Spock: People Search With A Man + Machine Approach
Last week, I wrote about the overall people search landscape, which, while not new, has been recently expanding and making full use of all that is shiny Web 2.0. Spock is the latest of these services to leverage social networks, intelligent web crawling, and community involvement. Spock launched in private beta in April, then in public beta in August.
With Spock, you can search on a person’s name or a keyword that may be associated with a person, and are returned a list of people with associated tags, photos, and web sites. From there, you can drill into more information about any of those aspects, add information of your own, or browse to people who are related in various ways.
Spock CEO Jaideep Singh says that their crawling and indexing infrastructure has the unique ability to identify people-specific data on web pages and extract only this information. Their algorithms combine natural language processing with machine learning and they augment this with human involvement—both community input and editorial oversight.
Tagging provides a unique twist on the search for people. As Singh points out, your search can turn into a discovery. You may initially search for information on Kristen Dunst, but then might follow the tags to see others associated with Bring It On and notice that several of those listed are also tagged with Buffy the Vampire Slayer, including Christophe Beck, who choreographed the music for both the awesome cheerleading movie and the kick ass TV show about the blonde girl with the pointy stick. Click “related people” for a list from another perspective.
In the case of Kristen Dunst, this gives you mostly a list of ex-boyfriends rather than costars. You view people in a larger context and can gain more insight with this type of system. You can also search directly over tags. For instance, a search for “Buffy fan” returns, well among other people, me. Singh says that only half of people-related searches are for names and the rest are topical, so having a way to categorize people into a variety of topics can be very useful.
Anyone can add a tag to anyone else. Spock is confident that the ability of the community to vote on tags will bring a Digg-like democracy to the results and will present an accurate picture of not only how people view themselves, but how the web views them as well. Singh notes, for instance, that while Google created an algorithm to keep George Bush from being returned as a top result for “miserable failure”, that’s a relevant result for Spock because it reflects the views of the community. (It’s currently the second tag listed for Bush.)
But tagging has not been without criticism. A Digg-like democracy can give power to the collective people or it can be a ripe environment for gaming, and in typical web fashion, that has already begun. Singh claims the malicious use of tagging has been extremely low and notes that anyone can request removal of a tag once they’ve claimed their profile. And then there’s Spock Power, which gives more or less weight to votes based on a person’s history (how often contributions have been voted up or down). They also point to their transparency. Since everyone can see how everyone else tags and votes, I know exactly who’s tagging me as a Buffy fan.
Spock is incorporating other community aspects as well. Once you’ve created an account and claimed your profile, you can add friends and mark people as favorites. It seems that Spock marks favorites for you as well based on matches it finds from your connections on social networks such as LinkedIn. You can also import your address books and start building up a full network of contacts. You can add information to each profile that only you can see, such as phone numbers and notes. With features like this, you can see a bit of Spock’s roots as the ultimate contact management system (after the founders become frustrated with the limitations of Outlook). Since Spock can be a combination of contact information you add about people and the information that Spock finds about them from the web, you can potentially do more useful searches over those contacts. Rather than just search by name, you can look for all your contacts who like to play golf or are experts in link building.
Creating an account and registering your profile not only enables you to take part in the community aspects of the site, but it allows you to engage in a bit of reputation management as well. Information that you add about yourself is weighted more heavily than information others add about you. Also, you can sign up for alerts about when others add information to your profile.
How relevant are the results? Spock is going after quality over comprehensiveness and are slowly expanding. They don’t include offline data, but are ambitiously aiming to crawl the entire web and extract all useful people-related information. They’re not there yet. Flickr photos are visibly missing, for instance, but they’re working on adding more data sources over time.
Singh acknowledges that extraction and aggregation are hard problems. They feel they have an advantage over Google regarding extraction because while Google is agnostic to the page type, Spock tries to identify pages and information specifically about people and then processes over that. Aggregation requires that you not only can classify the data as people-related, but can identify when data from disparate locations is about the same person. So far, Spock seems to be primarily concentrating on grouping profiles from social networks and adding links from blogs, news stories, and sites like Wikipedia.
They feel their “man plus machine” philosophy is a scalable and effective way to combine smart approaches to algorithmic classifications with crowdsourcing. While this multi-faced approach seems promising, they certainly have an uphill battle. Can they engage the community to add valuable information? Can they become comprehensive—both in the total number of people they profile and in the information about each person? Can they introduce a paradigm shift around where people search and how people manage their contacts that will trigger a move away from Google and Outlook for people-related data? And maybe most importantly, will all of these plays give them an edge over the other people search services in the space?
All of that remains to be seen, but certainly they’ll be expanding in these areas over time, so they’re well worth watching. Stay tuned in the coming weeks for the rundown on some of the other services and their approaches to winning the battle of the people search vertical.
Opinions expressed in the article are those of the guest author and not necessarily Search Engine Land.
Get the best search news, tips and resources, delivered each day.