A project completed with my group while completing the IS project at IE Business School. I thought that it might be helpful for the people who are interested in this topic.
Video search is the process of locating, interpreting, indexing and delivering video content (Exhibit 1). This process is enabled by the usage of the sophisticated technologies, allowing consumers to easily find relevant professional videos and user-generated content. The leading organizations in this sphere of video search technology are Youtube, Blinkx and the traditional search giants Google, Yahoo and Bing.
Contents 
1. Definition
2. Benefits
3. Costs
5. Risks
6. Current Technology Applications
9. Exhibits
10. Endnotes
The online video content is growing exponentially as a result of several factors:
- Availability of the broadband connection, which dramatically increased Internet speed
- Development of Web 2.0 technologies, which boosted the amount of user-generated content
- Growing number of the broadcasting giants targeting the Internet audience
This growth has led to two main trends: on one hand there is a huge Internet audience seeking convenient ways to browse through large amounts of digital video. On the other hand, there are a large number of Web publishers and other video content providers who need software to categorize, organize and index their clips and make them easily searchable by their sites' visitors[i].
In order to address the first trend both big search web-companies like Google, Yahoo!, as well as niche video search providers like Blinkx, CastTv, Truveo, have introduced relevant engines aimed to facilitate the video search. While traditional search engines depended on the accuracy of the metadata provided by publishers, the specialized video-search sites use technology that can search the actual content of the video.
These technologies[ii] include:
- Speech-recognition
- Video analysis
- Active tagging
- Speech-to-text
- Facial and optical recognition
- Computer vision
- Semantic categorization[iii
Some sites still rely heavily on human intelligence to be able to identify only high-quality videos, thereby increasing the relevancy of the product they provide to their users. Additionally, these sites have introduced sophisticated filters to enable easy search by type, content provider, relevance/date, popularity, user ratings etc. Moreover, the “community” function identifies consumer profiles, which can be further used to alert them to new content, suggest user groups and provide with recommendations upon the relevant video[iv]. Thus, due to their convenience such sites have become the single point of interest for digital video consumers.
Three major types of video search web-sites can be distinguished:
1. Designated video search engines (VSE). These sites, like Blinkx.com, CastTV, Truveo, do not host any videos, but serve purely as a search engine, allowing for the exploration and search for the specific video across the Internet, through a single gateway, with a high degree of accuracy, offering advanced filtering and sorting options.
2. Video-hosting sites, covering both video-sharing sites, which serve as a platform for uploading of user-generated content (YouTube, BrightCove and Metacafe), as well as internet TV broadcasters (ABC, CBS and Hulu). These sites not only host videos but also provide a search engine to facilitate a search of the hosted video.
3. Traditional Internet search engines such as Google and Yahoo! also introduced video search functions on their sites. Google Videos now serves primarily as a search engine for videos discovered by their search crawlers on other hosting services, in YouTube and user uploads.
Simultaneously, as organizations, including video content providers, produce more and more media, they face the challenge of managing all that data and providing it in a meaningful way to their users. Thus, the relevant technologies are being developed, like matching the publisher’s video with its text articles, suggesting relevant videos to users based on their preferences, categorization of clips in the video library and search engine optimization[v]. Some companies, specialized in video search technologies, provide their product via the Cloud Computing technology. For example, ReelSurfer has launched its Cloud Video Search Platform which allows a search for targeted dialog segments within a larger video and enables customers to implement this technology and avoid a large investment[vi].
The Web 2.0 capabilities that are afforded with video search further draw users in. Users are able to create new wikis for enhancing or commenting on video content. Furthermore, users are able to be the ones to do the video tagging, thereby potentially increasing the level of relevancy. On the other hand, this can also be detrimental to relevancy, as it may be tagged incorrectly. Video search engines are also fairly easy to use. Most search engines simply require its users to type a query into the search box, and the sophisticated technology allows the search engine to scour the web for relevant, and often, helpful information. Sites such as Blinkx have capitalized even further upon the attractiveness of its user-friendliness by displaying search results as 25-second clips of the video. This helps customers further zero in on what they are actually looking for.
As hosts of video content evolve with the video search technology, many other benefits have developed. One such benefit is that users are able to stream videos on their computer, rather than download them, thereby freeing up disk space.
