LinkedIn operates the world’s largest professional network on the Internet, designed to help members manage their professional brand, connect with others for business opportunities and gain insights to be great at what they do. Headquartered in Mountain View, California, the site launched in 2003 and has more than 238 million users around the world in over 200 countries. Included in LinkedIn’s suite of tools is the CardMunch app, which turns business cards into digital contacts and online connections to make it possible to get to know the person behind the card. Business cards are transformed by capturing images of each card using mobile phones and sending the information to the Amazon Mechanical Turk marketplace for transcription.

LinkedIn acquired CardMunch in 2011 as an extension of the company’s mission of connecting the world’s professionals to make them more productive and successful. When the CardMunch app was created in 2009, there were several options available for transcribing business cards for digital use, but none met the developers’ standards for accuracy. “The level of accuracy that most of them delivered would be fine for some apps, but wasn’t going to work for us,” says Sid Viswanathan, LinkedIn product manager and developer of CardMunch. “If you transcribe one character incorrectly in a phone number or an email address, the information becomes useless. We decided to launch an application that used crowdsourcing to get the accuracy we need.”

When building the CardMunch app, the developers decided to use Amazon Mechanical Turk, which provides on-demand access to a global community of more than 500,000 independent Workers in order to achieve the accuracy needed for a great customer experience. The company considered using off-the-shelf optical character recognition (OCR) solutions, but the OCR solutions didn’t provide the accuracy levels CardMunch needed. With Amazon Mechanical Turk, the company was able to achieve near-absolute accuracy.

CardMunch works by taking a picture of a business card, sending the image to Amazon Mechanical Turk to be transcribed by Workers, and then returning the transcribed information to the user as a mobile contact. To increase accuracy, each business card is sent to up to four Workers for transcription. Additionally, CardMunch uses test business cards (also referred to as Known Answers) to evaluate the quality of new Workers’ transcriptions and to evaluate each Worker’s performance on an ongoing basis. Workers who maintain accuracy scores of 90 percent or higher on Known Answers qualify as “trusted Workers.” When a business card is sent to a trusted Worker, only one Worker is used to transcribe the card—speeding processing time and reducing cost per card without sacrificing transcription accuracy.

LinkedIn also uses Amazon Simple Storage Service (Amazon S3) to store images of the business cards.

Millions of cards have been transcribed using Amazon Mechanical Turk since the launch of the CardMunch app. “With the Amazon Mechanical Turk marketplace, we manually transcribe tens of thousands of cards each day, while having the ability to burst up and scale down as needed,” Viswanathan says. “We’ve been able to drive down operational costs while not having to guess at our capacity needs.”

Using Amazon Mechanical Turk’s API, LinkedIn was able to implement an automated, scalable process for manually transcribing card images in real time with near-absolute accuracy. Business cards can be transcribed around the clock, providing on-demand results for CardMunch users within 24 hours. LinkedIn is able to access thousands of Workers on Mechanical Turk on a pay-as-you-go, variable-expense basis without long-term contracts or required minimum commitments.

The elasticity of Amazon Mechanical Turk enables LinkedIn to grow. “The flexibility and responsiveness of Amazon Mechanical Turk is helping us continue to scale the CardMunch experience,” Viswanathan says. "I think there are vast possibilities for crowdsourcing at large data-driven companies. Amazon Mechanical Turk is an integral way of surfacing data at LinkedIn."

To learn more about how Amazon Mechanical Turk can help you leverage a scalable workforce, visit our Amazon Mechanical Turk product page: