Real Search Research Articles

Mapping user search queries to product categories

Gathering detailed information about user product in-terests is becoming increasingly important for online advertisers. However, when gathering this information, maintaining the privacy of online users is a concern. This research is part of a larger project aiming to provide privacy preserving advertising. Specifically, this research aims to provide a method for mapping user search queries to actual product categories while pre-serving the users' privacy by storing user information on the local host. Product information gathered from two large shopping websites and real user search queries from a log file are used to match user search queries with the most relevant product categories. In matching search queries to product categories, we explore several issues including the algorithm used to rank product categories, the index size, which fields in the index are searched (product description or product name), and what type of product categories are used. Our findings indicate that the most successful algorithms on the user's computer, which preserve privacy, can match the results of those where information is sent to a central server. In addition, the description field of products is the most useful, particularly when searched as a phrase. Having specific fine-grained product categories would help advertisers, search engines and marketers by providing them more information about users while pre-serving user privacy.

Partitioning Search Spaces of a Randomized Search

This paper studies the following question: given an instance of the propositional satisfiability problem, a randomized satisfiability solver, and a cluster of n computers, what is the best way to use the computers to solve the instance? Two approaches, simple distribution and search space partitioning as well as their combinations are investigated both analytically and empirically. It is shown that the results depend heavily on the type of the problem (unsatisfiable, satisfiable with few solutions, and satisfiable with many solutions) as well as on how good the search space partitioning function is. In addition, the behavior of a real search space partitioning function is evaluated in the same framework. The results suggest that in practice one should combine the simple distribution and search space partitioning approaches.

Recommender Systems for Web Intelligence

Various concepts, methods, and technical architectures of recommender systems have been integrated into E-commerce storefronts, such as Amazon.com, Netflix, etc. Thereby, recently, Web users have become more familiar with the notion of recommendations. Nevertheless, little work has been done to integrate recommender systems into scientific information retrieval repositories, such as libraries, content management systems, online learning platforms, etc. This paper presents an implementation of a hybrid recommender system to personal the user’s experience on a real online learning repository and vertical search engine named HyperManyMedia. This repository contains educational content of courses, lectures, multimedia resources, etc. The main objective of this paper is to illustrate the methods, concepts, and architecture that we used to integrate a hybrid recommender system into the HyperManyMedia repository. This recommender system is driven by two types of recommendations: content-based (domain ontology model) and rule-based (learner’s interestbased and cluster-based). Finally, combining the contentbased and the rule-based models provides the user with hybrid recommendations that influence the ranking of the retrieved documents with different weights. Our experiments were carried out on the HyperManyMedia semantic search engine at Western Kentucky University. We used Top-n-Recall and Top-n-Precision to measure the effectiveness of re-ranking based on the learner’s semantic profile. Overall, the results demonstrate the effectiveness of the re-ranking based on personalization.

A Hybrid Recommender System Guided by Semantic User Profiles for Search in the E-learning Domain

Various concepts, methods, and technical architectures of recommender systems have been integrated into E-commerce storefronts, such as Amazon.com, Netflix, etc. Thereby, recently, Web users have become more familiar with the notion of recommendations. Nevertheless, little work has been done to integrate recommender systems into scientific information retrieval repositories, such as libraries, content management systems, online learning platforms, etc. This paper presents an implementation of a hybrid recommender system to personal the user's experience on a real online learning repository and vertical search engine named HyperManyMedia. This repository contains educational content of courses, lectures, multimedia resources, etc. The main objective of this paper is to illustrate the methods, concepts, and architecture that we used to integrate a hybrid recommender system into the HyperManyMedia repository. This recommender system is driven by two types of recommendations: content-based (domain ontology model) and rule-based (learner's interest- based and cluster-based). Finally, combining the content- based and the rule-based models provides the user with hybrid recommendations that influence the ranking of the retrieved documents with different weights. Our experiments were carried out on the HyperManyMedia semantic search engine at Western Kentucky University. We used Top-n-Recall and Top-n-Precision to measure the effectiveness of re-ranking based on the learner's semantic profile. Overall, the results demonstrate the effectiveness of the re-ranking based on personalization.

