Open postdoctoral position on Controlled randomness in semantic spreading activation

Context


This Post-Doc position is based on the results of a joint research project between Alcatel Bell Lucent Labs and Inria
and more precisely the Ph.D. thesis of Nicolas Marie in Wimmics that successfully proposed a new method to design
exploratory search engines over linked data on the Web.

Exploratory search refers to cognitive consuming search tasks that are open-ended, multi-faceted, and iterative like
learning or topic investigation. In parallel, semantic web and linked data offer new possibilities to solve complex
search queries and information needs including exploratory search ones. In this context the linked open data cloud
plays an important role by allowing advanced data processing and innovative interactions model elaboration.

We proposed a linked data based exploratory search based on a spreading activation algorithm and proposed new
diffusion formula exploiting the richness of typed graph. Starting from this formalization we proposed additional formalizations of several advanced querying modes in order to solve complex exploratory search needs. The result was
implemented and evaluated in the Discovery Hub (discoverhub.com) web application that retrieves the results and
present them in an interface optimized for exploration.

Job


In the evaluation of this work, two needs for controlled randomness where identified and they form the core research
questions of this post-doc subject:

(1) the current spreading activation is achieved through a greedy incremental sampling algorithm that retrieves subgraphs
of remote linked data sources to locally implement a semantic spreading activation. Random walks biased by
the types of arcs and resources visited could provide an alternative.

(2) the current approach to increase serendipity in exploratory search is a too simple random choice introduced on
structural criteria when exploring the graph. The result show that the impact on the relevance of results is too high and
that it makes it difficult to explain the results. There are other places where randomness could be used, for instance in
the choice of links types or resource categories during the propagation.

The objective of this post-doc is to propose and evaluate several methods relying on controlled randomness to improve
the performances of semantic spreading activation both in terms of performances and usability. The candidate will also
explore how to support new types of queries including negative search to eliminate some aspects from the results.

Profile


Ph.D. in computer science, knowledge of linked data, semantic web formalisms, graph formalisms and algorithms.

Details


Duration : 16 months
Salary : 2 621 ~ gross/month
Monthly salary after taxes : around 2 128 euros including health insurances

Business restaurant

Possibility of French courses

Scientific Resident card

Scientific advisor : Fabien Gandon fabien.gandon@inria.fr

The candidates are invited to contact the scientific advisor before applying, an preferably before april 1st.

Security and defense procedure
In the interests of protecting its scientific and technological assets, Inria is a restricted-access establishment. Consequently, it follows special regulations for welcoming any person who wishes to work with the institute. The final acceptance of each candidate thus depends on applying this security and defense procedure.

Warning
Applications must be submitted online on the Inria website. Processing applications submitted by other channels is not guaranteed.

APPLY ONLINE NOW