Query Logs Research Articles

Impact of response latency on sponsored search

Recent research in the human computer interaction and information retrieval areas has revealed that search response latency exhibits a clear impact on the user behavior in web search. Such impact is reflected both in users’ subjective perception of the usability of a search engine and in their interaction with the search engine in terms of the number of search results they engage with. However, a similar impact analysis has been missing so far in the context of sponsored search. Since the predominant business model for commercial search engines is advertising via sponsored search results (i.e., search advertisements), understanding how response latency influences the user interaction with the advertisements displayed on the search engine result pages is crucial to increase the revenue of a commercial search engine. To this end, we conduct a large-scale analysis using query logs obtained from a commercial web search. We analyze the short-term and long-term impact of search response latency on the querying and clicking behaviors of users using desktop and mobile devices to access the search engine, as well as the corresponding impact on the revenue of the search engine. This analysis demonstrates the importance of serving sponsored search results with low latency and provides insight into the ad serving policy of commercial search engines to ensure long-term user engagement and search revenue.

Location-aware query reformulation for search engines

Query reformulation, including query recommendation and query auto-completion, is a popular add-on feature of search engines, which provide related and helpful reformulations of a keyword query. Due to the dropping prices of smartphones and the increasing coverage and bandwidth of mobile networks, a large percentage of search engine queries are issued from mobile devices. This makes it possible to improve the quality of query recommendation and auto-completion by considering the physical locations of the query issuers. However, limited research has been done on location-aware query reformulation for search engines. In this paper, we propose an effective spatial proximity measure between a query issuer and a query with a location distribution obtained from its clicked URLs in the query history. Based on this, we extend popular query recommendation and auto-completion approaches to our location-aware setting, which suggest query reformulations that are semantically relevant to the original query and give results that are spatially close to the query issuer. In addition, we extend the bookmark coloring algorithm for graph proximity search to support our proposed query recommendation approaches online, and we adapt an A* search algorithm to support our query auto-completion approach. We also propose a spatial partitioning based approximation that accelerates the computation of our proposed spatial proximity. We conduct experiments using a real query log, which show that our proposed approaches significantly outperform previous work in terms of quality, and they can be efficiently applied online.

Summarizing database schema based on graph partition

As the underlying database schemas become larger and more complex, it is difficult for casual users to understand the schemas and contents of databases. Therefore, it has become an essential task to summarize the database schemas. However, most prior approaches pay little attention to the topological characteristics between tables, ignore the effect of the user feedback, and fail to accurately predict the number of clusters in the output. This seriously limits their accuracy of schema summarization. To deal with the problems, we propose a new schema summarization method based on a graph partition mechanism. First, we introduce a novel strategy to construct a similarity matrix between tables, which is based on the topology compactness, content similarity and query logs. Then we provide a calculation formula for table importance and a detection scheme of the most important nodes in local areas. Both are used for selecting the initial cluster centers and predicting the number of clusters in the graph partition mechanism. Finally, we evaluate the proposed method over the database TPC-E, and results demonstrate that it achieves high performance in summarizing accuracy.

Cross-Media Similarity Evaluation for Web Image Retrieval in the Wild

In order to retrieve unlabeled images by textual queries, cross-media similarity computation is a key ingredient. Although novel methods are continuously introduced, little has been done to evaluate these methods together with large-scale query log analysis. Consequently, how far have these methods brought us in answering real-user queries is unclear. Given baseline methods that use relatively simple text/image matching, how much progress have advanced models made is also unclear. This paper takes a pragmatic approach to answering the two questions. Queries are automatically categorized according to the proposed query visualness measure and later connected to the evaluation of multiple cross-media similarity models on three test sets. Such a connection reveals that the success of the state of the art is mainly attributed to their good performance on visual-oriented queries, which account for only a small part of real-user queries. To quantify the current progress, we propose a simple text2image method, representing a novel query by a set of images selected from large-scale query log. Consequently, computing cross-media similarity between the query and a given image boils down to comparing the visual similarity between the given image and the selected images. Image retrieval experiments on the challenging Clickture dataset show that the proposed text2image is a strong baseline, comparing favorably to recent deep learning alternatives.

