Which questions should I answer? Salience Prediction of Inquisitive Questions

Yating Wu,Ritika Mangla,Alexandros G. Dimakis,Greg Durrett,Junyi Jessy Li

2024-10-04

Abstract:Inquisitive questions -- open-ended, curiosity-driven questions people ask as they read -- are an integral part of discourse processing (Kehler and Rohde, 2017; Onea, 2016) and comprehension (Prince, 2004). Recent work in NLP has taken advantage of question generation capabilities of LLMs to enhance a wide range of applications. But the space of inquisitive questions is vast: many questions can be evoked from a given context. So which of those should be prioritized to find answers? Linguistic theories, unfortunately, have not yet provided an answer to this question. This paper presents QSALIENCE, a salience predictor of inquisitive questions. QSALIENCE is instruction-tuned over our dataset of linguist-annotated salience scores of 1,766 (context, question) pairs. A question scores high on salience if answering it would greatly enhance the understanding of the text (Van Rooy, 2003). We show that highly salient questions are empirically more likely to be answered in the same article, bridging potential questions (Onea, 2016) with Questions Under Discussion (Roberts, 2012). We further validate our findings by showing that answering salient questions is an indicator of summarization quality in news.

Computation and Language

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

The paper primarily aims to address the following issues:

1. **Importance Prediction of Exploratory Questions**:
- Exploratory questions are open-ended questions posed by readers out of curiosity during the reading process. These questions are an important part of understanding the text. However, not all exploratory questions are equally important. Therefore, the paper proposes a method to predict the importance (i.e., salience) of these questions to determine which questions deserve priority in answering.

2. **Development of a Salience Prediction Model**:
- The paper develops a salience prediction model named QSALIENCE, which is trained on a new dataset annotated by linguists. Through this model, the salience scores of exploratory questions posed in a given context can be predicted.

3. **Study of the Relationship Between Salience and Answerability**:
- The research shows that questions with higher salience are more likely to be answered within the same article, thereby validating the effectiveness of the salience scores. Additionally, the paper explores the relationship between salience scores and the answerability of questions.

4. **Application in Downstream Tasks**:
- The paper further demonstrates the value of salience prediction in the task of news summary expansion. Specifically, summaries that can answer more salient questions are generally of higher quality and better meet the needs of readers.

Through these studies, the paper aims to bridge the gap between theory and practice regarding the salience of exploratory questions and provide a practical tool to optimize the process of information retrieval and processing.