Rene Pickhardt

With related-work.net we introduce a scientific platform that is using neo4j to store an open citation graph of research articles and build a social networking application on top of it. We use Google Web Toolkit enhanced by the Model View Presenter framework GWTP.

Related-Work offers the following core functionality (http://www.rene-pickhardt.de/related-work-net-product-requirement-document-released/):

Convenient browsing of scientific literature using

recommendations (personalized) search functions ** (personalized) alerts for new publications

With related-work.net we introduce a scientific platform that is using neo4j to store an open citation graph of research articles and build a social networking application on top of it. We use Google Web Toolkit enhanced by the Model View Presenter framework GWTP.

Related-Work offers the following core functionality (http://www.rene-pickhardt.de/related-work-net-product-requirement-document-released/):

Convenient browsing of scientific literature using

recommendations (personalized) search functions ** (personalized) alerts for new publications.

Enhance scientific communication with Q&A features inspired by StackOverflow.

In this talk we show that graph databases are the natural technology to power not only a social networking application but are the best choice for a question and answer system.

Topics of the talk include:

Design and implementation of a discussion system on top of a social network application using a graph database. Why does it scale better than relational approaches? Why is it more flexible than a document store? Techniques for storing big graphs in neo4j and performing data mining tasks Benchmarks comparing neo4j’s core API to Cypher Query language. How to build personalized graph based search indices Performance boosts by smart client side caching for local graph queries. Currently only a preliminary python based demo is online at dev.related-work.net but we are working hard to get our +10’000 lines of code deployed in the beginning of 2013. That will already include a basic Q & A System as well as various data mining techniques on graphs, recommendation engines and personalized search and auto completion. The source code can be found at: https://github.com/renepickhardt/related-work.net

This is joint work with Heinrich Hartmann (Koblenz), David Shotton (Oxford) and Sun Xinruo (Shanghai Jiao Tong University).

You can find more information about it in my blog article: http://www.rene-pickhardt.de/graphity-an-efficient-graph-model-for-retrieving-the-top-k-news-feeds-for-users-in-social-networks/ and a demo can be found at http://gwt.metalcon.de/GWT-Modelling/#Home A key challenge of web platforms like social networking sites and services for news feed aggregation is the efficient and targeted distribution of new content items to the users. This can be formulated as the problem of retrieving the top-k news items out of the d-degree ego network of each given user, where the set of all users producing feeds is of size n, with n ≫ d ≫ k and typically k < 20. Existing approaches employ either expensive join operations on global indices or suffer from high redundancy through denormalization. This makes retrieving of different top-k news item sets for thousands of users per second very inefficient in a large network. In this paper, we propose a novel index GRAPHITY to remedy this problem. The GRAPHITY index is based on neo4j and allows retrieval of the top-k news items from a user’s ego network independent of the node degree of his ego network, thus reducing join efforts. In addition to efficient top-k retrieval, GRAPHITY does not introduce any redundancies in the content data.We prove that updates of the index are in O(1) and adding a new content item is linear in the user’s indegree. We compare the theoretical run time complexity of GRAPHITY against several baselines on two data sets of different characteristics and size. Our evaluation confirms that GRAPHITY is independent of the node degree d and network size n. In addition, GRAPHITY significantly outperforms the baselines in retrieving the news feeds as our theoretical examination predicted.