Entity-Based Query Recommendation for Long-Tail Queries

Query recommendation, which suggests related queries to search engine users, has attracted a lot of attention in recent years. Most of the existing solutions, which perform analysis of users’ search history (or query logs ), are often insufficient for long-tail queries that rarely appear in query logs. To handle such queries, we study the use of entities found in queries to provide recommendations. Specifically, we extract entities from a query, and use these entities to explore new ones by consulting an information source. The discovered entities are then used to suggest new queries to the user. In this article, we examine two information sources: (1) a knowledge base (or KB), such as YAGO and Freebase; and (2) a click log, which contains the URLs accessed by a query user. We study how to use these sources to find new entities useful for query recommendation. We further study a hybrid framework that integrates different query recommendation methods effectively. As shown in the experiments, our proposed approaches provide better recommendations than existing solutions for long-tail queries. In addition, our query recommendation process takes less than 100ms to complete. Thus, our solution is suitable for providing online query recommendation services for search engines.

Discovering Popular Clicks' Pattern of Teen Users for Query Recommendation

Search engines are still the most important gates for information search in internet. In this regard, providing the best response in the shortest time possible to the user's request is still desired. Normally, search engines are designed for adults and few policies have been employed considering teen users. Teen users are more biased in clicking the results list than are adult users. This leads to fewer clicks on the lowly-ranked search results. Such behavior reduces teen users’ navigation and result extraction skills. With an increase in information load and in teen’s demands, lack of efficient methods leads to inefficiency of search engines regarding teen users. For the purpose, this study discovers teen users’ search behavior and its application in yielding an improved search is strongly recommended. In this way, the pattern of teen users’ popular clicks is identified from a large search log through mining of users’ search transactions based on the frequency and similarity of the clicks in the search log. Then, using binary classification, the closest query into the teen user’s desired one is identified. To discover teen users’ behavior, we took advantage of the AOL query log. System efficiency was examined on the AOL query search log. Results reveal that click pattern improves approaching the query to the one desired by teen users. Generally, this study can demonstrate that in data recovery, application of click behavior and its binary classification can result in improved access of teen users to their desired results.

User Queries of the Online Standard Slovenian Dictionary and 2001 Slovenian Normative Guide at bos.zrc-sazu.si

This article analyzes the user query logs of the online versions of two Slovenian dictionaries—Slovar slovenskega knjižnega jezika (Standard Slovenian Dictionary, SSKJ) and Slovenski pravopis 2001 (2001 Slovenian Normative Guide, SP 2001)—at bos.zrc-sazu. si (BOS), which for almost a whole decade (2000–2010) served as a portal for searching through various text collections and dictionaries prepared or adapted for the online environment by the ZRC SAZU Fran Ramovš Institute of the Slovenian Language in Ljubljana. We focused on unsuccessful queries. It turned out that for new-generation dictionaries—that is, dictionaries primarily designed for online use—users need search engines with hints, search engines with corrections for unsuccessful queries, and combined search engines for various dictionaries, and in the dictionary itself they need information that is as complete and unabridged as possible.

Modeling behaviors and lifestyle with online and social data for predicting and analyzing sleep and exercise quality

While recent data studies have focused on associations between sleep and exercise patterns as captured by digital fitness devices, it is known that sleep and exercise quality are affected by a much broader set of factors not captured by these devices, such as general lifestyle, eating, and stress. Here, we conduct a large-scale data study of exercise and sleep effects through an analysis of 8 months of exercise and sleep data for 20 k users, combined with search query logs, location information and aggregated social media data. We analyze factors correlated with better sleep and more effective exercise, and confirm these relationships through causal inference analysis. Further, we build linear models to predict individuals’ sleep and exercise quality. This analysis demonstrates the potential benefits of combining online and social data sources with data from health trackers, and is a potentially rich computational benchmark for health studies. We discuss the implications of our work for individuals, health practitioners and health systems.

