Predicting Session Length for Product Search on E-commerce Platform

In this paper, a two-level deep learning framework is presented to model human information foraging behavior with search engines. A recurrent neural network architecture is designed using LSTM as the base unit to explicitly consider the temporal and spatial dependencies of information scents, the key concept in Information Foraging Theory. The target is to predict several major search behaviors, such as query abandonment, query reformulation, number of clicks, and information gain. The memory capability and the sequence structure of LSTM allow to naturally mimic not only what users are perceiving and performing at the moment but also what they have seen and learned from the past during the search dynamics. The promising results indicate that our information scent models with different input variations were better, compared to the state-of-the art neural click models, at predicting some search behaviors. When incorporating the knowledge from a previous query in the same search session, the prediction of current query abandonment, pagination, and information gain has been improved. Compared to the well known neural click models that model search behaviors under a single search query thread, this study takes a broader view to consider an entire search session which may contain multiple queries. More importantly, our model takes the search result relevance pattern on the Search Engine Results Pages (SERP) as a whole as the information scent input to the deep learning model, instead of considering one search result at each step. The results have insights on the impact of information scents on how people forage for information, which has implications for designing or refining a set of design guidelines for search engines.

The aim of this research proposal is to understand the extent in which Information Scent, part of Information Foraging Theory, influence the perception of relevance of retrieved search engine result pages (SERPs) in different environments (desktop and mobile). %We intend to study this influence by observing information seekers' behavior when they are searching for topics on a typical search engine. Firstly, we planned to investigate how previously found correlations between information scent and search behavior on desktop SERPs may also applicable on mobile. Secondly, we investigate if the differences in search behavior are also caused by SERP visibility and/or physical input behavior in different environments. Lastly, we would like to observe the extent in which changing snippet length can affect web search behavior. The perception of relevance on SERPs might be similar in both environments because the arrangement of SERPs on both mobile and desktop are identical; or different because the visibility of search results is less on mobile compared to desktop, due to limited screen sizes. We wanted to study the extent of which the physical limitations might significantly impact search behavior in different environments due to expected ``hidden scent. We conducted two user studies to observe these phenomenon and seek to understand search behavior by manipulating the SERPs presented to users. We observed both similarities and differences in web search behavior between the environments. On the desktop, increasing information scent led to lower positions of search results saved as relevant and lower positions of mouse hover. However, on mobile, increasing relevant search results beyond the initial screen size reduced number of documents examined.

Understanding how users examine result pages across a broad range of information needs is critical for search engine design. Cursor movements can be used to estimate visual attention on search engine results page (SERP) components, including traditional snippets, aggregated results, and advertisements. However, these signals can only be leveraged for SERPs where cursor tracking was enabled, limiting their utility for informing the design of new SERPs. In this work, we develop robust, log-based mouse movement models capable of estimating searcher attention on novel SERP arrangements. These models can help improve SERP design by anticipating searchers' engagement patterns given a proposed arrangement. We demonstrate the efficacy of our method using a large set of mouse-tracking data collected from two independent commercial search engines.

Toward creating a fairer ranking in search engine results

Search engine result pages (SERPs) usually display different information contents for web search users. In this paper, the effects of different information categories displayed on SERPs, including snippet, published date, and citation count, were investigated in the academic search context. In addition, the academic search behaviours and the satisfactions of 48 undergraduate students were explored. The results showed that participants were more satisfied with the SERPs displaying the snippet information than those displaying published date, and citation count information regardless of their familiarity with the search tasks' topics. Moreover, the participants were more satisfied with their final submitted options when they selected the submission from SERPs displaying the snippet information. The findings from this research could be used as the implication for SERPs interface design to provide valuable information for users without overwhelming them with excessive information categories.

Understanding user intent from her sequential search behaviors, i.e. predicting the intent of each user query in a search session, is crucial for modern Web search engines. However, due to the huge number of user behavior variables and coarse level intent labels defined by human editors, it is very difficult to directly model user behavioral dynamics or user intent dynamics in user search sessions. In this paper, we propose a novel Sparse Hidden-Dynamic Conditional Random Fields (SHDCRF) model for user intent learning from their search sessions. Through incorporating the proposed hidden state variables, SHDCRF aims to learn a substructure, i.e. a set of related hidden variables, for each intent label and they are used to model the intermediate dynamics between user intent labels and user behavioral variables. In addition, SHDCRF learns a sparse relation between the hidden variables and intent labels to make the hidden state variables explainable. Extensive experiment results, on real user search sessions from a popular commercial search engine show that the proposed SHDCRF model significantly outperforms in terms of intent prediction results that those classical solutions such as Support Vector Machine (SVM), Conditional Random Field (CRF) and Latnet-Dynamic Conditional Random Field (LDCRF).

