Extracting Hierarchies of Search Tasks & Subtasks via a Bayesian Nonparametric Approach

Abstract

A signi cant amount of search queries originate from some real world information need or tasks [13]. In order to improve the search experience of the end users, it is important to have accurate representations of tasks. As a result, signi cant amount of research has been devoted to extracting proper representations of tasks in order to enable search systems to help users complete their tasks, as well as providing the end user with be er query suggestions [9], for be er recommendations [41], for satisfaction prediction [36] and for improved personalization in terms of tasks [24,38]. Most existing task extraction methodologies focus on representing tasks as at structures. However, tasks o en tend to have multiple subtasks associated with them and a more naturalistic representation of tasks would be in terms of a hierarchy, where each task can be composed of multiple (sub)tasks. To this end, we propose an e cient Bayesian nonparametric model for extracting hierarchies of such tasks & subtasks. We evaluate our method based on real world query log data both through quantitative and crowdsourced experiments and highlight the importance of considering task/subtask hierarchies. KEYWORDS search tasks; bayesian non-parametrics; hierarchical model ACM Reference format:

Tempo-Lexical Context Driven Word Embedding for Cross-Session Search Task Extraction

“…A Chinese Restaurant Process (CRP) based posterior inference process was then used to extract the tasks from individual queries. In an extension of this work (Mehrotra and Yilmaz, 2017), the authors proposed a Bayesian non-parametric approach for extracting task hierarchies. The main difference between our approach and (Mehrotra et al, 2016;Mehrotra and Yilmaz, 2017) is that our focus is on finding cross-session tasks from a query log, rather than finding hierarchies of tasks.…”

“…In an extension of this work (Mehrotra and Yilmaz, 2017), the authors proposed a Bayesian non-parametric approach for extracting task hierarchies. The main difference between our approach and (Mehrotra et al, 2016;Mehrotra and Yilmaz, 2017) is that our focus is on finding cross-session tasks from a query log, rather than finding hierarchies of tasks. Further, instead of using similarities between embedded query vectors as one of the features to estimate the relatedness between two queries, we propose a task semantics driven embedding technique to transform a query in close proximity to its taskrelated counterpart.…”

Deep Sequential Models for Task Satisfaction Prediction

“…Sequence: Given a search impression and a list of possible user actions, a sequence is de ned as a time-ordered list of actions performed by the user when interacting with the search result page. Search Task: A search task is an atomic information need resulting in one or more queries [19] Complex Task: A complex search task is a multi-aspect or a multistep information need consisting of a set of related tasks, each of which might recursively be complex [33]. With this background, we formally de ne the problem of satisfaction prediction:…”

“…In order to make task satisfaction predictions, we leverage query level satisfaction estimates as well as subtask level satisfaction estimates. While few e cient approaches exist for automatically identifying subtasks [32,33], we assume access to subtask demarcation information (obtained via crowdsourced labeling) for the scope of this work.…”

“…Most existing work on gauging user satisfaction have focused on query level satisfaction [10,11,15,22], with some initial e orts aimed at measuring task satisfaction for simple tasks [14]. O en, independent information needs arise from an overall complex search task, where a complex search task refers to a multi-aspect or a multi-step information need consisting of a set of related tasks, each of which might recursively be complex [2,19,32,33]. While existing work has primarily focused on measuring user satisfaction on simple search tasks, work on understanding and measuring user satisfaction for complex search tasks remains in its infancy.…”