Inferring Interpersonal Relations in Narrative Summaries

Authors: Shashank Srivastava, Snigdha Chaturvedi, Tom Mitchell

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental On a dataset of movie summaries from Wikipedia, our structured models provide more than a 30% error-reduction over a competitive baseline that considers pairs of characters in isolation.
Researcher Affiliation Academia 1Carnegie Mellon University 2University of Maryland, College Park ssrivastava@cmu.edu, snigdhac@cs.umd.edu, tom.mitchell@cmu.edu
Pseudocode Yes Algorithm 1 Perceptron Training for Relations and Algorithm 2 Narrative-specific Model
Open Source Code No The paper states that the dataset is made publicly available, but there is no explicit statement or link indicating that the source code for the described methodology is available.
Open Datasets Yes We processed the CMU Movie Summary corpus, a collection of movie plot summaries from Wikipedia, along with aligned meta-data (Bamman, O Connor, and Smith 2013); and set up an online annotation task using BRAT (Stenetorp et al. 2012). ... The dataset is made publicly available for research on the first author s webpage.
Dataset Splits Yes Figure 4 shows the cross-validation performance of major feature families of text features on the training set. and We tune values of hyperparameters... through cross validation on training data.
Hardware Specification No The paper does not provide any specific hardware details used for running the experiments.
Software Dependencies No The paper mentions software tools such as "Stanford Core-NLP system", "Semafor parser", and "BRAT" but does not provide specific version numbers for any of them.
Experiment Setup Yes We tune values of hyperparameters, i.e. number of training epochs for the structured perceptron (10), the weighting parameter for the clustering model (α=0.8), and the number of clusters (K=2) through cross validation on training data. For the structured models, reported results are averages over 10 initializations.