By focusing on the point where the document triage process interacts with a search engine results page (SERP), this experiment extends our knowledge about both SERP design and document triage behavior. Prior SERP work has shown that longer meta descriptions in SERPs improve people's ability to answer information based questions, while document triage research has shown the importance of abstracts in making relevancy decisions. Using eye tracking equipment this work employed a repeated measure within factors experimental design method replacing the existing Google Scholar (GS) SERP meta descriptions with the abstracts of the corresponding retrieved articles. Undergraduate freshmen participants were asked to use two different GS SERPs, one with a control design and one with the experimental design and determine which resources are relevant to their assigned research task. The findings show that the participants changed how long they looked at the expanded meta description, while noticeably reducing how long they gazed at other parts of the page supporting other research findings. The addition of abstracts changed user behavior by reducing how often they made surrogate level document transitions, but did not change how often they sought out full-text documents, supporting the principle of least effort. The addition of abstracts did not contribute to changes in total time on task or participant's relevancy accuracy. This study's findings conflict with other work that found that longer meta descriptions corresponded with a reduction in total task time and an improvement in accuracy for informational tasks. Further research is needed to determine if this conflict was due to task differences or if the document triage task was not challenging enough.

This research focuses on what users know about search engine optimization (SEO) and how well they can identify results that have potentially been influenced by SEO. We conducted an online survey with a sample representative of the German online population (N = 2,012). We found that 43% of users assume a better ranking can be achieved without paying money to Google. This is in stark contrast to the possibility of influence through paid advertisements, which 79% of internet users are aware of. However, only 29.2% know how ads differ from organic results. The term ‘search engine optimization’ is known to 8.9% of users but 14.5% can correctly name at least one SEO tactic. Success in labelling results that can be influenced through SEO varies by search engine result page (SERP) complexity and devices: participants achieved higher success rates on SERPs with simple structures than on the more complex SERPs. SEO results were identified better on the small screen than on the large screen. 59.2% assumed that SEO has a (very) strong impact on rankings. SEO is more often perceived as positive (75.2%) than negative (68.4%). The insights from this study have implications for search engine providers, regulators, and information literacy.

The study of search satisfaction is one of the prime concerns in search performance evaluation research. Most existing works on search satisfaction primarily rely on the hypothesis that all results on search engine result pages (SERPs) are homogeneous. However, a variety of heterogeneous vertical results such as videos, images and instant answers are aggregated into SERPs by search engines to improve the diversity and quality of search results. In this paper, we carry out a lab-based user study with specifically designed SERPs to determine how verticals with different qualities and presentation styles affect search satisfaction. Users' satisfaction feedback and external assessors' satisfaction annotations are both collected to make a comparison regarding the perception of search satisfaction. Mouse click-through / movement data and eye movement information are also collected such that we can investigate the influence of vertical results from the perspectives of both benefit and cost. Finally, a vertical-aware learning-based prediction method is proposed to predict search satisfaction on aggregated SERPs. To the best of our knowledge, this paper is the first to analyze the effect of verticals on search satisfaction. The results show that verticals with different qualities, presentation styles and positions have different effects on search satisfaction, among which Encyclopedia verticals, as well as Download verticals, will bring the largest improvement. Furthermore, our proposed vertical-aware prediction method outperforms state-of-the-art methods that are designed for search satisfaction prediction in homogeneous environment.

PurposeGood abandonment behavior refers to users obtaining direct answers via search engine results pages (SERPs) without clicking any search result, which occurs commonly in mobile search. This study aims to better understand users' good abandonment behavior and perception, and then construct a good abandonment prediction model for mobile search with improved performance.Design/methodology/approachIn this study, an in situ user mobile search experiment (N = 43) and a crowdsourcing survey (N = 1,379) were conducted. Good abandonment behavior was analyzed from a quantitative perspective, exploring users' search behavior characteristics from four aspects: session and query, SERPs, gestures and eye-tracking data.FindingsUsers show less engagement with SERPs in good abandonment, spending less time and using fewer gestures, and they pay more visual attention to answer-like results. It was also found that good abandonment behavior is often related to users' perceived difficulty of the searching tasks and trustworthiness in the search engine. A good abandonment prediction model in mobile search was constructed with a high accuracy (97.14%).Originality/valueThis study is the first to explore eye-tracking characteristics of users' good abandonment behavior in mobile search, and to explore users' perception of their good abandonment behavior. Visual attention features are introduced into good abandonment prediction in mobile search for the first time and proved to be important predictors in the proposed model.

Search engines have become an important channel for increasing SMEs' global reach as well as competing with larger companies. Therefore, SMEs are enhancing their business visibility through search engine optimization (SEO). With effective SEO techniques, small companies can appear ahead of large, well-known companies on search results. However, in addition increasing the ranking on a search engine results page (SERP), there may be other benefits that SMEs may enjoy while engaging in an SEO campaign. This study uses multiple firms as cases to illustrate how SEO initiatives can impact performance for SMEs. The authors found that in addition to increasing the rankings on SERPs, SMEs saw an increase in number of visitors to the site, the average time duration of users visiting the site, more user engagement and an increase of annual sales revenue.

Article Info/Makale BilgisiNowadays, it can be said that studies on the intersection of "big data-network analysis" in the field of tourism have intensified, and the usage of Google and other social media platforms as the main big data sources has increased.This intersection study topic is important for determining tourist attraction centers and tourist routes, optimizing selection-decision systems, monitoring satisfaction and behavior, and thus determining tourist actions and usage patterns.In this regard, the main aim of the study is to analyze the spatial network structure created by the tourist attractions in Edirne city center and to evaluate it using the network analysis method.In this context, the locations defined in the "Things To-Do" category for Edirne city center on the Google Maps platform were considered.Depending on the content of the Google Search Engine Results Page (SERP), the rating level of the locations and the "People Also Searching For (PASF)" information, which is an important content in the field of Search Engine Optimization (SEO) today, were collected.PASF information reveals other places that users are searching for depending on a location, and as a result, a tourism search network is formed.Network analysis was applied, and evaluations were made considering the spatial dimensions of the locations.As a result, tourism places with high centrality value for Edirne city center, thematic neighborhood groups were identified, and route suggestions were presented.It is expected that the study will contribute to literature in terms of method and evaluation perspective. Gnmzde turizm alannda "byk veri-a analizi" kesiiminde almalarn younlat, Google ve dier sosyal medya platformlarnn temel byk veri kaynaklar kullanmnn artt sylenebilmektedir. Bu ara kesiim alma temas ise turistik cazibe merkezleri ve turistik rotalarn belirlenebilmesi, seim-karar sistemlerinin optimize edilebilmesi, memnuniyet ve davranlarn izlenebilmesi dolays turist davran ve kullanm rntlerinin tespit edilebilmesi iin nemlidir. Bu bakmdan almann temel amac Edirne kent merkezindeki turistik ekici noktalarn yaratm olduu meknsal a yapsnn zmlenmesi ve a analizi yntemi ile deerlendirilmesidir. Bu kapsamda Google Maps platformunda Edirne kent merkezi iin "ThingsTo-Do" kategorisinde tanmlanan konumlar ele alnmtr.Google Search Engine Results Page (SERP) ieriine bal olarak, konumlarn rating dzeyi ile gnmzde Search Engine Optimization (SEO) alannda nemli bir ierik olan "People Also Search For (PASF)" bilgisi toplanmtr.PASF bilgisi, kullanclarn bir konuma bal olarak aradklar dier konumlar ortaya koymakta sonucunda ise bir turizm arama a elde edilmektedir

Since Information Retrieval (IR) is an interactive process in general, it is important to study Interactive Information Retrieval (IIR), where we would attempt to model and optimize an entire interactive retrieval process (rather than a single query) with consideration of many different ways a user can potentially interact with a search engine. This tutorial systematically reviews the progress of research in IIR with an emphasis on the most recent progress in the development of models, algorithms, and evaluation strategies for IIR. It starts with a broad overview of research in IIR and then gives an introduction to formal models for IIR using a cooperative game framework and covering decision-theoretic models such as the Interface Card Model and Probability Ranking Principle for IIR. Next, it provides a review of some representative specific techniques and algorithms for IIR, such as various forms of feedback techniques and diversification of search results, followed by a discussion of how an IIR system should be evaluated and multiple strategies proposed recently for evaluating IIR using user simulation. The tutorial ends with a brief discussion of the major open challenges in IIR and some of the most promising future research directions.

Although Web search engines still answer user queries with lists of ten blue links to webpages, people are increasingly issuing queries to accomplish their daily tasks (e.g., finding a recipe , booking a flight , reading online news , etc.). In this work, we propose a two-step methodology for discovering tasks that users try to perform through search engines. First, we identify user tasks from individual user sessions stored in search engine query logs. In our vision, a user task is a set of possibly noncontiguous queries (within a user search session), which refer to the same need. Second, we discover collective tasks by aggregating similar user tasks, possibly performed by distinct users. To discover user tasks, we propose query similarity functions based on unsupervised and supervised learning approaches. We present a set of query clustering methods that exploit these functions in order to detect user tasks. All the proposed solutions were evaluated on a manually-built ground truth, and two of them performed better than state-of-the-art approaches. To detect collective tasks, we propose four methods that cluster previously discovered user tasks, which in turn are represented by the bag-of-words extracted from their composing queries. These solutions were also evaluated on another manually-built ground truth.

The World Wide Web plays a central role in many aspects of our modern life. In particular, using search engines to access information about products and services has become an integral part of our day-to-day activities. In this study we look at users' viewing behavior on search engine results pages (SERPs) through the lens of competition for attention theory. While this theory has been used for examining consumer behavior on e-commerce websites, little work has been done to test this theory for viewing behavior on SERPs. We use eye tracking data to analyze viewing behavior. The results show that viewing behavior can have an impact on a user experience and effective search, providing theoretical direction for studying the viewing behavior of SERPs.