all useful.)

3/ Consumerization of Enterprise: This has been an “eye-rolling” buzzword, but we have to accept it is an apt descriptor. Following the success of prosumer designs in apps like Slack, Asana, Wunderlist, and others (more on this below), Trello’s design delivers a lightweight experience to users with enough infrastructure and power to fuel large teams across many different platforms. Trello simply feels like a consumer product, something that may have been designed inside Google or Facebook — but much better, cleander.

4/ Capital Efficiency: Assuming Crunchbase and my sources are correct, Trello is (relatively) a modern case study in capital efficiency. Having only raised about ~$10M, Trello seemed to not only grow its team (over 100) and its user base (19M+) quickly, they also marketed a three-tier freemium product that charged more to small businesses and even more for enterprise customers. In VC-math terms, Trello likely produced a 8.5x realized (mostly in cash) exit for its investor in less than three (3) years (which positively impacts IRR) and didn’t have to raise round after round of capital. Compared to some of its peer products like Asana and Wunderlist, among others, Trello has been remarkably capital efficient relative to its exit value. (A reader notes that it’s spinout from Fog Creek also adds to its capital efficiency.)

5/ Enterprise SaaS consolidation: For years now, we have witnessed different varieties of M&A across enterprise SaaS, whether it’s an incumbent like Salesforce scooping up new products or private equity shops buying small-cap public companies, there’s more and more pressure in the environment for the larger companies to expand their offerings to grow, as well as financial incentives for buyouts led by managers who can profit from creatively rolling-up disparate end-point solutions. In a world where collaborative products like Slack or Facebook @ Work or Microsoft Teams are growing and/or boast infinite financial resources, other growing incumbents (like Atlassian) need to prepare for a long-term product and mindshare battle and scooping up Trello is a good step in that direction. As Fred Wilson predicted a few days ago for 2017, “The SAAS sector will continue to consolidate, driven by a trifecta of legacy enterprise software companies (like Oracle), successful SAAS companies (like Workday), and private equity firms all going in search of additional lines of business and recurring subscription revenue streams.”

6/ “If I Can Make It There, I’ll Make It Anywhere” – Another solid exit for the NYC startup market, and there are bigger ones to come. Despite Trello being young and a SMB/enterprise product from NYC, it recently internationalized to a few non-English-speaking markets worldwide. As a bonus, while I don’t know the team, from what I hear from friends, Spolsky, Pryor, and their team are well-respected and seem to have done things the right way — their way. Congrats on building a great product.

from Haystack http://blog.semilshah.com/2017/01/10/unpacking-atlassians-acquisition-of-trello/

New Study – Mobile Registration Forms on E-commerce Websites

(This branded content is brought to you by UseItBetter)

The European e-commerce market has witnessed substantial growth over the past several years. Last year, online shoppers contributed 455.3 billion euros to the market, and this figure goes up each year.

20 Gorgeous Mobile App Landing Pages

20 Gorgeous Mobile App Landing Pages

Most mobile app developers have an accompanying site that introduces and describes what their mobile app can do.…Read more

If you’re working in e-commerce, your list of resolutions for the New Year should include optimizing your registration and checkout forms for small screens. And, unless you are planning to quit smoking, it should be on the top of that list.

A great way to start is to compare how big brands design their forms. Well, the good news is that UseItBetter, the company behind highly effective form analytics, has done all the hard work.

useitbetter forms

UseItBetter conducted a comparative analysis of mobile registration forms on 40 top e-commerce websites from the perspective of a UK mobile shopper. They compiled all form fields, buttons and labels and turned this information into numbers and charts.

You can see the full study, including screenshots of those forms here. But let’s look together at some of the points that we found interesting.

Number of fields in registration forms

Customers prefer registration forms they can fill out quickly and easily. Especially when they are shopping on mobile. They accept information requests that seem appropriate, but usually balk at requests they regard as intrusive.

The business on the other hand, wants or needs certain data, and at times, data that can be used for marketing purposes, or to personalize a user’s shopping experience.

A rough rule of thumb is that for every additional question asked, the greater is the risk of abandonment. So how far websites are pushing their luck?

Out of the 40 websites that UseItBetter studied, 33 require users to take 10 or more actions (complete form fields and click buttons) in order to register an account.

useitbetter studies

Considering that all websites in the study are big successful online stores, we can assume that the risk pays off for them, and at their scale even a small percentage of users abandoning forms are worth a lot of money.

No “One-Form Fits All”

It’s interesting how similar some of those forms look like. Some of them have absolutely no branding, and so it’s practically impossible to say which is which. Yet, after a closer look at all those forms, it seems that there’s no fixed formula for building a perfect registration form.

no one form fits all

The UseItBetter study allows you to compare side by side, forms from different websites.

The number of form elements ranged from as few as five, to as many as 34, but it’s difficult to spot any patterns. Multi-department stores like jdwilliams.co.uk or very.co.uk requires 20 interactions but their competitor johnlewis.com just 5: email address, password, password confirmation, newsletter signup, and the registration button. Fashion brand hm.com has 8 form elements and zara.com has 17 – twice as many.

Requirement to enter the password

The UseItBetter study also revealed that 72% of the online retailers require users to enter their passwords twice, to correct any problems due to negligence. Even though it places an added burden on users, very few users object to this, as they are more prone to making typo errors using their mobile.

Roughly 20% of the retailers give the user an option to see their password while writing. A third requires user email addresses to be repeated.

For instance, qvick.com has the most complex registration process. Users must enter and confirm their email and password, respond to a security question, provide a security answer, and enter a PIN number.

Requirement of entering the Birth Dates

Birth dates are typically used by hospitals and clinics to ensure that patients do not receive the wrong treatments or medications, as well as to keep patient files organized. Retailers usually request this information for marketing and demographic reasons, to better personalize customers’ accounts, or to offer birthday greetings or birthday discounts. Some retailers require this information for security reasons. Or because they sell certain age-restricted items.

Examples:

  • asos.com – If you tell us, you’ll get a birthday treat.

Providing a birth date would appear to be optional in this example. However, in the following examples, it appears to be mandatory.

  • jdwilliams.co.uk – Please provide your date of birth as an additional security measure.
  • next.co.uk – You must be 18 years old or over to shop at nextco.uk
  • qvcuk.com – We request your date of birth because we sell certain age-restricted products.
useitbetter dropdown

How retailers put these requests forward is equally important as how a UX specialist presents these requests. A dropdown form, a calendar, and type and text are the most common options. Most retailers appear to prefer the dropdown menu approach. They also prefer that the number of input fields be kept to a minimum, due to issues that an empty field can sometimes cause.

How personal data is used

How retailers use personal data has been an important, and at times, a controversial topic. The study indicated that 85% of online retailers request this information for marketing purposes (for marketing strategy and demographic research, rather than for promotional activities), while 15% request this information for 3rd party marketing.

Approximately two-thirds of European retailers’ websites can take personal information by default, to be used for marketing purposes. Users can opt out, but mobile users especially may find it difficult to fathom the fine print.

Combine best practices with real data

Even when registration forms are designed in keeping with business practices, any attempt to create a perfect form for a given business will fail, and that’s because of the users. Different users have different mindsets and varying histories of online experience. A form that does not present a problem for one user, may present a challenge, or even appear objectionable, to another.

That’s why it is necessary to go beyond best-practices, usability testing and start looking at your real users with analytics. The analytics can provide the retailers field-after-field conversion funnels, tracks form validation errors and even auto-detects critical issues by comparing patterns among failed and successful user visits.

If not, then at least check their guidelines for form tracking which you can implement using Google Analytics or any other tool you use.

4 Form Design UX Tips You Should Know (With Examples)

4 Form Design UX Tips You Should Know (With Examples)

We tend to think about forms as simply a means to collect user data, but they are also…Read more

from Hongkiat.com http://www.hongkiat.com/blog/mobile-registration-forms-ecommerce-websites/

Isaac Asimov: How to Never Run Out of Ideas Again

3. Beware the Resistance

All creatives — be they entrepreneurs, writers or artists — know the fear of giving shape to ideas. Once we bring something into the world, it’s forever naked to rejection and criticism by millions of angry eyes.

Sometimes, after publishing an article, I am so afraid that I will actively avoid all comments and email correspondence…

This fear is the creative’s greatest enemy. In the The War of Art, Steven Pressfield gives the fear a name.

He calls it Resistance.

Asimov knows the Resistance too —

The ordinary writer is bound to be assailed by insecurities as he writes. Is the sentence he has just created a sensible one? Is it expressed as well as it might be? Would it sound better if it were written differently? The ordinary writer is therefore always revising, always chopping and changing, always trying on different ways of expressing himself, and, for all I know, never being entirely satisfied.

Self-doubt is the mind-killer.

I am a relentless editor. I’ve probably tweaked and re-tweaked this article a dozen times. It still looks like shit. But I must stop now, or I’ll never publish at all.

The fear of rejection makes us into “perfectionists”. But that perfectionism is just a shell. We draw into it when times are hard. It gives us safety… The safety of a lie.