Research on Web Page Classification Method Based on Query Log

Web page classification is an important application in many fields of Internet information retrieval, such as providing directory classification and vertical search. Methods based on query log which is a light weight version of Web page classification can avoid Web content crawling, making it relatively high in efficiency, but the sparsity of user click data makes it difficult to be used directly for constructing a classifier. To solve this problem, we explore the semantic relations among different queries through word embedding, and propose three improved graph structure classification algorithms. To reflect the semantic relevance between queries, we map the user query into the low-dimensional space according to its query vector in the first step. Then, we calculate the uniform resource locator (URL) vector according to the relationship between the query and URL. Finally, we use the improved label propagation algorithm (LPA) and the bipartite graph expansion algorithm to classify the unlabeled Web pages. Experiments show that our methods make about 20% more increase in F1-value than other Web page classification methods based on query log.

A semantic-preserving differentially private method for releasing query logs

Query logs are of great interest for data analysis. They allow characterizing user profiles, user behaviors and search habits. However, since query logs usually contain personal information, data controllers should implement appropriate data protection mechanisms before releasing them for secondary use. In the past, the anonymization of query logs was tackled from the perspective of statistical disclosure control and by relying on privacy models such as k-anonymity, which do not scale well with the high dimensionality and dynamicity of query logs. To offer better privacy protection, some authors have recently embraced the robust privacy guarantees of ɛ-differential privacy. However, this comes at the cost of limiting the number and types of analyses that can be made on the protected queries. To tackle this issue, in this paper we propose a privacy protection method for query logs that joins the flexibility and convenience of privacy-preserving data releases with the strong privacy guarantees of ɛ-differential privacy. Moreover, to retain the analytical utility of the protected query, we have put special care in capturing, managing and preserving the semantics of the queries during the protection process. The empirical experiments we report show that our method produces differentially private query logs that are more useful for analysis than related works.

User Generated SQL Queries Inform Evaluation of NSSP ESSENCE

Objective: The objective of this work is to describe the use and performance of the NSSP ESSENCE system by analyzing the structured query language (SQL) logs generated by users of the National Syndromic Surveillance Program’s (NSSP) Electronic Surveillance System for the Early Notification of Community-based Epidemics (ESSENCE).Introduction: As system users develop queries within ESSENCE, they step through the user-interface to select data sources and parameters needed for their query. Then they select from the available output options (e.g., time series, table builder, data details). These activities execute a SQL query on the database, the majority of which are saved in a log so that system developers can troubleshoot problems. Secondarily, these data can be used as a form of web analytics to describe user query choices, query volume, query execution time, and develop an understanding of ESSENCE query patterns.Methods: ESSENCE SQL query logs were extracted from April 1, 2016 to August 23th, 2017. Overall query volume was assessed by summarizing volume of queries over time (e.g., by hour, day, and week), and by Site. To better understand system performance the mean, median, and maximum query execution times were summarized over time and by Site. SQL query text was parsed so that we could isolate, 1) Syndromes queried, 2) Sub-syndromes queried, 3) Keyword categories queried, and 4) Free text query terms used. Syndromes, sub-syndromes, and keyword categories were tabulated in total and by Site. Frequencies of free text query terms were analyzed using n-grams, wordclouds, and term co-occurrence relationships. Term co-occurrence network graphs were used to visualize the structure and relationships among terms.Results: There were a total of 354,101 SQL queries generated by users of ESSENCE between April 1, 2016 and August 23rd, 2017. Over this entire time period there was a weekly mean of 4,785 SQL queries performed by users. When looking at 2017 data through August 23rd this figure increases to a mean of 7,618 SQL queries per week for 2017, and since May 2017 the mean number of SQL queries has increased to 10,485 per week. The maximum number of user generated SQL queries in a week was 29,173. The mean, median, and maximum query execution times for all data was 0.61 minutes, 0 minutes, and 365 minutes, respectively. When looking at only queries with a free text component the mean query execution time increases slightly to 0.94 minutes, though the median is still 0 minutes. The peak usage period based on number of SQL queries performed is between 12:00pm and 3:00pm EST.Conclusions: The use of NSSP ESSENCE has grown since implementation. This is the first time the ESSENCE system has been used at a National level with this volume of data, and number of users. Our focus to date has been on successfully on-boarding new Sites so that they can benefit from use of the available tools, providing trainings to new users, and optimizing ESSENCE performance. Routine analysis of the ESSENCE SQL logs can assist us in understanding how the system is being used, how well it is performing, and in evaluating our system optimization efforts.

