Toward Connecting Speech Acts and Search Actions in Conversational Search Tasks

Souvick Ghosh,Satanu Ghosh,Chirag Shah
DOI: https://doi.org/10.48550/arXiv.2305.04858
2023-05-09
Abstract:Conversational search systems can improve user experience in digital libraries by facilitating a natural and intuitive way to interact with library content. However, most conversational search systems are limited to performing simple tasks and controlling smart devices. Therefore, there is a need for systems that can accurately understand the user's information requirements and perform the appropriate search activity. Prior research on intelligent systems suggested that it is possible to comprehend the functional aspect of discourse (search intent) by identifying the speech acts in user dialogues. In this work, we automatically identify the speech acts associated with spoken utterances and use them to predict the system-level search actions. First, we conducted a Wizard-of-Oz study to collect data from 75 search sessions. We performed thematic analysis to curate a gold standard dataset -- containing 1,834 utterances and 509 system actions -- of human-system interactions in three information-seeking scenarios. Next, we developed attention-based deep neural networks to understand natural language and predict speech acts. Then, the speech acts were fed to the model to predict the corresponding system-level search actions. We also annotated a second dataset to validate our results. For the two datasets, the best-performing classification model achieved maximum accuracy of 90.2% and 72.7% for speech act classification and 58.8% and 61.1%, respectively, for search act classification.
Human-Computer Interaction,Information Retrieval
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to predict system - level search actions by identifying users' speech acts in conversational search tasks, thereby improving the natural language understanding of the conversational search system and its ability to perform complex search tasks. Specifically, the research aims to: 1. **Improve natural language understanding**: By identifying speech acts in user utterances, enhance the understanding ability of the conversational search system for user utterances, especially when dealing with non - factual information needs. 2. **Connect speech acts and search actions**: After accurately predicting speech acts, use this information to guide the system to perform corresponding search actions, achieving an effective conversion from user utterances to system actions. To achieve the above goals, the author has carried out research work in the following aspects: - **Data collection and annotation**: Through the Wizard - of - Oz research method, data of 75 search sessions were collected, and topic analysis was carried out on these sessions. 1,834 utterances and 509 system actions were annotated, forming a high - quality annotated data set. - **Model development**: A deep neural network model based on the attention mechanism was developed to automatically classify users' utterance behaviors and further predict corresponding system search actions. - **Experimental verification**: The effectiveness of the model was verified through experiments. The best model achieved speech act classification accuracies of 90.2% and 72.7% on two data sets respectively, and search action classification accuracies of 58.8% and 61.1% respectively. In general, this paper is committed to improving the intelligence level and user experience of the conversational search system by connecting users' speech acts and the system's search actions.