An Eye Movement Analysis of Evolving User Preferences during Internet Search

Using an eye tracker we examined decision making processes during a simulated real estate internet search. Specifically, we studied discrepancies between “stated preferences”, or preferences that are revealed through verbalizations or answers on a questionnaire, and “revealed preferences”, or preferences that become apparent from an actual decision or choice. Participants viewed photographs of homes, similar to those seen on actual real estate websites. Average dwell time, fixation duration/counts, and saccade counts/amplitudes were compared to their stated preferences on home survey questionnaires and assessed by gender. This information was used to determine if discrepancies exist between stated and revealed preferences, what qualities on the website drew the homebuyer's attention, how this may differ by gender, and ways in which these websites can more effectively be designed. In addition, we examined the effect of negative externalities, or physical properties that may affect an individual's internet search decisions. We manually altered several homes to reveal Level 1 externalities (or, properties that can easily be changed) such as pink paint on the wall, and Level 2 externalities (or, properties that cannot easily be changed) such as power lines in front of the house. Results revealed that men demonstrated a more aggressive search pattern than women, and the men had a greater number of saccades with shorter saccade amplitudes for each of the web pages seen. Dwell time, fixation duration, and fixation count were also found to change when participants were shown a home with the presence of a Level 1 (pink paint) or Level 2 (power line) externality, depending on gender. Data also suggested that a discrepancy exists between stated and revealed preferences. While a participant stated that they greatly disliked a room on the home specific survey, their dwell times, fixation durations, and fixation counts did not significantly decrease. These findings have implications for web design.

An automated CAE system for multidisciplinary structural design: its application to micro accelerometer

This paper describes a new computer-aided engineering system for multidisciplinary structural design. An automatic finite element mesh generation technique, based on fuzzy knowledge processing and computational geometry, is incorporated into the system, together with one of the commercial finite element analysis codes. A bubble is generated if its distance from existing bubble points is similar to the bubble-spacing function at the point. The bubble-spacing function is well controlled by fuzzy knowledge processing. The Delaunay method is employed as a basic tool for element generation. Automatic finite element generation for three-dimensional MEMS holds great benefits for analyses. An optimum design solution or satisfactory solutions will be automatically searched using the genetic algorithms modified for real search space, together with the automated finite element analysis system. A novel CAE system was developed, and successfully applied to the shape design of a micro accelerometer based on a tunneling current concept.

Boosting video tracking performance by means of Tabu Search in intelligent visual surveillance systems

In this paper, we present a fast and efficient technique for the data association problem applied to visual tracking systems. Visual tracking process is formulated as a combinatorial hypotheses search with a heuristic evaluation function taking into account structural and specific information such as distance, shape, color, etc. We introduce a Tabu Search algorithm which performs a search on an indirect space. A novel problem formulation allows us to transform any solution into the real search space, which is needed for fitness calculation, in linear time. This new formulation and the use of auxiliary structures yields a fast transformation from a blob-to-track assignment space to the real shape and position of tracks space (while calculating fitness in an incremental fashion), which is key in order to produce efficient and fast results. Other previous approaches are based on statistical techniques or on evolutionary algorithms. These techniques are quite efficient and robust although they cannot converge as fast as our approach.

TWO-STAGE SEGMENTATION OF AERIAL IMAGES FOR SEARCH AND RESCUE

This paper presents a novel two-stage image segmentation approach for detection of artificial materials and objects in non-urban terrain. It is based on the assumption that an object and a natural background have different color/saturation. High resolution image of the unknown terrain taken with digital camera from a distance of about 100 m is divided into smaller sub-images for further processing. Each sub-image is clustered using mean shift algorithm, and information about obtained cluster centers is transferred to the next stage. Second stage uses information about cluster centers from all sub-images and applies the same clustering method to this data set. Finally, a decision-making module evaluates all clusters and eventually proposes image segments that have high possibility of presenting the artificial material or the object in the input image. The proposed method has been tested on 22 aerial images typical for search and rescue. Observed results are validated through recall and precision and it has been shown that the proposed method can be successfully used for real search and rescue operations.

Robustness of optimal intermittent search strategies in one, two, and three dimensions

Search problems at various scales involve a searcher, be it a molecule before reaction or a foraging animal, which performs an intermittent motion. Here we analyze a generic model based on such type of intermittent motion, in which the searcher alternates phases of slow motion allowing detection and phases of fast motion without detection. We present full and systematic results for different modeling hypotheses of the detection mechanism in space in one, two, and three dimensions. Our study completes and extends the results of our recent letter [Loverdo, Nat. Phys. 4, 134 (2008)] and gives the necessary calculation details. In addition, another modeling of the detection case is presented. We show that the mean target detection time can be minimized as a function of the mean duration of each phase in one, two, and three dimensions. Importantly, this optimal strategy does not depend on the details of the modeling of the slow detection phase, which shows the robustness of our results. We believe that this systematic analysis can be used as a basis to study quantitatively various real search problems involving intermittent behaviors.

