Question Answering via Integer Programming over Semi-Structured Knowledge

Authors: Daniel Khashabi, Tushar Khot, Ashish Sabharwal, Peter Clark, Oren Etzioni, Dan Roth

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

Reproducibility Variable Result LLM Response
Research Type Experimental On a dataset of real, unseen science questions, our system significantly outperforms (+14%) the best previous attempt at structured reasoning for this task, which used Markov Logic Networks (MLNs).
Researcher Affiliation Collaboration University of Illinois at Urbana-Champaign, IL, U.S.A. {khashab2,danr}@illinois.edu Allen Institute for Artificial Intelligence (AI2), Seattle, WA, U.S.A. {tushark,ashishs,peterc,orene}@allenai.org
Pseudocode No The paper describes its ILP formulation and variables but does not present any pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes Table Corpus and the ILP model are available at allenai.org.
Open Datasets Yes We use the same question set as Clark et al. [2016], which consists of all non-diagram multiple-choice questions from 12 years of the NY Regents 4th Grade Science exams.
Dataset Splits Yes The set is split into 108 development questions and 129 hidden test questions based on the year they appeared in (6 years each).
Hardware Specification Yes For evaluations, we use a 2-core 2.5 GHz Amazon EC2 linux machine with 16 GB RAM.
Software Dependencies No The paper mentions the use of the "open source SCIP engine [Achterberg, 2009]" and refers to "Word Net-based" functions, but it does not specify version numbers for these or other software components.
Experiment Setup Yes Given a question Q, we select the top 7 tables from the Table Corpus using the the standard TF-IDF score of Q with tables treated as bag-of-words documents. For each selected table, we choose the 20 rows that overlap with Q the most.