Identifying task-based sessions in search engine query logs

Abstract

The research challenge addressed in this paper is to devise e↵ective techniques for identifying task-based sessions, i.e. sets of possibly non contiguous queries issued by the user of a Web Search Engine for carrying out a given task. In order to evaluate and compare di↵erent approaches, we built, by means of a manual labeling process, a ground-truth where the queries of a given query log have been grouped in tasks. Our analysis of this ground-truth shows that users tend to perform more than one task at the same time, since about 75% of the submitted queries involve a multi-tasking activity. We formally define the Task-based Session Discovery Problem (TSDP) as the problem of best approximating the manually annotated tasks, and we propose several variants of well known clustering algorithms, as well as a novel e cient heuristic algorithm, specifically tuned for solving the TSDP. These algorithms also exploit the collaborative knowledge collected by Wiktionary and Wikipedia for detecting query pairs that are not similar from a lexical content point of view, but actually semantically related. The proposed algorithms have been evaluated on the above groundtruth, and are shown to perform better than state-of-the-art approaches, because they e↵ectively take into account the multi-tasking behavior of users.

Discovering tasks from search engine query logs

“…In a previous work [Lucchese et al 2011], we already proved that user tasks can effectively be found by exploiting the lexical and semantic content similarity of queries issued by individuals within specific search contexts (i.e., time-bounded subsessions of the original query session). Conversely, the same approach would not be able to discover collective tasks if applied directly to users' queries, because queries that are issued by two users, which are lexically or semantically similar, might refer to different latent needs.…”

“…In Lucchese et al [2011], we already showed that users perform multitasking search activities in the query streams issued to a search engine . Multitasking refers to the way users interact with a search engine, by intertwining different tasks within the same time period.…”

“…Unlike the unsupervised learning approach introduced in Lucchese et al [2011], here the task relatedness is learned by training several classifiers in our ground truth. In particular, we exploit the binary classifiers introduced by Jones and Klinkner [2008] and use the prediction provided by these classifiers in order to determine whether two queries belong to the same task.…”

Task-aware search assistant

“…We also use many of the same lexical features: Jaccard coefficient, tri-gram character overlap, and Levenshtein distance. We borrow a feature presented by Lucchese et al [4], which computes the semantic similarity between two queries by the cosine similarity between vectors of tf-idf scores over Wikipedia documents. We also use their weighted connected components clustering method once the pair-wise classifications have been made.…”

“…Previous studies have observed an average of 2.6-3.3 queries per search task [2,3,4] and Jones and Klinkner [2] found 16% of tasks were revisited over a three-day span of Yahoo! logs.…”