Robotic Exploration Utility for Urban Search and Rescue Tasks

In the 2006 RoboCup Virtual Rescue Robots competition, teams from different research labs developed methods for controlling teams of mobile robots in a simulated urban search and rescue scenario. The scoring procedure used in this inaugural competition rewards participants for the number of victims found, the amount of area explored in the environment, the quality of the maps created by the robot teams and penalties participants for colliding with a victim or relying on human operators. The analysis of the strategies and scores suggests that the scoring procedure may lead teams to adopt strategies that are not consistent with the needs of a real search and rescue scenario. This paper introduces Robotic Exploration Utility as a measure of exploration quality and analyzes the results of the competition based on this measure. Individual robot contributions to the system were reviewed to account for the costs associated with adding a robot to the environment, indicating that value added per robot is an important measure that is overlooked. The analysis also revealed substantial performance variation, depending on the behavior that was being rewarded, which may indicate a lack of focus for evaluative performance measures of robotic urban search and rescue systems. The Robotic Exploration Utility metric enables the research community to focus on a performance measure which reflects the needs of the domain, while allowing task performance to be easily compared across systems.

Evaluating a metadata‐based term suggestion interface for PubMed with real users with real requests

AbstractThis paper reports results of an evaluation study of MAP (Multi‐faceted Access to PubMed), a metadata induced query suggestion interface for PubMed bibliographic search.A novel evaluation methodology was used to address the challenges involved in evaluating an IIR (Interactive Information Retrieval) system such as the MAP interface. The most significant aspect of this methodology is that, instead of using assigned tasks common in traditional IR evaluation, it asks real users with real search requests to search with real systems in an experimental setting. Several performance measures were created based on which comparisons were made between MAP and PubMed baseline. MAP was shown to perform better in several of these measures, especially when the search requests had not been attempted before.The finding pointed to search characteristics as an important intervening variable in IIR evaluation. The advantages of and potential threats to our methodology were also discussed.

Identification of ambiguous queries in web search

It is widely believed that many queries submitted to search engines are inherently ambiguous (e.g., java and apple). However, few studies have tried to classify queries based on ambiguity and to answer “what the proportion of ambiguous queries is”. This paper deals with these issues. First, we clarify the definition of ambiguous queries by constructing the taxonomy of queries from being ambiguous to specific. Second, we ask human annotators to manually classify queries. From manually labeled results, we observe that query ambiguity is to some extent predictable. Third, we propose a supervised learning approach to automatically identify ambiguous queries. Experimental results show that we can correctly identify 87% of labeled queries with the approach. Finally, by using our approach, we estimate that about 16% of queries in a real search log are ambiguous.

Towards Deeper Understanding of the Search Interfaces of the Deep Web

Many databases have become Web-accessible through form-based search interfaces (i.e., HTML forms) that allow users to specify complex and precise queries to access the underlying databases. In general, such a Web search interface can be considered as containing an interface schema with multiple attributes and rich semantic/meta-information; however, the schema is not formally defined in HTML. Many Web applications, such as Web database integration and deep Web crawling, require the construction of the schemas. In this paper, we first propose a schema model for representing complex search interfaces, and then present a layout-expression based approach to automatically extract the logical attributes from search interfaces. We also rephrase the identification of different types of semantic information as a classification problem, and design several Bayesian classifiers to help derive semantic information from extracted attributes. A system, WISE-iExtractor, has been implemented to automatically construct the schema from any Web search interfaces. Our experimental results on real search interfaces indicate that this system is highly effective.

Three-Level Caching for Efficient Query Processing in Large Web Search Engines

Large web search engines have to answer thousands of queries per second with interactive response times. Due to the sizes of the data sets involved, often in the range of multiple terabytes, a single query may require the processing of hundreds of megabytes or more of index data. To keep up with this immense workload, large search engines employ clusters of hundreds or thousands of machines, and a number of techniques such as caching, index compression, and index and query pruning are used to improve scalability. In particular, two-level caching techniques cache results of repeated identical queries at the frontend, while index data for frequently used query terms are cached in each node at a lower level. We propose and evaluate a three-level caching scheme that adds an intermediate level of caching for additional performance gains. This intermediate level attempts to exploit frequently occurring pairs of terms by caching intersections or projections of the corresponding inverted lists. We propose and study several offline and online algorithms for the resulting weighted caching problem, which turns out to be surprisingly rich in structure. Our experimental evaluation based on a large web crawl and real search engine query log shows significant performance gains for the best schemes, both in isolation and in combination with the other caching levels. We also observe that a careful selection of cache admission and eviction policies is crucial for best overall performance.

Platform‐independent authoring and production of rapid e‐learning content

One of today’s major problems in the field of e‐learning is that the creation of high‐quality content is still rather time consuming and expensive. In the past, many efforts have been made to produce educational content on the fly, but the results were mainly static blocks of recorded lecture lacking sophisticated navigation facilities. Facing this challenge the authors developed the concept of hyper‐presentations. During the live presentation content‐ and time‐based metadata is captured and stored in a lightweight and player‐independent format. With this metadata powerful navigation facilities like real time navigation and full text search in audio or video data can be generated automatically. This improves flexibility and interoperability of technical solutions, which are both key factors in the emerging rapid e‐learning market.

