Exploring Question Understanding and Adaptation in Neural-Network-Based Question Answering

By Junbei Zhang, Xiao-Dan Zhu, Qian Chen, Lirong Dai, Si Wei, Hui Jiang. 2017

54 citations recorded by the source corpus.

The last several years have seen intensive interest in exploring neural-network-based models for machine comprehension (MC) and question answering (QA). In this paper, we approach the problems by closely modelling questions in a neural network framework. We first introduce syntactic information to help encode questions. We then view and model different types of questions and the information shared among them as an adaptation task and proposed adaptation models for them. On the Stanford Question Answering Dataset (SQuAD), we show that these approaches can help attain better results over a competitive baseline.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Question Answering & Reading Comprehension

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 Question Answering & Reading Comprehension

Citations in the corpus

QANet: Combining Local Convolution with Global Self-Attention for Reading Comprehension (2018, neutral): “…(Weissenborn et al., 2017) 70.8 / 78.9 70.8 / 78.9 ReasoNet (Shen et al., 2017b) 69.1 / 78.9 70.6 / 79.4 Document Reader (Chen et al., 2017) 70.0 / 79.0 70.7 / 79.4 Ruminating Reader (Gong & Bowman, 2017) 70.6 / 79.5 70.6 / 79.5 jNet (Zhang et al., 2017) 70 LSTMs as is used in most existing models.”
The Effect of Natural Distribution Shift on Question Answering Models (2020, neutral): “jNet (Ensemble) (Zhang et al., 2017) https://worksheets.”
Adversarial Examples for Evaluating Reading Comprehension Systems (2017, background): “, 2017), jNet (Zhang et al., 2017), Ruminating Reader (Gong and Bowman, 2017), MultiPerspective Context Matching (MPCM) Single version (Wang et al.”
Gated Self-Matching Networks for Reading Comprehension and Question Answering (2017, neutral): “…Dynamic Coattention Networks (Xiong et al., 2016) 65.4 / 75.6 66.2 / 75.9 RaSoR (Lee et al., 2016) 66.4 / 74.9 - / - BiDAF (Seo et al., 2016) 68.0 / 77.3 68.0 / 77.3 jNet (Zhang et al., 2017) - / - 68.7 / 77.4 Multi-Perspective Matching (Wang et al., 2016) - / - 68.9 / 77.8 FastQA (Weissenborn et”
Enhanced LSTM for Natural Language Inference (2016, methodology): “The difference and element-wise product are then concatenated with the original vectors, ā and ã, or b̄ and b̃, respectively (Mou et al., 2016; Zhang et al., 2017).”
ReasoNet: Learning to Stop Reading in Machine Comprehension (2016, positive): “…RaSoR [14] - - 67.4 75.5 Multi-Perspective Matching [28] 68 Iterative Co-attention Network α 67.5 76.8 - - FastQA [30] 68.4 77.1 70.8 78.9 jNet [36] 68.7 77.4 - - Document Reader [4] 69.9 78.9 - - R-Net [27] 71.3 79.7 75.9 82.9 Vocab Size : We use the python NLTK tokenizer 6 to preprocess…”
Phase Conductor on Multi-layered Attentions for Machine Comprehension (2017, methodology): “Following Hu et al. (2017), we also use question type (what, how, who, when, which, where, why, be, and other) features (Zhang et al., 2017) where each type is represented by a trainable embedding.”
Reinforced Mnemonic Reader for Machine Comprehension (2017, methodology): “Following (Zhang et al. 2017), we count the key word frequency of all queries and obtain top-9 query categories: what , how , who , when , which , where , why , be , other .”
Ruminating Reader: Reasoning with Gated Multi-hop Attention (2017, methodology): “Comparing to the jNet (Zhang et al., 2017) whose success answers occupy 69.”
Multilingual Extractive Reading Comprehension by Runtime Machine Translation (2018, background): “Previous work reported that the SQuAD models including BiDAF tend to rely on some superficial cues or interrogative of questions (Hu, Peng, and Qiu 2017; Zhang et al. 2017; Jia and Liang 2017), and thus incorrectly translating a single word in a question sentence could sometimes result in a fatal er”
Hierarchical Attention Flow for Multiple-Choice Reading Comprehension (2018, background): “Zhang et al. (2017) incorporate syntactic information to explore better understanding and adaptation, but limited to the question only.”
Mnemonic Reader for Machine Comprehension (2017, methodology): “…model the query category, we count the key word frequency of all queries and obtain top-9 most frequent query categories, which is similar in (Zhang et al., 2017): what, how, who, when, which, where, why, be, other, in which be stands for queries beginning with different forms of key word…”
Neural Natural Language Inference Models Enhanced with External Knowledge (2017, methodology): “…for other NLP tasks, including word embedding (Chen et al., 2015; Faruqui et al., 2015; Liu et al., 2015; Wieting et al., 2015; Mrksic et al., 2017), machine translation (Shi et al., 2016; Zhang et al., 2017b), language modeling (Ahn et al., 2016), and dialogue systems (Chen et al., 2016b).”
Enhancing Sentence Embedding with Generalized Pooling (2018, methodology): “To help replicate our results, we publish our code3, which is developed from our codebase for multiple tasks (Chen et al., 2018; Chen et al., 2017a; Chen et al., 2016; Zhang et al., 2017).”