Web Search Sessions Research Articles

Web Search Training and Query Formulation Behaviour of Undergraduates of Kano State University of Science and Technology (KUST) Wudil, Kano State, Nigeria

The Web is predominant source of information upon which most university students depend for accomplishing their academic tasks. This study investigated the influence of web search training program on query formulation behaviour of undergraduates of Kano State University of Science and Technology (KUST), Wudil, Kano State, Nigeria. The study was set to answer two research questions and tested a hypothesis. Quasi-experimental design of the pretest and posttest type was used for the study. All 20,679 undergraduates of KUST made up the target population. Convenience sampling technique was used to recruit only 35 students, who participated in a six-week training program and lab sessions. A structured and validated questionnaire was used for data collection at both pretest and posttest levels. Cronbach’s alpha reliability coefficient for the constructs ranged 0.78 to 0.88. Two adapted search task tests were used to assess students’ query formulation behaviour on the web before and after the training. Data were analyzed with descriptive statistics and t-test. Findings revealed that web search training had significant effect on the query formulation behaviour of participants (t(34) = 17.680; p <0.05,η2= 0.857). The study concluded that web search training is essential for all undergraduate students of KUST Wudil to enable them to develop appropriate web search query formulation behaviour and also accomplish their academic search tasks. The study recommended that academic librarians should include practical web search sessions in their use of library classes for freshmen. Keywords: Boolean logic, Query formulation behavior, Undergraduate students, Web Searching, Web search training

Proactive information retrieval

Users interact with digital systems with some underlying tasks in their minds. In our research scope, a task can be either single or multi-staged. A single-staged task is associated with a single information need, whereas a multi-staged one is associated with more than one information needs. An example of a single-staged task is suggesting related papers to an author while they are writing a section of the research papers. An example of a multi-staged task is planning for a vacation, where the different underlying information needs could be 'places to visit', 'booking accommodation' etc. In the process of accomplishing their task objectives, a user often needs to interact with an information retrieval (IR) system to address one or more information needs. For instance, for writing a research paper on a chosen topic, a user needs to look for existing research work related to the topic. Traditional IR systems do not take into account a user's task intent while showing search results to the user. In our work, we propose a methodology towards developing next generation IR systems (i.e. proactive IR systems) which seek to anticipate the task intent of a user from their interactions with digital systems in order to proactively suggest potentially relevant information sources to assist them to complete their tasks. Specifically speaking, in this PhD, we proposed an embedding approach that captures the task semantics from the interactions of a user with a digital system (e.g. laptop, desktop, smartphone etc.). The proposed embedding approach is then applied for the downstream tasks of providing proactive suggestions in both single and multi staged scenarios. For the single-staged task, we propose a simulation setup to simulate a user's reading and writing interactions in a desktop environment. For the multi-staged task, we focus on web search sessions where a user can have multiple information needs corresponding to a search task. We also proposed a reproducible evaluation framework to compare between different proactive suggestion models. Awarded by : Dublin City University, Ireland on 10 August 2021. Supervised by : Gareth Jones. Available at : https://procheta.github.io/sprocheta/Thesis.pdf.

Coherent Dialog Generation with Query Graph

Learning to generate coherent and informative dialogs is an enduring challenge for open-domain conversation generation. Previous work leverage knowledge graph or documents to facilitate informative dialog generation, with little attention on dialog coherence. In this article, to enhance multi-turn open-domain dialog coherence, we propose to leverage a new knowledge source, web search session data, to facilitate hierarchical knowledge sequence planning, which determines a sketch of a multi-turn dialog. Specifically, we formulate knowledge sequence planning or dialog policy learning as a graph grounded Reinforcement Learning (RL) problem. To this end, we first build a two-level query graph with queries as utterance-level vertices and their topics (entities in queries) as topic-level vertices. We then present a two-level dialog policy model that plans a high-level topic sequence and a low-level query sequence over the query graph to guide a knowledge aware response generator. In particular, to foster forward-looking knowledge planning decisions for better dialog coherence, we devise a heterogeneous graph neural network to incorporate neighbouring vertex information, or possible future RL action information, into each vertex (as an RL action) representation. Experiment results on two benchmark dialog datasets demonstrate that our framework can outperform strong baselines in terms of dialog coherence, informativeness, and engagingness.

