Tools
Generic and Trend-aware Curriculum Learning for Relation Extraction in Graph Neural Networks. Graph Text Neural Network (GTNN) is a generic and trend-aware curriculum learning approach that effectively integrates textual and structural information in text graphs for relation extraction between entities, which we consider as node pairs in graphs. The proposed model extends existing curriculum learning approaches by incorporating sample-level loss trends to better discriminate easier from harder samples and schedule them for training.
Attentive Multiveiw Neural Network (AMNN) Model. This is framework that can combine different views (representations) of the same input through effective data fusion and attention strategies for ranking purposes. We developed the model to find the most probable diseases that match with clinical descriptions of patients, using data from the Undiagnosed Diseases Network.
RbF Curriculum Learner. This tool demonstrates spaced repetition for training any neural network. RbF is inspired by the broad evidence in psychology that shows human ability to retain information improves with repeated exposure and exponentially decays with delay since last exposure. RBF works based on spaced repetition in which training instances are repeatedly presented to the network on a schedule determined by a spaced repetition algorithm. RbF shortens or lengthens review intervals for training instances with respect to a few indicators.
Leitner system. Implementation of Leitner system for simultaneously training a given neural network and identifying spurious instances (those with wrong labels) in its input dataset.
Twitter Crawler. A crawler for searching on Twitter and obtaining tweets and user networks.
Clinical Decisions. A demo of decision detection in clinical discharge notes.
Datasets
Gene, Disease, Phenotype Relation (GDPR) dataset: This dataset is obtained by combining and linking entities across two freely-available datasets: Online Mendelian Inheritance in Man (OMIM, https://omim.org/) and Human Phenotype Ontology (HPO, https://hpo.jax.org/). The dataset contains relations between genes, diseases and phenotypes. See this paper (Vakil et al., NAACL 2022) for the details.
Churn Dataset: The dataset contains labeled tweets about three telco brands: Verizon, AT&T, and T-Mobile. Tweet are labeled as churny or not-churny, where churny tweets indicate a high risk of canceling the brand's service. Labels are obtained through crowdsourcing and each tweet is labeled by at least three annotators. Fleiss’ kappa is 0.62, which indicates substantial agreement among annotators. Cohen's kappa computed over 1073 instances related to T-Mobile, that were independently labeled by one of our team members and compared against the aggregation of the three annotators judgments over these instances, is 0.93, which indicates substantial annotation agreement as well.