QUAREL: A Dataset and Models for Answering Questions about Qualitative Relationships

Authors: Oyvind Tafjord, Peter Clark, Matt Gardner, Wen-tau Yih, Ashish Sabharwal7063-7071

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present QUAREL, a dataset of diverse story questions involving qualitative relationships that characterize these challenges, and techniques that begin to address them. ... We contribute ... (3) two novel models for this task, built as extensions of type-constrained semantic parsing. The first of these models (called QUASP+) significantly outperforms off-the-shelf tools on QUAREL. The second (QUASP+ZERO) demonstrates zero-shot capability... The dataset and models are available at http://data.allenai.org/quarel.
Researcher Affiliation Industry Oyvind Tafjord, Peter Clark, Matt Gardner, Wen-tau Yih, Ashish Sabharwal Allen Institute for AI, Seattle, WA {oyvindt,peterc,mattg,scottyih,ashishs}@allenai.org
Pseudocode No The paper describes model architectures and processes, but it does not include a dedicated pseudocode block or algorithm labeled as such.
Open Source Code Yes The dataset and models are available at http://data.allenai.org/quarel.
Open Datasets Yes We present QUAREL, a dataset of diverse story questions involving qualitative relationships that characterize these challenges, and techniques that begin to address them. ... The dataset and models are available at http://data.allenai.org/quarel.
Dataset Splits Yes Table 1: Summary statistics for the QUAREL dataset. ... # questions train/dev/test 1941/278/552
Hardware Specification No The paper describes the models and training process but does not specify any particular hardware used for running the experiments.
Software Dependencies No The paper mentions software components like Allen NLP, LSTMs, Glove, and ELMo, but does not provide specific version numbers for any of these dependencies.
Experiment Setup No The paper describes the training objective and general setup (e.g., using beam search and specific embeddings) but does not provide concrete hyperparameters such as learning rate, batch size, or number of epochs.