From the perspective of sites with capabilities for video searching, there is an added benefit that allows them to have insight into the user’s mind. They are able to follow what they are looking for, and in return, sell this information to an advertiser. For instance, if someone enters into a search box the keyword “ice skates,” this information can be matched and sold to an advertiser who can provide items surrounding this search to the user. This type of contextual advertising[vii] makes VSE, such as Google and BlinkxTv, have such strong business model.
As the video-hosting sites provide hosting and streaming of the videos they have to cover huge infrastructure costs in order to maintain these services. These costs include the payments for broadband connection (e.g. YouTube bandwidth costs are about $360 million a year[viii]), as well as the purchase of servers and electricity. Additionally, these companies spend a large amount of money to acquire the licenses for professional video (for YouTube it’s about $250 million annually[ix]), as well as to cover the cost of lawsuits (see Risks section). These companies struggle to cover these significant costs due to the fact that user-generated content is not easily monetized.
The video-search dedicated engines have a different cost-structure. Instead of being a repository for video content, they serve as a single entry point to browse across all the video-hosting sites on the Internet. As one of the crucial factors of their success is the provision of the efficient video-search technology, these companies invest a lot into Research and Development. Thus, Blinkx, one of the leaders in this industry, spends $10.6 million annually on R&D, which represents about 34% of their revenues[x]. Other major cost items are Sales and Marketing, Engineering Headcount and other administrative expenses.
As video is a complicated medium, combining speech, images and sounds, the development of video search technology faces some major constraints. While textual features are well defined, video and acoustic features are ill-defined, computationally expensive to extract and bulky to represent[xi]. Queries by image content involve specifying image or video attributes, require the use of an index that is typically derived from the media streams. However, a lot of videos lack detailed metadata, making it hard for search engines to automatically categorize the content. Additionally some videos may be tagged incorrectly (a “tag spam”). These factors decrease significantly the quality of search and relevancy of the content provided.
The challenges of video search as compared to text search are summarized as follows[xii]:
1. Crawling: video is more likely to be accessed via a web application and inaccessible to simple spiders. This is a problem for text search (invisible web) but to a lesser extent. Access to video is more often restricted to resisted users.
2. File formats: Text search engines handle HTML format and perhaps PDF, while video media comes in a wide variety of formats.
3. Link Rot: video files are often too large to be maintained indefinitely by on web servers. News content in particular may be published only for a limited time.
4. Duplication: compared to text, media files are much more likely to appear on multiple sites, and duplication is much more difficult to detect
5. Caching: the size of video media makes caching costly, and caching may violate copyrights.
6. Ranking: referral based ranking may be used, but it is more difficult to implement due to the duplication and web application issues (above). Term frequency ranking methods are typically inapplicable.
7. Browsing: summary generation or hit context extraction is more involved with video, and may introduce copyright issues. The time cost of viewing a document is higher due to buffering delays and the nature of linear media.
All these factors put a lot of constraints on the development of the video search technologies.
Search on video-hosting sites is limited to the content solely provided by the site. Thus the consumers striving to get more objective information prefer to use designated video-search engines. Additionally, video-hosting sites are often criticized for failing to comply with copyright law. For example, some companies have filed lawsuits against YouTube, claiming that it has done too little to prevent the uploading of copyrighted material[xiii]. Besides, some of the hosted content, as well as the user comments, are criticized for being controversial. Given the amount of user-generated content uploaded daily on these sites, it becomes unrealistic to make a proactive review and control of the video and user comments.
Privacy of the users is also a big concern, as web-sites have access to the detailed data about the viewing habits of the users. For example, Apple policy for the iTunes Store is criticized for missing the criteria which would trigger the sharing of personal real-time information with government entities[xiv].
Current Applications of the Technology
VSE Technology has been applied in several fields and industries throughout the development of the technology. As the technology advances and becomes more enhanced, its utility broadens. For instance, when VSE technology came out its main use was to watch videos online, and this was revolutionary at the time (given that only common Search Engines existed). Currently, the technology has evolved so much that its been applied in countless areas. Yovisto[xv], for instance, is a VSE website dedicated exclusively to academic and educational material. It is a VSE that hosts academic videos from professors, students and lecturers around the world. It is interesting that although Yovisto is a hosting VSE and it narrows its business to academic content, it is still regarded by some as one of the top 10 VSE in the market.
