The KryGle Project

Title: “Introduction to Network Mathematics”
Author: Bruce Hoppe, Professor at the Boston University
Publish Date: Live online document from spring 2007

This is the first paper that purely fit in the Math’s category. Basically its content is a bunch of mathematical concepts well organisez and easy to understand by not purely math-people (like me .

First parts of the document can be a bit boring as it takes a quite big walk around basic concepts on sets, graph’s theory, trees, … But there are some interesting things specially in the graph’s section. We are used to be explained maths as a purely theorical thing, that’s really boring. Instead, in this document you’ll find real approaches where solutions based on this theory are explained and that’s something to be thanked of.

One of my faovourite sections in this document is the one dedicated to the concept of Structural Equivalence, I do really recommend it if you’re interested in understanding the basics on social networks and recommendation systems. Also is interesting when the author explains how bipartite graphs can be really useful when working in a tags-based system.

Another important section you’ll find here is about the way the Google’s PageRank does its work. Maybe not so deeply that other document may do, but enough to understand the basics and getting conscience of the complexity it has, something that makes you think about how to face up to a highly recursively problem. It uses as an example the HopRank, a simpler PageRank concept for better understanding.

Once again, the big point you’ll find in this document is the amount of real and easy to understand examples against the minimun theory concepts necessary to understand them.

It seems to be a good idea writting a post per paper read, just to highlitght important ideas and make a brief (or not so brief) about my thoughts and the way it can apply on my research. Having this in mind, here’s the first post about papers.

Title: “Social Search and Discovery Using a Unified Approach”
Author/s: Einat Amitay, David Carmel, Nadav Har’El, Shila Ofek-Koifman, Aya Soffer, Sivan Yogev, Nadav Golbandi, from the IBM Research Lab in Haifa, Israel.
Publishing Date: Friday, April 24, 2009

This papers describes a research and some experiments done in the IBMs intranet searching engine. It was tried to improve the search results using some social approach. The idea is similar to the one who motivates me but there are some important differences:

• They worked out some social improvements in a closed environment, and intranet. My idea is to open krygle wildly to the internet or, better said, to the googlenet as krygle will work with results obtained from Google.
• They worked on a well-known model, this is, they built up a model of the items they would take care of. As the intranet was built on blogs and social bookmarks, they defined a model where blog entries, comments, tags, bookmars and people where the elements used. Here they stabilsh some relationship that are quite easy to see: a blog entry is more relevant as more activity (comments, bookmarks, tag) gets. People is more relevant depending on how much relevant blogs entries do they have, and so on… I liked the model: easy, simple and they proved it works. But, taking care on the whole internet you need a more opened model taking in count the services you want to take care of.

In this paper, I’ve seen some thing I need to have a look at, for example, they talk about already known algorithms I’ve never heard about, that’s why I’ll have to make some diagonal reading on the references in the papers. Pointing out the most important ones, I’ll have a look at these:

• Faceted Search
• Normalized Discounted Cumulative Gain (aka NDCG)
• Normlized Google Distance (measures the similarity of meaning between terms)

All this added information from a two-pages-paper… don’t want to thing about the addition readign I’ll get from the 32-pages one waitting at the bottom of the paper’s pile… xD

Ok, I’m a noob talking about searching technologies. Maybe that’s one of the points I like the most about working in this field. It’s kind of a “fresh start”, I’m here, I have the tools, the idea and the motivation so… what’s missing? It can be said in one work: READING.

Looking for information is quite easy sometimes, you just need to google it and you’ll get plenty of information, but… all kind of information: good, poor, wrong, … When you are researching you need contrasted information, usually published as papers in workshops.

I’m lucky to have some friends used to work in the research field. Counting whith Jose Carlos Cortizo (the perfect guy for a mid-afternoon coffee) saved me a lot of time, he knows were to look and has worked in social search as a part of Wipley so, who could point me in the right direction better than him? These the papers he recommended me to get the basics on search technologies and social search in particular:

• A Survey om Web Information Retrieval Technologies
• Deeper Inside the PageRank
• Towards a Model of Understanding Social Search
• Social Search in Small-World Experiments
• Social Search and Discovery Using an Unified Approach
• Social Search – Exploring and Searching Social Architectures in Digital Networks

With this papers I’ll get the basics on the technology side but there is another field that has to be covered, the maths. I’m not a math-lover but graph’s theory and statictics are not so bad. Here I’ve gotten help from another person, Regino Criado, the director of the Applied Maths Department at the Rey Juan Carlos University who recommended me this starters to see if I run away or keep in the edge:

• Centrality and Network Flow
• Introduction to Network Mathematics

These days I’ll be working with these papers (now I’m finishing my reading on Social Search and Discovery Using a Unified Approach) and writting some posts about them. One thing I’ve noticed when reading that first paper is that papers seem to be recursive, almost all bibliography entries are worth at least a quick look so… more reading!

The University Libre Software Contest (aka Concurso Universitario de Software Libre, CUSL) is been held again this year, as they have been doing for the last three years.

I have already taken part in this contest before. Two years ago I was part of the team who developed the R4P Project, we were very proud of being in the finals and being the winners in the innovation category. My colleague in that project (Luis Ignacio Díaz del Dedo) is taking part this year too, with his investigation project Open Virtual Object Viewer which seems to have a great future! Good luck from here pal!

As you can see in this post’s title, I’m taking part in the contest this year with the KryGle Project. These year seems to be much more competition, and that’s great, as the point of the contest is to promote libre software development. This makes much more difficult to be on the finals again, but that’s not the target of KryGle, don’t misunderstand, it would be great, but this project is an investigation one, and that means that implementation is less important that researching.

Anyway, I’ll do my best for making KryGle a good project, and at least some implementation has to be done as it’s necessary to experiment with search results from different people.

KryGle is the codename of an investigation project that aims to improve the way search engines ponderates its results taking in count the social factor in the web.

Everyday social relationships on the web get more and more important. Some years ago information was spread from search engines and, in some cases, freed readers. Now there are lots of points spreading information. We can see how some social services such as twitter became into important information providers. The point is, what’s more important? Comments on a blog post or retweets?We’ll try to figure it out.

Also, we can say that we are connected to people who are interested n subject we are interested in too. Have a look to this… wouldn’t it be great that a search engine would take in count people you are connected to in order to give you the best search results? People used to like what they friends like, so this idea isn’t so bad.

The point here is to build a search engine based on Google. We’ll take Google’s results and reponderate its results according to people you are connected to, and see if the reponderation applied to Google’s results make a better way for you to find the info you’re looking for.

We’ll need your help so… stay tunned!