Dutch real estate search engine found to provide lawful deep linking to websites of realtors

Dutch real estate search engine found to provide lawful deep linking to websites of realtors

Usability and user perceptions of a thesaurus‐enhanced search interface

PurposeThis paper seeks to report an investigation into the ways in which end‐users perceive a thesaurus‐enhanced search interface, in particular thesaurus and search interface usability.Design/methodology/approachThirty academic users, split between staff and postgraduate students, carrying out real search requests were observed during this study. Users were asked to comment on a range of thesaurus and interface characteristics including: ease of use, ease of learning, ease of browsing and navigation, problems and difficulties encountered while interacting with the system, and the effect of browsing on search term selection.FindingsThe results suggest that interface usability is a factor affecting thesaurus browsing/navigation and other information‐searching behaviours. Academic staff viewed the function of a thesaurus as being useful for narrowing down a search and providing alternative search terms, while postgraduates stressed the role of the thesaurus for broadening searches and providing new terms.Originality/valueThe paper provides an insight into the ways in which end‐users make use of and interact with a thesaurus‐enhanced search interface. This area is new since previous research has particularly focused on how professional searchers and librarians make use of thesauri and thesaurus‐enhanced search interfaces. The research reported here suggests that end‐users with varying levels of domain knowledge are able to use thesauri that are integrated into search interfaces. It also provides design implications for search interface developers as well as information professionals who are involved in teaching online searching.

인공지능을 이용한 3차원 구조물의 최적화 설계 : 마이크로 가속도계에 적용

본 논문은 실질적인 최적화 구조물의 설계를 위한 시스템에 대한 것으로 퍼지이론에 바탕을 둔 자동 유한요소 생성 망 기술과 계산 기하학적 기술, 해석코드 및 솔리드모델러를 시스템에 통합시켰다. 최적해 또는 만족해는 자동해석 시스템과 함께 탐색공간을 위한 유전자 알고리즘을 이용하여 자동적으로 탐색되어 진다. 또한, 유전자 알고리즘을 이용함으로써 본 설계 시스템은 다차원 해를 얻을 수 있다. 개발된 시스템은 터널전류에 바탕을 둔 마이크로 가속도계의 형상설계에 적용하였다. This paper describes an optimal design system for multi-disciplinary structural design. An automatic finite element (FE) mesh generation technique, which is based on the fuzzy knowledge processing and computational geometry technique, is incorporated into the system, together with a commercial FE analysis code and a commercial solid modelers. An optimum design solution or satisfactory solutions are then automatically searched using the genetic algorithms modified for real search space, together with the automated FE analysis system. With an aid of genetic algorithms, the present design system allows us to effectively obtain a multi-dimensional solutions. The developed system is successfully applied to the shape design of a micro accelerometer based on a tunnel current concept.

ALGORITHMS FOR SEARCHING FOR GAMMA-GRAVITY CORRELATIONS

We report the development of algorithms for searching for possible correlations between the data from gravitational-wave and gamma-ray detectors due to common astrophysical sources. We account for the time delay between the two signals and also use information on the detailed time structure of individual gamma-ray bursts (GRBs). Three possible algorithms of data processing are suggested with different degrees of noise smoothing. The algorithms have been tested using real Burst and Transient Search Experiment (BATSE) GRB data and records of the cryogenic resonant bar detector Explorer. The expected filtering properties of the algorithms are confirmed by the tests. The use of more extended data records is required to detect any correlations or to establish upper limits. A possible generalization for non-Gaussian noises is discussed.

Improved sampling of the pareto-front in multiobjective genetic optimizations by steady-state evolution: a pareto converging genetic algorithm.

Previous work on multiobjective genetic algorithms has been focused on preventing genetic drift and the issue of convergence has been given little attention. In this paper, we present a simple steady-state strategy, Pareto Converging Genetic Algorithm (PCGA), which naturally samples the solution space and ensures population advancement towards the Pareto-front. PCGA eliminates the need for sharing/niching and thus minimizes heuristically chosen parameters and procedures. A systematic approach based on histograms of rank is introduced for assessing convergence to the Pareto-front, which, by definition, is unknown in most real search problems. We argue that there is always a certain inheritance of genetic material belonging to a population, and there is unlikely to be any significant gain beyond some point; a stopping criterion where terminating the computation is suggested. For further encouraging diversity and competition, a nonmigrating island model may optionally be used; this approach is particularly suited to many difficult (real-world) problems, which have a tendency to get stuck at (unknown) local minima. Results on three benchmark problems are presented and compared with those of earlier approaches. PCGA is found to produce diverse sampling of the Pareto-front without niching and with significantly less computational effort.