Summary

Zemanta is one of a number of semantic computing startups attempting to address the problem on how to enhance the content that a publisher authors.  Zemanta attempts to answer this question by processing the text in the content in real time to suggest four things that are contextually relevant: images, stories, in-text links and tags. For now the company has focused on enhancing the content for blog publishers. However, it seems in the future, they have plans of introducing support for distribution channels such email clients and other content management systems.

Context
Zemanta received a seed funding of $ 650,000 on 15 September 2008, from Union Square ventures raising the total to $1.5 million. The investment from Union Square ventures seems to be the first ever investment by a US VC in a Slovenian tech company. The other investors are Eden Ventures, The Accelerator Group and Saul Klien.

Zemanta was launched in Alpha on 28 May 2008.  Aleš Špetič is the CEO who earlier ran one of the earliest trading companies in Europe overseeing its operations for a couple of years. Andraž Tori is a co founder and the CTO who also seems to be the technology brain behind the service. Andraž had previously cofounded Cyberpipe and worked as a core developer in Cinelerra, an open source video editing solution. Boštjan Špetič is another co founder and Product Director who had previously worked on Cyberpipe.

Technical Background

Zemanta uses natural language processing algorithms to process the content that a blogger authors. The semantic parsing of the blogger’s content is done in real time and searched for recommendations on the data indexed by Zemanta. The source for this indexed data is identified by Zemanta which includes blogs, Wikipedia, IMDB, Amazon’s book listings, Flickr and other popular news and entertainment websites such as The New York Times, CNN among others. All this indexed content is validated of not violating any copyright laws before recommending them. The company notes that the images sometime may not be relevant to the content because of the limitations in semantic organization of the image data.

Services

The company’s flagship service of content suggestion can be accessed in following mechanisms:

1. A plug-in for parsing of the content and a widget for distribution of recommended content. The plug-in and widget is supported on popular browsers and publishing platforms. A complete list of supported platforms is available here.

2. Access through an API. Example extension services such as Researchr and Faviki were built using the API service.

The company plans to offer other distribution channels for example integrating with an email client.

Buzz
Zemanta does not provide the number of installations of its widget. However every release of the product seems to be creating enough buzz that may result in accelerated service adoption. For example the release last month in September resulted in more than twenty reactions from the blogosphere in less than a day.

The Race
While Zemanta might lead the buzz around its launches there are a number of services that provide somewhat similar services such as: Evri, Calcais, Extensions based on Calcais (Tagaroo) among others. However the key strategic differentiator among the players in this business would be the number of widget installations, because widget is the key component that distributes the suggested content and probably will be utilized to push advertisements for monetization. The motivation for widget installation from users would be influenced by two key factors: 1. relevancy of the suggestions and 2. suggested content must be free of copyright issues. Zemanta already has these two features.

For those who are interested in building recommendation systems on the web can check out open source raw components such as Apache Mahout, Lingpipe, WEKA, Apache Nutch, Apache Lucene and IBM-Yahoo OmniFind.