Compositional Language Understanding with
Text-based Relational Reasoning
Koustuv Sinha
∗
Mila
McGill University, Canada
koustuv.sinha@mail.mcgill.ca
Shagun Sodhani
Mila
Université de Montréal, Canada
sshagunsodhani@gmail.com
William L. Hamilton
Mila
McGill University, Canada
Facebook AI Research (FAIR), Montreal
will.leif.hamilton@gmail.com
Joelle Pineau
Mila
McGill University, Canada
Facebook AI Research (FAIR), Montreal
jpineau@cs.mcgill.ca
Abstract
Neural networks for natural language reasoning have largely focused on extractive,
fact-based question-answering (QA) and common-sense inference. However, it is
also crucial to understand the extent to which neural networks can perform rela-
tional reasoning and combinatorial generalization from natural language—abilities
that are often obscured by annotation artifacts and the dominance of language
modeling in standard QA benchmarks. In this work, we present a novel bench-
mark dataset for language understanding that isolates performance on relational
reasoning. We also present a neural message-passing baseline and show that this
model, which incorporates a relational inductive bias, is superior at combinatorial
generalization compared to a traditional recurrent neural network approach.
1 Introduction
Neural language understanding systems have been extremely successful at information extraction
tasks, such as question answering (QA). An array of existing datasets are available, which test
a system’s ability to extract factual answers text [
1
–
5
], as well as datasets that emphasize simple,
commonsense inference (e.g., entailment between sentences) [
6
,
7
]. However, it is difficult to evaluate
a model’s reasoning ability in isolation using existing datasets. Most datasets combine several chal-
lenges of language processing into one, such as co-reference / entity resolution, incorporating world
knowledge, and semantic parsing. Moreover, the state-of-the-art on all these existing benchmarks
relies heavily on large, pre-trained language models [
8
,
9
], highlighting that the primary difficulty in
these datasets is incorporating the statistics of natural language, rather than reasoning.
In this work, we see to directly evaluate and innovate on the compositional reasoning ability of
a QA system. Inspired by CLEVR [
10
]—a synthetic computer vision dataset that isolates the
challenges of relational reasoning—we propose a text based dataset for Compositional Language
Understanding with Text-based Relational Reasoning (CLUTRR). Our initial version, CLUTTR v0.1,
requires reasoning and generalizing about kinship relationships, and we plan to use our proposed data
generation pipeline to extend the set of tasks in the future. We develop and evaluate strong baselines
on CLUTTR v0.1, including a recurrent LSTM model and a message-passing graph neural network
∗
Work done while being an intern at Samsung Advanced Institute of Technology (SAIT), Montreal
Proceeedings of Relational Representation Learning Workshop, 32nd Conference on Neural Information Pro-
cessing Systems (NIPS 2018), Montréal, Canada.
arXiv:1811.02959v2 [cs.CL] 8 Nov 2018