Topic-independent modeling of user knowledge in informational search sessions

Web search is among the most frequent online activities. In this context, widespread informational queries entail user intentions to obtain knowledge with respect to a particular topic or domain. To serve learning needs better, recent research in the field of interactive information retrieval has advocated the importance of moving beyond relevance ranking of search results and considering a user’s knowledge state within learning oriented search sessions. Prior work has investigated the use of supervised models to predict a user’s knowledge gain and knowledge state from user interactions during a search session. However, the characteristics of the resources that a user interacts with have neither been sufficiently explored, nor exploited in this task. In this work, we introduce a novel set of resource-centric features and demonstrate their capacity to significantly improve supervised models for the task of predicting knowledge gain and knowledge state of users in Web search sessions. We make important contributions, given that reliable training data for such tasks is sparse and costly to obtain. We introduce various feature selection strategies geared towards selecting a limited subset of effective and generalizable features.

Lymelight: forecasting Lyme disease risk using web search data

Lyme disease is the most common tick-borne disease in the Northern Hemisphere. Existing estimates of Lyme disease spread are delayed a year or more. We introduce Lymelight—a new method for monitoring the incidence of Lyme disease in real-time. We use a machine-learned classifier of web search sessions to estimate the number of individuals who search for possible Lyme disease symptoms in a given geographical area for two years, 2014 and 2015. We evaluate Lymelight using the official case count data from CDC and find a 92% correlation (p < 0.001) at county level. Importantly, using web search data allows us not only to assess the incidence of the disease, but also to examine the appropriateness of treatments subsequently searched for by the users. Public health implications of our work include monitoring the spread of vector-borne diseases in a timely and scalable manner, complementing existing approaches through real-time detection, which can enable more timely interventions. Our analysis of treatment searches may also help reduce misdiagnosis of the disease.

Toward a web search model: Integrating multitasking, cognitive coordination, and cognitive shifts

AbstractLimited research has investigated the role of multitasking, cognitive coordination, and cognitive shifts during web search. Understanding these three behaviors is crucial to web search model development. This study aims to explore characteristics of multitasking behavior, types of cognitive shifts, and levels of cognitive coordination as well as the relationship between them during web search. Data collection included pre‐ and postquestionnaires, think‐aloud protocols, web search logs, observations, and interviews with 42 graduate students who conducted 315 web search sessions with 221 information problems. Results show that web search is a dynamic interaction including the ordering of multiple information problems and the generation of evolving information problems, including task switching, multitasking, explicit task and implicit mental coordination, and cognitive shifting. Findings show that explicit task‐level coordination is closely linked to multitasking, and implicit cognitive‐level coordination is related to the task‐coordination process; including information problem development and task switching. Coordination mechanisms directly result in cognitive state shifts including strategy, evaluation, and view states that affect users' holistic shifts in information problem understanding and knowledge contribution. A web search model integrating multitasking, cognitive coordination, and cognitive shifts (MCC model) is presented. Implications and further research also are discussed.

A capture–recapture sampling standardization for improving Internet meta-search

This work describes a novel sampling standardization for improving precision of third-party results in web meta-search. The standardization uses the capture–recapture methodology which is mainly applied for estimating evolution rates in wildlife biological studies, as well as a mechanism that records the users' browsing behavior during their web search sessions. The paper provides the implementation details and the initial assessment of a third-party results ranking algorithm which employs both mechanisms. It is proved that an important quality factor in providing relevant information is how frequent Internet search services refresh their databases. It is also proved that when users' browsing behavior is jointly examined with the ability of the Internet search services to offer new and fresh results, a more effective meta-search is provided. Experimental results have shown that the precision levels of third-party results were significantly increased in several recall levels, over a six-month period. For acquiring third-party results we used five known web search services namely, AltaVista, Google, Lycos, MSN, and Yahoo!

