Generating Instructive Questions from Multiple Articles to Guide Reading in E-Bibliotherapy

Abstract

E-Bibliotherapy deals with adolescent psychological stress by manually or automatically recommending multiple reading articles around their stressful events, using electronic devices as a medium. To make E-Bibliotherapy really useful, generating instructive questions before their reading is an important step. Such a question shall (a) attract teens' attention; (b) convey the essential message of the reading materials so as to improve teens' active comprehension; and most importantly (c) highlight teens' stress to enable them to generate emotional resonance and thus willingness to pursue the reading. Therefore in this paper, we propose to generate instructive questions from the multiple recommended articles to guide teens to read. Four solutions based on the neural encoder-decoder model are presented to tackle the task. For model training and testing, we construct a novel large-scale QA dataset named TeenQA, which is specific to adolescent stress. Due to the extensibility of question expressions, we incorporate three groups of automatic evaluation metrics as well as one group of human evaluation metrics to examine the quality of the generated questions. The experimental results show that the proposed Encoder-Decoder with Summary on Contexts with Feature-rich embeddings (ED-SoCF) solution can generate good questions for guiding reading, achieving comparable performance on some semantic similarity metrics with that of humans.

Keywords: E-bibliotherapy; dataset; encoder-decoder; instructive question; reading guidance.

Figure 1

The interfaces of TeenRead. Teens first select several undergoing stress categories, then TeenRead recommends 4 articles based on the stress categories and teens’ instant stressful events.

Figure 2

The crawling process of TeenQA. Quora provides a related-topic-list for each topic, we recursively crawl 2-layer related topics of the topic seed.

Figure 3

Lengths of QAs in TeenQA.

Figure 4

Number of questions’ follows.

Figure 5

Number of answers.

Figure 6

Solution 1: ED-SoO. E, D denote the encoder and decoder, ${\mathit{e}}_{ij}$ is the word embedding of the jth word in the ith article, ${\mathit{c}}_i$ is the context vector of the ith article.

Figure 7

Words overlap between questions and the first m words of answers in TeenQA.

Figure 8

Solution 2: ED-SoI. E, D and C denote the encoder, decoder and context vector, respectively. ${\mathit{e}}_j$ is the word embedding of the jth word in the multi-document summary.

Figure 9

Words overlap between questions and the first s words of the multi-doc summary of 4 answers in TeenQA.

Figure 10

The ED-SoC solution and ED-SoCF solution. SE denotes the summary encoder that sequentially encodes ${\mathit{c}}_1$, …, ${\mathit{c}}_N$ into an overall context vector C. ${\mathit{e}}_{ij}$ and ${\mathit{e}}_{ij}^f$ are the word embedding and feature embedding of the jth word in the ith article. ${\alpha }_i^t$ and ${\beta }_{ij}^t$ is the article-level attention weight and the word-level attention weight.

Figure 11

Article attention visualization.