The truth is, all of us have ideas. Little seeds of creativity waft in through the windowsills of the mind. The difference between Asimov and the rest of us is that we reject our ideas before giving them a chance.

After all, never having ideas means never having to fail.

from Sidebar http://sidebar.io/out?url=https%3A%2F%2Fmedium.com%2Fpersonal-growth%2Fisaac-asimov-how-to-never-run-out-of-ideas-again-b7bf8e09cc91%23.e9plvz3ba

Has the Internet Killed Curly Quotes?

The trouble with being a former typesetter is that every day online is a new adventure in torture. Take the shape of quotation marks. These humble symbols are a dagger in my eye when a straight, or typewriter-style, pair appears in the midst of what is often otherwise typographic beauty. It’s a small, infuriating difference: “this” versus “this.”

Many aspects of website design have improved to the point that nuances and flourishes formerly reserved for the printed page are feasible and pleasing. But there’s a seemingly contrary motion afoot with quotation marks: At an increasing number of publications, they’ve been ironed straight. This may stem from a lack of awareness on the part of website designers or from the difficulty in a content-management system (CMS) getting the curl direction correct every time. It may also be that curly quotes’ time has come and gone.

Major periodicals have fallen prey, including those with a long and continuing print edition. Not long ago, Rolling Stone had straight quotes in its news-item previews, but educated them for features; the “smart” quotes later returned. Fast Company opts generally for all “dumb” quotes online, while the newborn digital publication The Outline recently mixed straight and typographic in the same line of text at its launch. Even the fine publication you’re currently reading has occasionally neglected to crook its pinky.

This baffles Matthew Carter, a type designer whose work spans everything from metal type’s last stand to digital’s first, and whose dozens of typefaces, like Verdana and Georgia, are viewed daily by a billion-odd people. “I have no idea why people don’t use proper quotes. They are always [included] in the font,” Carter says.

This lack of quote sophistication is odd, because the web’s design origins owe a lot to choices Steve Jobs made at Apple and later at his second computer firm, Next. Jobs’s attachment to type famously stems from a calligraphy class taken at Reed College, and he ensured that the first Mac had a mix of bespoke and classic typefaces that included curly quotes and all the other punctuation a designer could want. At Next, he went further, and the web’s father, Tim Berners-Lee, built the first browser and server on a Next.

But in the early days of the web, different computing platforms—Unix, Mac, and Windows, primarily—didn’t always agree with how text was encoded, leading to garbled cross-platform exchanges. The only viable lingua franca was 7-bit ASCII, which included fewer than 100 characters, and omitted letters from alphabets outside English and curly quotes.

Marcin Wichary, the current design lead at Medium responsible for pushing forward on typographic niceties, grew up in Poland, and says in his youth, most computers simply omitted his language’s ę, ł, and other diacriticals. He says he felt privately glad that his first and last names lacked a missing Polish letter. It took years before one of his middle names was easy to type.

But ASCII and a few similar small character sets acted as a limitation only early on. With the right effort, even by the late 1990s, a browser could properly show the right curly quotes. But effort is the right word: While browsers could show typographers’ quotes, it was hard for users to type them.

* * *

Straight quotes appear as an abomination in a typeface, because their designers rarely love them; they’re included by necessity and often lack cohesion with other characters. The non-curly quote comes from the typewriting tradition, and arose from cost. As U. Sherman MacCormack wrote in The Stenographer in 1893: “For some time past the manufacturers of typewriters have adopted straight quotation marks, for the reason that the same character can be used at the beginning and end of the sentence, thus saving one key.”

At the time of the single quote’s popularization in the 1870s, the use of paired quotation marks was just over a hundred years old. Keith Houston, the author of The Book, has traced the history of many punctuation marks back hundreds and thousands of years on his blog Shady Characters, and in a book of the same name. “There was no quotation mark for a very long time,” he says. One first appeared in the third century BCE alongside the invention of basic punctuation. It resembled a right angle bracket, >, which was resurrected in the 1970s for quoting email without any apparent connection. (The history of that newer use of > remains undocumented.)

Scribes and printers chose different symbols and conventions, Houston says, until a regular comma and an inverted one—one rotated 180 degrees—used in the left and right margins came into vogue as “quotations marks” in 1525. “You’d see it on the outer margin or inner margin depending on who printed it,” sometimes pointing toward and sometimes away from the text.

But it took Samuel Richardson to make a consistently used paired set within the text. While he remains best known for his invention of the epistolary novel in English with Pamela in 1740, followed by other literary innovations, his trade as a printer long predated his novel writing. Until he came along, quotations remained marked in the margin for each line that contained any referenced text, not the starting and end point of the quote within the text itself. (It was also commonly used only for excerpts from other documents, not dialog.)