The Effect of Big Data on Recommendation Quality. The Example of Internet Search

Are there economies of scale to data in internet search? This paper is first to use real search engine query logs to empirically investigate how data drives the quality of internet search results. We find evidence that the quality of search results improve with more data on previous searches. Moreover, our results indicate that the type of data matters as well: personalized information is particularly valuable as it massively increases the speed of learning. We also provide some evidence that factors not directly related to data such as the general quality of the applied algorithms play an important role. The suggested methods to disentangle the effect of data from other factors driving the quality of search results can be applied to assess the returns to data in various recommendation systems in e-commerce, including product and information search. We also discuss the managerial, privacy, and competition policy implications of our findings.

Search engines, news wires and digital epidemiology: Presumptions and facts

BackgroundDigital epidemiology tries to identify diseases dynamics and spread behaviors using digital traces collected via search engines logs and social media posts. However, the impacts of news on information-seeking behaviors have been remained unknown. MethodsData employed in this research provided from two sources, (1) Parsijoo search engine query logs of 48 months, and (2) a set of documents of 28 months of Parsijoo’s news service. Two classes of topics, i.e. macro-topics and micro-topics were selected to be tracked in query logs and news. Keywords of the macro-topics were automatically generated using web provided resources and exceeded 10k. Keyword set of micro-topics were limited to a numerable list including terms related to diseases and health-related activities. The tests are established in the form of three studies. Study A includes temporal analyses of 7 macro-topics in query logs. Study B considers analyzing seasonality of searching patterns of 9 micro-topics, and Study C assesses the impact of news media coverage on users’ health-related information-seeking behaviors. ResultsStudy A showed that the hourly distribution of various macro-topics followed the changes in social activity level. Conversely, the interestingness of macro-topics did not follow the regulation of topic distributions. Among macro-topics, “Pharmacotherapy” has highest interestingness level and wider time-window of popularity. In Study B, seasonality of a limited number of diseases and health-related activities were analyzed. Trends of infectious diseases, such as flu, mumps and chicken pox were seasonal. Due to seasonality of most of diseases covered in national vaccination plans, the trend belonging to “Immunization and Vaccination” was seasonal, as well. Cancer awareness events caused peaks in search trends of “Cancer” and “Screening” micro-topics in specific days of each year that mimic repeated patterns which may mistakenly be identified as seasonality. In study C, we assessed the co-integration and correlation between news and query trends. Our results demonstrated that micro-topics sparsely covered in news media had lowest level of impressiveness and, subsequently, the lowest impact on users’ intents. ConclusionOur results can reveal public reaction to social events, diseases and prevention procedures. Furthermore, we found that news trends are co-integrated with search queries and are able to reveal health-related events; however, they cannot be used interchangeably. It is recommended that the user-generated contents and news documents are analyzed mutually and interactively.

