Refer-to-as Relations as Semantic Knowledge

Authors: Song Feng, Sujith Ravi, Ravi Kumar, Polina Kuznetsova, Wei Liu, Alexander Berg, Tamara Berg, Yejin Choi

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that the proposed approaches outperform competitive baselines, confirming our hypothesis to view entrylevel categorization as a collective inference problem over lexical structure knowledge.
Researcher Affiliation Collaboration Song Feng IBM T. J. Watson Research Center & Stony Brook University sfeng@us.ibm.com Sujith Ravi, Ravi Kumar Google Mountain View, CA sravi@google.com ravi.k53@gmail.com Polina Kuznetsova Computer Science Department Stony Brook University polina.sbu@gmail.com Wei Liu, Alexander C. Berg, Tamara L. Berg Computer Science Department University of North Carolina at Chapel Hill {wliu, aberg, tlberg}@cs.unc.edu Yejin Choi Computer Science & Engineering University of Washington yejin@cs.washington.edu
Pseudocode No The paper describes algorithms (ILP, min-cost flow) and provides a diagram for the min-cost flow formulation, but does not include structured pseudocode or an explicit algorithm block.
Open Source Code No We are making the resulting entry-level categories and visual similarities publicly available at http://homes.cs.washington.edu/ yejin/refer2as/. This link provides data, not explicitly the source code for the methodology.
Open Datasets Yes We use Word Net (Fellbaum 1998) as the encyclopedia knowledge base. We use word frequencies derived from Google Ngram 1T data (Brants and Franz. 2006) ... using the Image Net dataset (Deng et al. 2009) ... Our contributions include a new labeled data set, the collective inference and optimization approach, and the computed mappings and similarities.
Dataset Splits No The paper mentions 'cross-validated performance' and 'development dataset' but does not provide specific numerical percentages or counts for training, validation, and test splits needed for reproduction.
Hardware Specification No The paper mentions processing a large number of images and using a convolutional neural network (Caffe) but does not provide specific hardware details (e.g., CPU/GPU models, memory) used for experiments.
Software Dependencies No The paper mentions 'open source Caffe (Jia 2013)' and 'CPLEX (2006)' but does not specify exact version numbers for these software dependencies or other libraries.
Experiment Setup Yes We use θ = 0.5, a = 2, k = 2.