Posts Tagged ‘Product Development’

reCAPTCHA Recycles Human Effort

Friday, January 9th, 2009

The American astronauts were rumored to be using a special Space Pen, which was developed specifically to work in zero-gravity conditions. It used a pressurized ink cartridge, enabling it to write upside down (Seinfeld fans should smile by now) and, of course, be used in space. It is believed that NASA spent a total amount of $11 million dollars to develop that pen.

The Russians gave their Cosmonauts a pencil.

Whether this story is accurate or not, it brings out a common product development truth – all things being equal, the simplest solution tends to be the right one. Hence the abbreviation K.I.S.S. – Keep It Simple, Stupid! And reCAPTCHA certainly did achieve a simple solution to several problems at once.

reCAPTCHA is a special captcha form component for identifying humans and rejecting spam and bots.  It is done through the traditional means of displaying distorted words for the users to type and verify. However, instead of generating a random word, reCAPTCHA displays words that were scanned from real text books, thus helping in the global effort of digitizing old texts.

I ran into reCAPTCHA a few months ago and was impressed by the concept of “recycling” human actions – taking something we do so commonly, almost without thinking, like verifying our humanity with forms through captchas, and harnessing it to both combat evil (spam and bots) and do good (digitize texts).

Name: reCAPTCHA
Website: recaptcha.net
Type: Captcha component
reCAPTCHA is a system that uses captcha to help digitize the text of books while protecting websites from bots attempting to access restricted areas.
reCAPTCHA

Value

Let’s take a second to ponder over a very disturbing fact: if we didn’t have spam or bots that were driving our websites off the wall – we wouldn’t have had reCAPTCHA, and as a result, many textbooks wouldn’t have been digitized as part of a universal effort.

When thinking of it that way, reCAPTCHA’s value starts to get a whole new meaning. It becomes very much like modern medicine – it wouldn’t have been discovered had we not gotten ill; I suppose we owe plenty of our discoveries to urgent needs or even accidents, such as the genome project, antibiotics, Champagne, and so forth. It’s strange to vocalize it, but many people around the world will be exposed to important texts digitally thanks to the plague called spam and bots.

That being said, lets examine reCAPTCHA’s value in two aspects: intrinsic value and global contribution.

Intrinsic value

reCAPTCHA has a very granular offering – it provides your product with a solution to something you’ll probably need as part of your registration form, or any other form that requires user authentication. You need to sort out the humans from the bots that attack your site.

Your R&D will most likely tell you that developing such a component is very simple, and there are plenty of open-source solutions out there that you can use. And they’ll be right to say so. Basically, all you need is a tiny utility that generates a random word, and displays it in a funny way. You can add colored squiggly lines if you’re overly creative. And that’s about that.

But reCAPTCHA’s solution has two main advantages:

  1. It’s off the shelf. You just cut and paste a little piece of code, and you get all of the intrinsic functionality already contained within it; also, the processing is done on the reCAPTCHA servers, which might free your own servers slightly, but on the other hand relies on a server-farm that is out of your control latency-wise. However, if Twitter and Facebook use it, then I guess it’s all right.
  2. It’s fully loaded. Two paragraphs ago, it sounded quite simple to make a workable captcha. However, to make it right, you need to consider people who might not be able to read your distorted image. Sometimes, the text is illegible, and needs to be changed. Again – implementing something like this is not like inventing the space pen, but why bother when this elegant component already exists?

Unreadable

(I’m not sure my dad would be able to read this one…)

Global contribution

I love recycling; not only in my trash habits at home, but also as a goal to live a greener life. Don’t worry – I don’t spend my spare time hugging trees, but I definitely believe that there’s a practical approach to recycling, which can ultimately be applied to business.

“I think there is a world market for maybe five computers,” according to Thomas J. Watson Sr., formerly the president of IBM. I’m afraid Mr. Watson may have missed the mark slightly. Our collective processing power today is gargantuan, and some applications make use of it very effectively.

File sharing, legal or not, is a perfect example for recycling processing power, as well as the wonderful SETI@Home project at Berkeley, which is currently the largest distributed computing effort around the world, utilizing over 3 million computers worldwide in order to analyze data from radio telescopes in search for extra-terrestrial intelligence. reCAPTCHA goes along those lines, but instead of recycling wasted processing time, they’re recycling wasted human effort.

Call me an idealist, but as a product junkie I think that part of making the world a better place revolves around reducing waste and channeling effort. Most of us share their hosing with others, use platforms developed by the online community, incorporate maps drawn by other companies, and make it an effort to make sure we reduce our costs and increase the productivity of our company.

With reCAPTCHA we have a double benefit – we get a ready-to-use component, and join a world effort to bring knowledge to the masses. That alone is a value in itself.

Value summary: Provides an off-the-shelf component that answers a specific requirement, while helping in a global effort to digitize texts.

Usability

reCAPTCHA

One thing to consider when creating such a generic and simple component is not to overdo it. I’m sure product development people would have loved to sink their teeth into this product and design their heart out with plenty of cool features, like changing the image size, making it full-screen, creating a hovering magnify-glass to make the letters bigger, etc. But that would have turned this component into a monster.

When dealing with a component so basic in its functionality, the product developer needs to constantly take a step back and look at the big picture. It’s terribly easy to look at your product under the microscope and be so caught up with its tiny world – I’m sure many of us have been there and realized our overzealousness.

One has to remember that this type of granular component lives inside the context of a foreign web page, which contains a form in it. It’s mostly the last, and least favorite, field the user needs to fill in. So it’s better to keep it as simple and as non-intrusive as possible. We’ve already agreed that forms are a pain – let’s not make them any worse.

reCAPTCHA managed to keep it simple. That’s their first great achievement.

Other than that – the usability is quite basic; after all, this is merely a form field. You can change the text with a new set of words, in case the image that’s presented to you is too hard to read, or you can change the text into an audio challenge taken from an old time radio show (therefore also transcribing old recorded shows).

That’s it; in terms of usability, the product fits the requirements.

Usability summary: Simple and complete.

Layout and design

Once again – such a simple widget doesn’t have much room for an elaborate layout and design schemes. However, one thing that I would like to see is customizable display: my website and forms are usually of a specific design, and I would have liked to make this field look and feel as a native field to my site. I have no problem with sharing credit or double branding, but reCAPTCHA’s choice of Bordeaux colors does not necessarily fit my overall website design.

Layout and design summary: Simple, though cannot be customized.

Under the hood

When I initially heard of reCAPTCHA, the first question that popped into my mind was “how does it work?” If the words that are shown to me are the ones that are used to digitize book, then how does the captcha verify my input as correct? Or else – if the captcha engine already knows the meaning of the word, then what part am I playing here exactly?

And here is where reCAPTCHA came up with the pencil solution.

After the texts are scanned, they go through an OCR process to interpret it. However, OCR is never perfect, and many words remain unresolved. Yet the OCR mechanism is clever enough to at least indicate which words were not interpreted correctly.

So reCAPTCHA came up with a very simple, yet ingenious solution: each user is presented with two words, not one. One word is a word that was successfully acquired by the OCR, while the other one was not. After typing these words, reCAPTCHA verifies the known word with your input; if the match is correct, it assumes that the second, unknown word is also correct.

This user-interpreted word is then given to a few other users for recapturing, and then, after cross verification, the word is declared as properly interpreted.

Why was I so impressed? Because, as a former engineer, I would definitely have overcomplicated something this simple. Can’t help but to doff your cap – well done!

Bottom line

When creating your online forms, it is great to be able to take part in a worldwide effort, and harness the power of the masses to a global endeavor. When your website joins the ranks of websites like Facebook, TicketMaster, Twitter, StumbleUpon and Craigslist, who already use reCAPTCHA, you know that you’re with good company.

Formalizers and Search Engines

Thursday, January 8th, 2009

Google search

Search engines are possibly the most commonly used, and perhaps the oldest type of formalizers out there, which many of us take for granted.

A search form has one major formalizer field, which requires a search query. The user can be quite expressive when filling in this field, while the formalizer field takes care of certain types of input which is expected. Google have been true pioneers in this sense of formalizer usage.

Surely, we all know that there is a specific format for search – we use quotation marks (“”) to group words together, we use logical operators such as OR and AND, or sometimes even use the plus (+) and minus signs (-), especially if you’re an advanced user. However, there are a few things popular search engines have taken care of:

  • Stop words. Also known as noise words in some circles, they are words that are filtered out of search queries, usually since they’re too common and appear far too many times in searchable documents. These words, like “is”, “are”, “the”, “you”, “me”, and many others, are omitted from search queries automatically by the search engine, in order to provide a more relevant search result list.
  • Spelling. Google really made a difference here, both by finding spelling mistakes in your search query and suggesting a correction for you in the search results, and providing the means for other applications to utilize their spelling mechanism with their SOAP search API.
  • Contextual search. When typing a specific search phrase that requires a specific type of data, Google makes sure they provide you with this data as quickly as they can, with the tools that they have. You can get stock quotes, airline schedules and word definitions simply by asking for them. This saves a lot of time and hassle by simply thinking like people do, so a simple phrase like “kilometer in mile” typed into the Google search box will make the simple conversion for us, and display it at the top of the page.

What can we learn from Google?

Many applications, especially web applications, provide search capabilities to their users. With search being such a common use-case, not to mention a perfect way to learn of your users’ habits, product development managers need to make their search engines as usable and formalizer-friendly as they possibly can.

Here’s a short checklist:

  1. Auto-suggest. When users start typing search queries, try to assist them with suggestions based on common search queries.
  2. Spelling. As mentioned before, there are services for checking the spelling of search queries out there, like the one Google offers us. Therefore, there’s no need to develop it in-house, and you can literally use the best tool out there. This way, users won’t get frustrated when a misspelled query gets sent back with no results.
  3. Natural language. This is probably the most specific, and therefore most complicated aspect of formalizer-friendly search engine. It requires you to define a way for the user to communicate with your application’s search engine in a natural language. Try to find common search phrases and to identify patterns that occur the most, and you’ll be able to see how you can go towards your users and help them by understanding their queries better.

Combining these three elements together creates a powerful search user experience, and positions you immediately above your competitors in terms of usability and barrier of entry.

For example: say you have a dating website, and your search box allows users to look for possible people to meet. You could provide a boring form that would require an “advanced search” option to be clicked (possibly the worst use-case a search engine could have, but more on that in a later post).

On the other hand, you can define a few search patterns that upgrades your search field into a formalizer. Such fields are not search-boxes in their classical sense, and are usually preceded with a question, such as “what kind of date are you looking for?” to which the user can answer “middle-aged men in the greater LA area.”

And now, the hard work begins: your language processor should be able to understand this sentence and extract meaningful information out of it.  It should understand that “middle-aged” means a certain age group, and that “the greater LA area” is a geographic region. Again – not the easiest of tasks, but once implemented, it goes a very long way in terms of usability.

This, combined with a spelling service and an auto-suggest utility, makes sure the user types the words correctly, or at least minimizes the number of mistakes the user makes. This allows your formalizer mechanism to be right more times than wrong and ultimately makes your users happier. And happy users will return again tomorrow, no doubt.