Query Auto-Completion Based on Word2vec Semantic Similarity

Query auto-completion (QAC) is the first step of information retrieval, which helps users formulate the entire query after inputting only a few prefixes. Regarding the models of QAC, the traditional method ignores the contribution from the semantic relevance between queries. However, similar queries always express extremely similar search intention. In this paper, we propose a hybrid model FS-QAC based on query semantic similarity as well as the query frequency. We choose word2vec method to measure the semantic similarity between intended queries and pre-submitted queries. By combining both features, our experiments show that FS-QAC model improves the performance when predicting the user’s query intention and helping formulate the right query. Our experimental results show that the optimal hybrid model contributes to a 7.54% improvement in terms of MRR against a state-of-the-art baseline using the public AOL query logs.

Personalized query suggestion based on user behavior

Query suggestions help users refine their queries after they input an initial query. Previous work mainly concentrated on similarity-based and context-based query suggestion approaches. However, models that focus on adapting to a specific user (personalization) can help to improve the probability of the user being satisfied. In this paper, we propose a personalized query suggestion model based on users’ search behavior (UB model), where we inject relevance between queries and users’ search behavior into a basic probabilistic model. For the relevance between queries, we consider their semantical similarity and co-occurrence which indicates the behavior information from other users in web search. Regarding the current user’s preference to a query, we combine the user’s short-term and long-term search behavior in a linear fashion and deal with the data sparse problem with Bayesian probabilistic matrix factorization (BPMF). In particular, we also investigate the impact of different personalization strategies (the combination of the user’s short-term and long-term search behavior) on the performance of query suggestion reranking. We quantify the improvement of our proposed UB model against a state-of-the-art baseline using the public AOL query logs and show that it beats the baseline in terms of metrics used in query suggestion reranking. The experimental results show that: (i) for personalized ranking, users’ behavioral information helps to improve query suggestion effectiveness; and (ii) given a query, merging information inferred from the short-term and long-term search behavior of a particular user can result in a better performance than both plain approaches.

Application of hybrid of Fuzzy Set, Trust and Genetic Algorithm in query log mining for effective Information Retrieval

The precision of Information Retrieval (IR) System is low due to imprecise user queries as well as because of information overload on web. The Fuzzy set infers the user’s information need from vague and imprecise queries and web recommender systems are used to overcome information overload problem. The performance of recommender system is still low due to data sparsity. The concept of trust is used to deal with data sparseness problem and improves the performance of recommender system. Optimization techniques like Genetic Algorithm(GA) have been applied in domain of information retrieval for effective web search. In this research hybrid of Fuzzy set, GA and Trust has been used together in query log mining for personalized web search based on using fuzzy queries for recommendation of optimal set of trusted documents. Thus the use of hybrid of Fuzzy set, trust and GA together infer the user’s information need from vague and imprecise user’s queries and optimize the web page ranking of trusted web pages for effective personalized web search. The experimental results were analyzed statistically as well as compared with GA IR, and Fuzzy Trust based IR. Hence based on comparative analysis of results, thus hybrid of Fuzzy Set, Trust and GA shows the improvement in average precision of search results and confirms the effective personalization of web search.

The Characteristics of Voice Search

The growing popularity of mobile search and the advancement in voice recognition technologies have opened the door for web search users to speak their queries rather than type them. While this kind of voice search is still in its infancy, it is gradually becoming more widespread. In this article, we report a comprehensive voice search query log analysis of a commercial web search engine’s mobile application. We compare voice and text search by various aspects, with special focus on the semantic and syntactic characteristics of the queries. Our analysis suggests that voice queries focus more on audio-visual content and question answering and less on social networking and adult domains. In addition, voice queries are more commonly submitted on the go. We also conduct an empirical evaluation showing that the language of voice queries is closer to natural language than the language of text queries. Our analysis points out further differences between voice and text search. We discuss the implications of these differences for the design of future voice-enabled web search tools.

LAILAPS-QSM: A RESTful API and JAVA library for semantic query suggestions.