In the 1748 edition of Pamela, however, Richardson included not just these per-line commas, but also what we see as an opening quote and mirrored closing quote at the beginning and end of excerpts at the exact start and end in the run of text. The pairing and intraline practices quickly became standard, although it varied in style among countries. French writing instead features guillemets, « and », close relatives of the ancient > mark.

Typesetters likely first inverted commas until type foundries started casting proper quotes as separate pieces of type. When a “hot-metal” mechanized typesetter appeared in the late 1800s, it followed the tradition: The earliest Linotype keyboards had paired curly quotes and no straight ones. But practical typewriters, which began to appear around the same time as the Linotype, followed a different path. As a tool for note-takers like stenographers, telegraphers, and business secretaries, the typewriter had no need for the flourish of the curled quote—and it would have added cost, as MacCormack noted.

As metal typesetting equipment moved on an inexorable path towards extinction, typewriters begat teletypewriters, and those begat computer keyboards. Medium’s Wichary, at work on a book about keyboards, says he’s found just one computer keyboard that has curly quotes instead of straight: the Xerox Star 8010. Virtual keyboards have mostly followed the physical style.

Paul Ford, a writer and programmer known for his thoughts about how code affects culture, notes that even on a mobile device “the energy to type a curly quote feels prohibitive. You have to hold down the quote. The effort of typing one on a regular keyboard [also] can be prohibitive.” Some software automatically swaps in the “smart” quote, but doesn’t always get the right curl (decades should always be ’90s, but autoformat software often drops in ‘90s). For wonks, you can find cheatsheets for explicit shortcuts on desktop machines, like Shift-Option-] for a curly apostrophe on the Mac, but it requires additional effort and memorization.

* * *

Even when a writer gets things right, the CMS remains a stumbling block. “Smart quotes are traditionally one of the things that get turned into weird garbage characters when the character encoding is set poorly,” Ford says.

The result of the variation in input from Word documents and other sources, explains Claudia Rojas, Fast Company’s website product manager, led that publication’s website (but not print publication) to standardize on straight quotes for consistency. Fast Company doesn’t seem alone, as any survey of sites quickly finds others that have made the same choice. As Greg Knauss, a humorist and programmer who has built CMSes, elaborates: “If you use [straight] ASCII quotes, you know that they’re going to survive the cut-and-paste transition that often happens with text, as well as old or broken email servers and other 7-bit indignities.”

Straight quotes are a way to play things safe, in other words—but they’re not the only solution. Wichary has taken the opposite tack at Medium, developing code to guess a user’s intent as they type and format quotes automatically. “We took it further than I originally thought was possible,” he says, and estimates the site covers about 95 percent of possible situations. “A fraction of people who type rock ’n’ roll ask, ‘Why do those point the same way?’”

Conceivably, if they wanted to, all CMS designers could employ algorithms to always make the curl happen. It’s ultimately a software choice when quotes either all get converted to typewriter versions or remain inconsistent in the final product. Because of this, there’s a temptation to read the push toward straight quotes as a principled, pragmatic stand against the needless embellishment of a curl. But Anil Dash, once the chief evangelist of Six Apart, makers of Movable Type, argues there’s a different underlying issue with the current generation of widely used CMS software. “Typography is the kind of refinement that happens at the end of a generation of CMS tech,” he says.

While it might seem that CMSes—like WordPress and others—are mature, Dash says periodicals’ systems are in the midst of continuous updates to deal with all the formats required by content partners: Facebook’s Instant Articles, Google’s support for Accelerated Mobile Pages (AMP), Apple’s News feed requirements, and others. “Once this stuff is nice and boring in a year or two, their developers will probably refocus on type and layout details,” Dash says.

So perhaps curled quotation marks will again have their day. Or, by then, it’s possible conventions will have changed enough that people cease to notice. Wichary says in Poland, the lack of Polish-style quotation marks („ and ”)  have led the current generation to use American-style quotes and think the native ones look wrong.

Maybe periodicals, which sometimes commission typefaces or pay to adapt existing ones, will demand type designers draw better-looking, harmonious straight quotes that don’t seem pulled from typewriter typebars. Paul Ford is just plain resigned: “They sure do look nicer to old people like you and me, but frankly do they actually add any magical semantic value to a given text? Not really.”

from Sidebar http://sidebar.io/out?url=https%3A%2F%2Fwww.theatlantic.com%2Ftechnology%2Farchive%2F2016%2F12%2Fquotation-mark-wars%2F511766%2F

Bakarema © 2024.