It’s been a great week here at Topsy. This past Tuesday (at 8am!) our new Google+ search went live. Quickly following that, we got a ton of love from news agencies (Forbes, Techcrunch and the Wall Street Journal) and prominent tech bloggers alike. Some of our favorites are Danny Sullivan’s (editor-in-chief, Search Engine Land) post on Google+,
..or Robert Scoble’s concise reaction to the news: “Topsy rocks.”
This means that you can now use Topsy to search through public Google+ posts, which is really exciting given the high-quality of these posts. Google+’s longer format and threaded conversations has allowed it to grow, even in its short lifespan, into a rich forum with in-depth discussion across many topics, with a focus on technology and other technical topics.
Needless to say, the Google+ search results are also ranked by our influence algorithm (just like our Twitter index) to help you find the most relevant results on a topic. And all the other Topsy features are available too, including advanced search options, results grouped into trackbacks, multilingual support, and a listing of experts for any query. You can now see what is trending in both Twitter and Google+ (by choosing one or the other on the left sidebar, along with the other search options).
Richard Lee, our tech lead for the Google+ project, gives us some insight into the effort this involved.
We had already put a lot of thought into diversifying our data sources when the decision was made to build a product around Google Plus data. We knew that our underlying technologies could handle whatever data we were able to get. A far more interesting challenge was imagining how to represent search results in a meaningful way to users.
In the classic Topsy interface, all tweets are rendered the same way. This is because by their nature, all tweets are tweets regardless of how many mentions, hashtags, or retweets they have. Because of that, you’ll find that it’s not the tweets themselves that are most significant in our search results, it’s the external citation that rules. We spent a lot of time making arbitrary links, videos and images show up in a useful way.
In the case of Google Plus, we found that individual posts held a lot of value and served as a point of reference for the tweet-like comments that were associated with them. It was a little like looking at a Twitter conversation and knowing, without guessing, what the conversation was about. This posed a bit of a problem though. Simply displaying a disparate comment from some far flung post on its own would be utterly useless without including a significant portion of the context it was made in. We also could not render a complete context because posts themselves get quite large and we have a very finite amount of space available on a search result page. It was nice to see that in the searchengineland.com coverage of our release, Greg Finn picked up on our efforts, “Topsy’s results featured 6 results above the fold while Google+ managed a little over one-half of a result.” As we learn how to pull more data from the social graph, it will become an increasing challenge to represent higher densities of data concisely without it getting overwhelming.