In order to access and filter content of life-science databases, full text search is a widely applied query interface. But its high flexibility and intuitiveness is paid for with potentially imprecise and incomplete query results. To reduce this drawback, query assistance systems suggest those combinations of keywords with the highest potential to match most of the relevant data records. Widespread approaches are syntactic query corrections that avoid misspelling and support expansion of words by suffixes and prefixes. Synonym expansion approaches apply thesauri, ontologies, and query logs. All need laborious curation and maintenance. Furthermore, access to query logs is in general restricted. Approaches that infer related queries by their query profile like research field, geographic location, co-authorship, affiliation etc. require user’s registration and its public accessibility that contradict privacy concerns. To overcome these drawbacks, we implemented LAILAPS-QSM, a machine learning approach that reconstruct possible linguistic contexts of a given keyword query. The context is referred from the text records that are stored in the databases that are going to be queried or extracted for a general purpose query suggestion from PubMed abstracts and UniProt data. The supplied tool suite enables the pre-processing of these text records and the further computation of customized distributed word vectors. The latter are used to suggest alternative keyword queries. An evaluated of the query suggestion quality was done for plant science use cases. Locally present experts enable a cost-efficient quality assessment in the categories trait, biological entity, taxonomy, affiliation, and metabolic function which has been performed using ontology term similarities. LAILAPS-QSM mean information content similarity for 15 representative queries is 0.70, whereas 34% have a score above 0.80. In comparison, the information content similarity for human expert made query suggestions is 0.90. The software is either available as tool set to build and train dedicated query suggestion services or as already trained general purpose RESTful web service. The service uses open interfaces to be seamless embeddable into database frontends. The JAVA implementation uses highly optimized data structures and streamlined code to provide fast and scalable response for web service calls. The source code of LAILAPS-QSM is available under GNU General Public License version 2 in Bitbucket GIT repository: https://bitbucket.org/ipk_bit_team/bioescorte-suggestion

Learning a unified embedding space of web search from large-scale query log

In the procedure of Web search, a user first comes up with an information need and a query is issued with the need as guidance. After that, some URLs are clicked and other queries may be issued if those URLs do not meet his need well. We advocate that Web search is governed by a unified hidden space, and each involved element such as query and URL has its inborn position, i.e., projected as a vector, in this space. Each of above actions in the search procedure, i.e. issuing queries or clicking URLs, is an interaction result of those elements in the space. In this paper, we aim at uncovering such a unified hidden space of Web search that uniformly captures the hidden semantics of search queries, URLs and other involved elements in Web search. We learn the semantic space with search session data, because a search session can be regarded as an instantiation of users’ information need on a particular semantic topic and it keeps the interaction information of queries and URLs. We use a set of session graphs to represent search sessions, and the space learning task is cast as a vector learning problem for the graph vertices by maximizing the log-likelihood of a training session data set. Specifically, we developed the well-known Word2vec to perform the learning procedure. Experiments on the query log data of a commercial search engine are conducted to examine the efficacy of learnt vectors, and the results show that our framework is helpful for different finer tasks in Web search.

Cleaning Antipatterns in an SQL Query Log

Today, many scientific data sets are open to the public. For their operators, it is important to know what the users are interested in. In this paper, we study the problem of extracting and analyzing patterns from the query log of a database. We focus on design errors (antipatterns), which typically lead to unnecessary SQL statements. Such antipatterns do not only have a negative effect on performance. They also introduce bias on any subsequent analysis of the SQL log. We propose a framework designed to discover patterns and antipatterns in arbitrary SQL query logs and to clean antipatterns. To study the usefulness of our approach and to reveal insights regarding the existence of antipatterns in real-world systems, we examine the SQL log of the SkyServer project, containing more than 40 million queries. Among the top 15 patterns, we have found six antipatterns. This result as well as other ones gives way to the conclusion that antipatterns might falsify refactoring and any other downstream analyses.