Abstract
In recent years, the Sq-MRC (machine reading comprehension for separate questions) task, where the questioner poses a separate question each time, has experienced rapid development. However, we often encounter circumstances where the questioner poses a series of nonstop questions. The corresponding task, called Cq-MRC (machine reading comprehension for continuous questions), has been rarely investigated and is nontrivial due to (1) the incomplete expressions of subsequent questions, such as anaphora or subject ellipsis, and (2) the shortage of annotated samples for model training and performance evaluation. We explore this challenging task, in this article, and propose a completion-then-prediction approach. To tackle the first issue, we identify the key-entity in the first question and employ it to complete the subsequent ones, and hence decompose the task into a set of Sq-MRC tasks which are accomplished using a BERT-based model. In addition, to mitigate the shortage of annotated samples, we train a meta-learner on the Sq-MRC task which is related to Cq-MRC and has abundant labeled data, and then perform few-shot learning by fine-tuning the model on our target task, viz. Cq-MRC, using the limited number of samples available. The results of extensive experiments conducted on a multi-document Chinese machine reading comprehension dataset show the effectiveness and superiority of our method, achieving 2.3% to 3.7% absolute improvements in terms of major metrics.
Highlights
As a key component of open-domain question answering [1], machine reading comprehension (MRC), which is defined as the task of enabling machines to understand natural language documents and answer related questions, has received wide research attention in the past few years
VOLUME 9, 2021 we train a meta-learner on the Sq-MRC task which is related to Cq-MRC and has abundant labeled data, and perform few-shot learning by fine-tuning the model on our target task, viz. Cq-MRC, using the limited number of samples available
We propose a completion--prediction approach to the Cq-MRC task, which identifies the key-entity using a deep neural model and uses it to complete the subsequent questions with a heuristic method, and to employ a BERT-based model to predict answers for each of the resulting questions
Summary
As a key component of open-domain question answering [1], machine reading comprehension (MRC), which is defined as the task of enabling machines to understand natural language documents and answer related questions, has received wide research attention in the past few years. We propose a completion--prediction approach to the Cq-MRC task, which identifies the key-entity using a deep neural model and uses it to complete the subsequent questions with a heuristic method, and to employ a BERT-based model to predict answers for each of the resulting questions. To address the issue of the shortage of annotated samples, we propose to utilize the meta-learner trained on the Sq-MRC task, which is related to Cq-MRC and has abundant labeled data, as prior knowledge to guide the few-shot learning of the model for our target task.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have