What is Toluu?
Toluu is a free service for sharing the feeds you read and discovering new ones.		 Get Invite

Iterend Search blog

The official blog about the Iterend search service


Leweb 3.0 in ParisDecember 2

We (and 29 other companies) have been selected to give a short presentation about our company on the leweb conference in Paris on the 9th/10th of december.

As on the Web 2.0 Conference in Berlin, we hope to meet some really interesting people. Christophe and myself will be attending the conference this time.


Iterend at Web 2.0 Expo in BerlinOctober 6

We will be attending the web 2.0 conference in Berlin, the altsearchengines dinner and the latecrunch party thereafter.

Hope to see you all there.


Short downtimeSeptember 25
As we were making a few changes to our blog spider a few days ago, the spider stopped working a little more than 24 hours ago. Unfortunately we didn't notice this. This is also why there was a 20 hours delay when readwriteweb wrote their article (Thanks!).
Normally articles are added between 30 minutes or an hour after they are posted (depending on how often the blog is being fetched).
We are fixing that bug right now. Sorry for the inconvenience. It should be fixed soon and you can access our site normally.


EDIT:
Everything is working normally again and incoming posts are normally added to the search engine.





Searching... or why Summize is often fasterSeptember 23
Searching on our site is an expensive operation:
  1. When you make a query, your query is sent to all the machines which might have matching articles
  2. Based on the list of those articles, our system then creates the related phrases, categories or calculates the sentiment of all the matching articles (this is not yet a public feature) and returns them to the searcher.
Especially the second part is an expensive operation. While the first step "only" needs to distribute the search, map docids and sentenceids (for sentence/post level searching), the second part has to dynamically create lists of keywords (do some calculations and also sort them!) based on all the search results we found in the first step.
This is all being done in a distributed fashion, as you can't have all the information on one machine. Even if the information is distributed over multiple machines, you still have locality problems: If you have seperated your document id space over multiple machines (and you can locally extract a list of keywords), you end up with seperate lists of keywords on multiple machines, which you have to merge. So you have to come up with a clever way of doing this and maybe even change the initial distribution of documents, especially when you want to calculate variations of frequency over time.


Summize (they were independent at first from twit





What makes us different from other search enginesSeptember 16
We have less team members and less servers than our competitors ;). But besides that: