How long does it take to do a Google News for B2B?

written by Bharath Mohan - View Comments

Google News rocks. It has the ability to collate related stories together, giving orthogonal perspectives on the same story. How many people have wanted to do something like Google News for their own verticals? How many days would it take to do something like that?

Today, Channel Tech Center launched a curated portal for B2B industry news, ChannelToday.

Take a look at that page. Its fully integrated into their site. Reflects their look and feel. Has their branding. It curates industry news on B2B. Every article is tagged with relevant concepts that drive the article, along with related news. Very much like Google News. How many days do you think this took for Joseff Betancourt?

2!

Yes, 2. Joseff approached us and said he loved our technology and what it can do. We piloted our content discovery widget on their site. Then powered rich hub pages on ChannelTechNews. He replaced Google Custom Search with ours. Our products were being put to good use. But he was still hungry. He wanted to have a curated portal of all B2B news from the channel network. He wanted to know if we can help him do this.

Busy in the past week, we just dropped him a link to our API documentation. We allow many ways to ingest relevant data (articles) into a document collection. Given a set of articles, we could bubble up the most relevant concepts. Given any articles we could mine other relevant articles, and the concepts that drive them. Maybe this can help get some distance? What ensued was some mashery, and some support from our side. Joseff wrote a cute PHP wrapper to our API, and used it to create his own Google News. His algorithm was pretty simple:

1. Pull up all the data he cares about into a “session”. A session is our way of calling a document collection.
2. Our text mining/ content discovery engine has already started working. Its pulled articles, indexed and understood them.
3. He just asks for the list of articles, sorted by time. He gets a JSON array for this that can be iterated through.
4. For every article in this list, he asks for related articles, and the top concepts that matter. Another JSON array follows.
5. Now its just about rendering the page based on all the information he got. To ensure good response times, he caches the page and hits our server once every 10 minutes for updates.

Pretty simple you’d say. Why did he take 2 days for this? Well, this includes the initial learning curve of our API, and its nuances – we’ll get better at this over time.

Is Joseff satisfied now? Not yet. He wants source ranking, the ability to train concepts, etc etc. He’s keeping us busy. But boy! Are n’t we happy?