One interesting example of VSE technology being applied so far is what the US Military is doing with their state of the art Video Imaging Retrieval and Analysis Tool (VIRAT) [xvi]. The US Military under the Defense Advance Project Agency (DARPA) [xvii] realized that the amount of video footage and data they have stored is massive, and as machines (military airplanes, tanks, video cameras, etc.) become more and more sophisticated, such large amounts of video content will never be seen by human analysts. VIRAT is trying to develop software technology and tools that will analyze the content of the videos, thus making defense and national security more effective[xviii].
The entertainment and television industry has been the driving force of the VSE technology. More and more video content of TV shows, movies, advertisements, sports, etc., are available online (some of them in HD quality), but this content has been hard to find, as it is all scattered around the web. However, VSE technology allows users to easily find TV shows in VSE sites such as Blinkx, Youtube, and Google Video. VSE technology also allows users to search for the very focused, specific content within a video, such as an actor behaving in a specific way.
Of course, the Marketing and Advertisement industry has been intimately link with VSE technology from the beginning. With Video Search Engine Optimization (VSEO)[xix] advancing daily, companies are more enthusiastic with sponsoring specific videos which they might see (based on the VSEO) as a relevant channel to promote their brand or product.
Altavista[xx], which was ultimately acquired by Yahoo.com, was the first Internet search engine to provide Video search results in 2000. By 2004, it was rumored thatxxi] the big search engines - google.com, msn.com, yahoo.com - where developing VSE technologies and were to launch them.
In December 2004, Yahoo.com launched its VSE Beta version[xxii]. It consisted simply of crawling the web to find videos based on the text (metadata) of the video itself. The site only included videos from Yahoo’s paying partners (which included Altavista). Yahoo expressed, at the time of the launch, that they will index relevant links such as XML and RSS. Another reason that Yahoo had for launching their VSE, was that “it is not easy for a web crawler to find downloadable and streaming video content: In many cases, they're hidden behind complex JavaScript, Flash-based players, and other non-crawler friendly obstacles." By asking video producers to supply information about files and their content, Yahoo! is trying to get around that hurdle.
Shortly after Yahoo VSE, Google.com, msn.com and aol.com all followed with different and more enhanced versions of VSE. For instance, Google.com advertised that they would get TV shows and movies videos on their VSE. By this time Blinkx.com was also immersed in the work of launching their VSE and developing their technologies.
By 2004-2005 the internet was evolving from being the World Wide Web, or just a library online, to a media companion of the Television. Not surprisingly, this era was when Internet dial up connections were being updated for faster broadband speed. The Internet Multimedia was just emerging, and the big search engine companies were there to harvest this niche with the VSE technology.
- Public Relations - Agents within the public relations industry can employ this tool to seek out and monitor what is being said about their client. The use of video searching for this purpose is particularly useful in the political arena. The popularity of a presidential candidate can be assessed by adopting this technology and searching for specific keywords attributed to the candidate, his name, as well as his competitor. This assessment will then facilitate in changing a campaign strategy to better address the needs and concerns of voters.
- Consumer Packaged Goods - Consumer packaged goods company- such as Kraft foods - are often interested in the awareness their products yield among consumers. While media agencies are able to provide information on product placement as well as advertisements (magazine, TV, radio, etc), it is not possible to provide comprehensive details of every possible placement. In addition, video search technologies have allowed for the gathering of statistics - how many customers viewed a specific video, what are the demographics of those customers, and what types of comments are they posting. This technology is currently being used by YouTube. (Exhibit 2)
- Retail Industry - Theft is a major concern in the retail industry. Surveillance cameras capture activity, thereby containing a large number of hours of footage. Whether trying to zero in on whom might be to blame for a particular incident, or training new employees to be vigilant of potential thieves, video searching would aide in saving a lot of time not having to go through hours of security footage, and has the potential of avoiding a future incident, thereby saving the company thousands of dollars.
- Medicine - Applying the video search technology to the medical field can have a significant benefit on mankind. A dermatologist, for instance, is confronted from time to time with patients displaying rare skin conditions. They are able to take a picture of this condition and run it through a video search in order to, inevitably, find a lot of targeted videos to help him figure out the prognosis, as well as treatment.
- Legal Industry - Video search can be applied in the legal industry to benefit lawyers, judges, as well individuals outside the industry involved in a legal situation. The technology can be used to search through millions of hours of courtroom footage to focus in on the specific information one is looking for.
Posts