Identifying Web search session patterns using cluster analysis: A comparison of three search environments

AbstractSession characteristics taken from large transaction logs of three Web search environments (academic Web site, public search engine, consumer health information portal) were modeled using cluster analysis to determine if coherent session groups emerged for each environment and whether the types of session groups are similar across the three environments. The analysis revealed three distinct clusters of session behaviors common to each environment: “hit and run” sessions on focused topics, relatively brief sessions on popular topics, and sustained sessions using obscure terms with greater query modification. The findings also revealed shifts in session characteristics over time for one of the datasets, away from “hit and run” sessions toward more popular search topics. A better understanding of session characteristics can help system designers to develop more responsive systems to support search features that cater to identifiable groups of searchers based on their search behaviors. For example, the system may identify struggling searchers based on session behaviors that match those identified in the current study to provide context sensitive help.

Geographic intention and modification in web search

Web searchers signal their geographic intent by using place‐names in search queries. They also indicate their flexibility about geographic specificity by reformulating their queries. By examining this data we can learn to understand web searcher flexibility with respect to geographic intent. We examine aggregated data of queries with locations, and locations identified from IP addresses, to identify overall distance preferences, as well as distance preferences by search topic. We also examine query rewriting: both deliberate query rewriting, conducted in web search sessions, and automated query rewriting, with manual relevance judgments of geo‐modified queries. We find geo‐specification in 12.7% of user query rewrites in search sessions, and show the breakdown into sub‐classes such as same‐city, same‐state, same‐country and different‐country. We also measure the dependence between US‐state‐name and distance‐of‐modified‐location‐from‐original‐location, finding that Vermont web searchers modify their locations greater distances than California web searchers. We find that automatically‐modified queries are perceived as much more relevant when the geographic component is unchanged. We look at the relationship between the non‐location part of a query and the distance from the user. We see that people search for child day‐care near their locations and maps far from where they are located. We also give distance profiles for the top topics which cooccur with place‐names in queries, which could be used to set document priors based on document proximity and query topic.

Mining web search behaviors: Strategies and techniques for data modeling and analysis

AbstractThere is a growing interest in modeling Web searching behaviors using query log data. In this project, we identified some gaps in current research. We propose to model Web search behaviors along three dimensions: interactions, linguistic and cognitive behaviors. We propound Web search session as a vital important concept to study interactive behaviors using query logs. A highly granular, comprehensive relational model is presented for data extraction and transformation along with strategies and methods for session identification. To facilitate analysis, we developed an interactive Web tool for exploring different session thresholds. We demonstrate statistically that the 80‐20 empirical rule shows promise for setting session boundaries. In addition, we recommend that decisions for session boundary thresholds should be determined based on specific query corpus characteristics such as type and size of the database searched, and type of searchers who submit the queries. Our approach is based on the fact that data mining researchers do not always know all the hypotheses that the data can answer at the outset and the log data are diverse across environments due to the lack of standardization. This model maximizes transactional data inclusion, is flexible in handling data content, and can be extended easily to incorporate new hypotheses and new data elements as mining progresses.

Internet usage transaction log studies: The next generation

Internet usage transaction log studies: The next generation

Multitasking during Web search sessions

A user's single session with a Web search engine or information retrieval (IR) system may consist of seeking information on single or multiple topics, and switch between tasks or multitasking information behavior. Most Web search sessions consist of two queries of approximately two words. However, some Web search sessions consist of three or more queries. We present findings from two studies. First, a study of two-query search sessions on the Alta Vista Web search engine, and second, a study of three or more query search sessions on the Alta Vista Web search engine. We examine the degree of multitasking search and information task switching during these two sets of Alta Vista Web search sessions. A sample of two-query and three or more query sessions were filtered from Alta Vista transaction logs from 2002 and qualitatively analyzed. Sessions ranged in duration from less than a minute to a few hours. Findings include: (1) 81% of two-query sessions included multiple topics, (2) 91.3% of three or more query sessions included multiple topics, (3) there are a broad variety of topics in multitasking search sessions, and (4) three or more query sessions sometimes contained frequent topic changes. Multitasking is found to be a growing element in Web searching. This paper proposes an approach to interactive information retrieval (IR) contextually within a multitasking framework. The implications of our findings for Web design and further research are discussed.

Multitasking information behavior and information task switching: an exploratory study

Recent studies show that humans engage in multitasking information behaviors, often in libraries, as they seek and search for information on more than one information task. Multitasking information behaviors may consist of library search and use behaviors, or database or Web search sessions on multiple information tasks. However, few human information behavior models of seeking, searching or use, or library use models, include considerations of multitasking information behavior. This paper reports results from a case study exploring multitasking information behavior by an information seeker in a public library using diary, observation and interview data collection techniques. The information seeker sought information on four unrelated personal information tasks during two public library visits. Findings include a taxonomy of information behaviors; a sequential flowchart of the information seeker's complex and iterative processes, including multitasking information behavior, electronic searches, physical library searches, serendipitous browsing, and successive searches; and that the information seeker engaged in a process of 17 information task switches over two library visits. A model of information multitasking and information task switching is presented. Implications for library services and bibliographic instruction are also discussed.

Measuring the affective information environment of web searchers

AbstractInformation seeking research and theory is focusing increasingly on the role of affect in information behavior and how it influences cognitive operations. Affective variables that have been explored include need, preference, attitude, task motivation, expected and felt effort, uncertainty, self‐efficacy, optimism, relevance, satisfaction, and acceptance of or loyalty to the system. This study gives operational definitions for measuring several affective variables in the form of rating scales filled out by college students at the beginning and end of weekly Web search sessions throughout a semester. Intercorrelations and ANOVA analyses showed that there is a dynamic and coherent interaction among these affective variables. It is shown that the affective environment of searchers can be monitored objectively and continuously by means of such measures. A new concept termed “affective load” is introduced and defined, along with “user coping skills” which can counteract and reduce the negative effects of uncertainty, frustration, anxiety, irritation and rage during searching.

Information task switching and multitasking web search

AbstractA user's single session with a Web search engine may consist of seeking information on single or multiple topics. Some Web search sessions consist of three queries of one word or more. We present findings from a study of three or more‐query search sessions on the Alta Vista Web search engines. We examine the degree of multitasking search and information task switching by these Web searchers. A sample of three or more query sessions were filtered from Alta Vista transaction logs from 2002 and qualitatively analyzed. Findings include: (1) 91.3% of three or more query sessions contained multiple topics, (2) multitasking sessions contained frequent topic changes, and (3) there are a broad variety of search topics in multitasking search sessions. Multitasking is a key element within Web search. The implications of our findings for Web design and further research are discussed.

Multitasking information seeking and searching processes

AbstractRecent studies show that humans engage in multitasking behaviors as they seek and search information retrieval (IR) systems for information on more than one topic at the same time. For example, a Web search session by a single user may consist of searching on single topics or multitasking. Findings are presented from four separate studies of the prevalence of multitasking information seeking and searching by Web, IR system, and library users. Incidence of multitasking identified in the four different studies included: (1) users of the Excite Web search engine who completed a survey form, (2) Excite Web search engine users filtered from an Excite transaction log from 20 December 1999, (3) mediated on‐line databases searches, and (4) academic library users. Findings include: (1) multitasking information seeking and searching is a common human behavior, (2) users may conduct information seeking and searching on related or unrelated topics, (3) Web or IR multitasking search sessions are longer than single topic sessions, (4) mean number of topics per Web search ranged of 1 to more than 10 topics with a mean of 2.11 topic changes per search session, and (4) many Web search topic changes were from hobbies to shopping and vice versa. A more complex model of human seeking and searching levels that incorporates multitasking information behaviors is presented, and a theoretical framework for human information coordinating behavior (HICB) is proposed. Multitasking information seeking and searching is developing as major research area that draws together IR and information seeking studies toward a focus on IR within the context of human information behavior. Implications for models of information seeking and searching, IR/Web systems design, and further research are discussed.