A Parallel-Hierarchical Model for Machine Comprehension on Sparse Data

By Adam Trischler, Zheng Ye, Xingdi Yuan, Jing He, Philip Bachman. 2016

47 citations recorded by the source corpus.

Understanding unstructured text is a major goal within natural language processing. Comprehension tests pose questions based on short text passages to evaluate such understanding. In this work, we investigate machine comprehension on the challenging {\it MCTest} benchmark. Partly because of its limited size, prior work on {\it MCTest} has focused mainly on engineering better features. We tackle the dataset with a neural approach, harnessing simple neural networks arranged in a parallel hierarchy. The parallel hierarchy enables our model to compare the passage, question, and answer from a variety of trainable perspectives, as opposed to using a manually designed, rigid feature set. Perspectives range from the word level to sentence fragments to sequences of sentences; the networks operate only on word-embedding representations of text. When trained with a methodology designed to help cope with limited training data, our Parallel-Hierarchical model sets a new state of the art for {\it MCTest}, outperforming previous feature-engineered approaches slightly and previous neural approaches by a significant margin (over 15\% absolute).

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

Citations in the corpus

NewsQA: A Machine Comprehension Dataset (2016, methodology): “Nevertheless, recent comprehension models have performed well on MCTest (Sachan et al., 2015; Wang et al., 2015), including a highly structured neural model (Trischler et al., 2016a).”
Option Comparison Network for Multiple-choice Reading Comprehension (2019, methodology): “dn’t want to visit the factory any more. C. The writer had visited the factory before. D. The writer couldn’t wait to visit the factory. Answer: D Table 1: An MCRC examplefrom the RACE dataset. 2016; Trischler et al., 2016; Dhingra et al., 2017; Tay et al., 2018; Parikh et al., 2018; Zhu et al., 201”
End-to-End Answer Chunk Extraction and Ranking for Reading Comprehension (2016, methodology): “EpiReader (Trischler et al. 2016b) adopted a joint training model for answer extractor and reasoner, where the extractor proposes top candidates, and the reasoner weighs each candidate by examining entailment relationship between question-answer representation and the document.”
A Co-Matching Model for Multi-choice Reading Comprehension (2018, methodology): “Inspired by some previous work (Wang and Jiang, 2016; Trischler et al., 2016), we ﬁrst use bi-directional LSTMs (Hochre-iter and Schmidhuber, 1997) to pre-process the sequences as follows: where H p ∈ R l × P , H q ∈ R l × Q and H a ∈ R l × A are the sequences of hidden states generated by the…”
Universal Natural Language Processing with Limited Annotations: Try Few-shot Textual Entailment as a Start (2020, background): “Deep learning has not achieved signiﬁcant success on it because of the limited training data (Trischler et al., 2016)—this is exactly our motivation that applying few-shot textual entailment to handle annotation-scarce NLP problems.”
Using Natural Language Relations between Answer Choices for Machine Comprehension (2019, background): “Trischler et al. (2016b), Kumar et al. (2016), Liu and Perez (2017), Min et al. (2018) and Xiong et al. (2016) proposed successful models on those datasets.”
A Compare-Aggregate Model for Matching Text Sequences (2016, methodology): “Compare-Aggregate network: This kind of framework is to perform the word level matching (Wang & Jiang, 2016a; Parikh et al., 2016; He & Lin, 2016; Trischler et al., 2016; Wan et al., 2016).”
MITRE at SemEval-2018 Task 11: Commonsense Reasoning without Commonsense Knowledge (2018, methodology): “Trischler et al. (2016) compute a version of word overlap by ﬁnding cosine similarity of word representations be-tween sections.”
Prerequisite Skills for Reading Comprehension: Multi-Perspective Analysis of MCTest Datasets and Systems (2017, result): “Their performance, however, is still suboptimal: e.g., Trischler et al. (2016) yielded an accuracy of 77.5% on MCTest, which was poor in comparison with human performance (>95%).”
Globally Normalized Reader (2017, methodology): “Trischler et al. (2016b) also explored this approach in the context of question answering.”
Hierarchical Attention Flow for Multiple-Choice Reading Comprehension (2018, methodology): “Trischler et al. (2016) propose to conduct the match in multiple parallel perspectives with the hierarchical structure, and a training technique for the proposed neural model to converge on MCTest.”