Reranking web search results for diversity

Search engine results are often biased towards a certain aspect of a query or towards a certain meaning for ambiguous query terms. Diversification of search results offers a way to supply the user with a better balanced result set increasing the probability that a user finds at least one document suiting her information need. In this paper, we present a reranking approach based on minimizing variance of Web search results to improve topic coverage in the top-k results. We investigate two different document representations as the basis for reranking. Smoothed language models and topic models derived by Latent Dirichlet?allocation. To evaluate our approach we selected 240 queries from Wikipedia disambiguation pages. This provides us with ambiguous queries together with a community generated balanced representation of their (sub)topics. For these queries we crawled two major commercial search engines. In addition, we present a new evaluation strategy based on Kullback-Leibler divergence and Wikipedia. We evaluate this method using the TREC sub-topic evaluation on the one hand, and manually annotated query results on the other hand. Our results show that minimizing variance in search results by reranking relevant pages significantly improves topic coverage in the top-k results with respect to Wikipedia, and gives a good overview of the overall search result. Moreover, latent topic models achieve competitive diversification with significantly less reranking. Finally, our evaluation reveals that our automatic evaluation strategy using Kullback-Leibler divergence correlates well with α-nDCG scores used in manual evaluation efforts.     

Coverage-based search result diversification

Traditional retrieval models may provide users with less satisfactory search experience because documents are scored independently and the top ranked documents often contain excessively redundant information. Intuitively, it is more desirable to diversify search results so that the top-ranked documents can cover different query subtopics, i.e., different pieces of relevant information. In this paper, we study the problem of search result diversification in an optimization framework whose objective is to maximize a coverage-based diversity function. We first define the diversity score of a set of search results through measuring the coverage of query subtopics in the result set, and then discuss how to use them to derive diversification methods. The key challenge here is how to define an appropriate coverage function given a query and a set of search results. To address this challenge, we propose and systematically study three different strategies to define coverage functions. They are based on summations, loss functions and evaluation measures respectively. Each of these coverage functions leads to a result diversification method. We show that the proposed coverage based diversification methods not only cover several state-of-the-art methods but also allows us to derive new ones. We compare these methods both analytically and empirically. Experiment results on two standard TREC collections show that all the methods are effective for diversification and the new methods can outperform existing ones.     

Non-english web search: an evaluation of indexing and searching the Greek web

The study reports on a longitudinal and comparative evaluation of Greek language searching on the web. Ten engines, five global (A9, AltaVista, Google, MSN Search, and Yahoo!) and five Greek (Anazitisi, Ano-Kato, Phantis. Trinity, and Visto), were evaluated using (a) navigational queries in 2004 and 2006; and (b) by measuring the freshness of the search engine indices in 2005 and 2006. Homepage finding queries for known Greek organizations were created and searched. Queries included the name of the organization in its Greek and non-Greek, English or transliterated equivalent forms. The organizations represented ten categories: government departments, universities, colleges, travel agencies, museums, media (TV, radio, newspapers), transportation, and banks. The freshness of the indices was evaluated by examining the status of the returned URLs (live versus dead) from the navigational queries, and by identifying if the engines have indexed 32480 active (live) Greek domain URLs. Effectiveness measures included (a) qualitative assessment of how engines handle the Greek language; (b) precision at 10 documents (P@10); (c) mean reciprocal rank (MRR); (d) Navigational Query Discounted Cumulative Gain (NQ-DCG), a new heuristic evaluation measure; (e) response time; (f) the ratio of the dead URL links returned, (g) the presence or absence of URLs and the decay observed over the period of the study. The results report on which of the global and Greek search engines perform best; and if the performance achieved is good enough from a user’s perspective.     

Incorporating web browsing activities into anchor texts for web search

Anchor texts complement Web page content and have been used extensively in commercial Web search engines. Existing methods for anchor text weighting rely on the hyperlink information which is created by page content editors. Since anchor texts are created to help user browse the Web, browsing behavior of Web users may also provide useful or complementary information for anchor text weighting. In this paper, we discuss the possibility and effectiveness of incorporating browsing activities of Web users into anchor texts for Web search. We first make an analysis on the effectiveness of anchor texts with browsing activities. And then we propose two new anchor models which incorporate browsing activities. To deal with the data sparseness problem of user-clicked anchor texts, two features of user’s browsing behavior are explored and analyzed. Based on these features, a smoothing method for the new anchor models is proposed. Experimental results show that by incorporating browsing activities the new anchor models outperform the state-of-art anchor models which use only the hyperlink information. This study demonstrates the benefits of Web browsing activities to affect anchor text weighting.     

