Date(s) - 20/03/2014
2:00 pm - 6:00 pm
Category(ies) No Categories
This month, the Biomedical Research Computing Forum will play host to an afternoon of speakers covering topics related to information extraction. All welcome.
Richard Jackson [14:00-14:45]
Text Mining in the BRC-MH – past, present and future
Natural Language Processing is a fundamental tool to generate data for many applications in clinical informatics. This talk explores the BRC’s information extraction approach for common variables of interest, and how machine learning is employed to target concepts specific to researcher questions.
Richard Jackson is the Text Mining Lead for the BRC-MH CRIS project.
Angus Roberts [14:45-15:30]
Using semantic spaces for text categorisation
The distributional hypothesis states that the contexts in which words appear correlate with their meaning. This talk explores the use of distributional semantics to represent the “semantic space” of words as high dimensional feature vectors, and the construction of text classifiers from these. The talk will present two simple examples of building semantic spaces over text collections, using these to train classifiers capable of categorising documents and sentences.
Angus Roberts is a Senior Researcher in the Natural Language Processing Group, University of Sheffield, and leads life science work for GATE, a widely used language engineering tool-kit.
Zina Ibrahim [15:45-16:30]
Towards Using Rule-based Multi-agent Systems for the Early Detection of Adverse Drug Reactions
Adverse Drug Reactions (ADRs) represent troublesome and potentially fatal side effects of medication treatment. To address the burden induced by ADRs, a preventive approach is necessary whereby clinicians are provided with new data-driven decision-support systems to foresee the factors leading to ADRs and plan precautionary activities effectively. We present the first phase of a multi-agent system which monitors the factors leading to the onset of ADRs using information found in the patient records in a hospital setting. The system uses a fuzzy rule-based reasoning engine utilizing decision rules recommended by experts.
Zina Ibrahim [16:30-17:15]
Towards Category-Driven Rule Association Mining
The quality of rules generated by ontology-driven rule association mining algorithms is highly constrained by the algorithm’s effectiveness in exploiting the usually large ontology in the mining process. We present a framework built around superimposing a structure on a given ontology, enabling the division of the rule mining problem into disjoint subproblems which can be solved individually. We present a new metric for evaluating the rule’s interestingness based on where its constructs fall within the ontology. Our metric is anti-monotonic with respect to subsets, making it usable in an apriori-like algorithm which we present here. The algorithm categorises the ontology into disjoint subsets and uses the metric to find associations in each, joining the results iteratively using its anti-monotonicity as a guide. The framework includes built-in definition of user-specified filters which can be optionally embedded in the algorithm reflecting user preferences.