felipemoraes/dynamic

Code for my Master's thesis and ICTIR 2017 paper:"On effective dynamic search in specialized domains'

dynamic's Issues

Offline evaluation and Cube Test metric

Implement offline evaluation in Python. The folder with the Python scripts will be called eval.

Elaborate more Term Features, features dependent of query

Dynamic Diversification with Multiple Information Sources using CA

After completing #6 and I will run experiments with CA using the same algorithm to reformulate a sub-query for each feedback passage related to a sub-topic.

Refactor code for Online Learning Models

Implement Coordinate Ascent

Implement Coordinate Ascent as described in Donald Metzler thesis.

https://pdfs.semanticscholar.org/02f5/5a795ba82ae0064cba98ae9bf0e3843c5fe2.pdf

Iterative Query Reformulation Model with Multiple Information Sources.

In this model for Dynamic Search, I will estimate the parameters with CA algorithm.
At each iteration we reformulate the original query and add more terms that appear on the feedback passages for that query.
As Query Reformulation from Multiple Resources Model, I will test the one proposed in the following papers:

• Effective query formulation with multiple information sources. Michael Bendersky, Donald Metzler, and W. Bruce Croft. In Proc. of WSDM 2012.
• Parameterized concept weighting in verbose queries. Michael Bendersky, Donald Metzler, and W. Bruce Croft. In Proc. of SIGIR 2011.

Dueling Bandits Gradient Discent