On the role of novelty for search result diversification

Re-ranking the search results in order to promote novel ones has traditionally been regarded as an intuitive diversification strategy. In this paper, we challenge this common intuition and thoroughly investigate the actual role of novelty for search result diversification, based upon the framework provided by the diversity task of the TREC 2009 and 2010 Web tracks. Our results show that existing diversification approaches based solely on novelty cannot consistently improve over a standard, non-diversified baseline ranking. Moreover, when deployed as an additional component by the current state-of-the-art diversification approaches, our results show that novelty does not bring significant improvements, while adding considerable efficiency overheads. Finally, through a comprehensive analysis with simulated rankings of various quality, we demonstrate that, although inherently limited by the performance of the initial ranking, novelty plays a role at breaking the tie between similarly diverse results.     

Enhancing web search with queries of equivalent intents

Users often issue all kinds of queries to look for the same target due to the intrinsic ambiguity and flexibility of natural languages. Some previous work clusters queries based on co-clicks; however, the intents of queries in one cluster are not that similar but roughly related. It is desirable to conduct automatic mining of queries with equivalent intents from a large scale search logs. In this paper, we take account of similarities between query strings. There are two issues associated with such similarities: it is too costly to compare any pair of queries in large scale search logs, and two queries with a similar formulation, such as “SVN” (Apache Subversion) and support vector machine (SVM), are not necessarily similar in their intents. To address these issues, we propose using the similarities of query strings above the co-click based clustering results. Our method improves precision over the co-click based clustering method (lifting precision from 0.37 to 0.62), and outperforms a commercial search engine’s query alteration (lifting \(F_1\) measure from 0.42 to 0.56). As an application, we consider web document retrieval. We aggregate similar queries’ click-throughs with the query’s click-throughs and evaluate them on a large scale dataset. Experimental results indicate that our proposed method significantly outperforms the baseline method of using a query’s own click-throughs in all metrics.     

Search markets and search results: The case of Bing

Bing and Google customize their results to target people with different geographic locations and languages but, despite the importance of search engines for web users and webometric research, the extent and nature of these differences are unknown. This study compares the results of seventeen random queries submitted automatically to Bing for thirteen different English geographic search markets at monthly intervals. Search market choice alters a small majority of the top 10 results but less than a third of the complete sets of results. Variation in the top 10 results over a month was about the same as variation between search markets but variation over time was greater for the complete results sets. Most worryingly for users, there were almost no ubiquitous authoritative results: only one URL was always returned in the top 10 for all search markets and points in time, and Wikipedia was almost completely absent from the most common top 10 results. Most importantly for webometrics, results from at least three different search markets should be combined to give more reliable and comprehensive results, even for queries that return fewer than the maximum number of URLs.     

Click-based evidence for decaying weight distributions in search effectiveness metrics

Search effectiveness metrics are used to evaluate the quality of the answer lists returned by search services, usually based on a set of relevance judgments. One plausible way of calculating an effectiveness score for a system run is to compute the inner-product of the run’s relevance vector and a “utility” vector, where the ith element in the utility vector represents the relative benefit obtained by the user of the system if they encounter a relevant document at depth i in the ranking. This paper uses such a framework to examine the user behavior patterns—and hence utility weightings—that can be inferred from a web query log. We describe a process for extrapolating user observations from query log clickthroughs, and employ this user model to measure the quality of effectiveness weighting distributions. Our results show that for measures with static distributions (that is, utility weighting schemes for which the weight vector is independent of the relevance vector), the geometric weighting model employed in the rank-biased precision effectiveness metric offers the closest fit to the user observation model. In addition, using past TREC data as to indicate likelihood of relevance, we also show that the distributions employed in the BPref and MRR metrics are the best fit out of the measures for which static distributions do not exist.     

The whens and hows of learning to rank for web search

Web search engines are increasingly deploying many features, combined using learning to rank techniques. However, various practical questions remain concerning the manner in which learning to rank should be deployed. For instance, a sample of documents with sufficient recall is used, such that re-ranking of the sample by the learned model brings the relevant documents to the top. However, the properties of the document sample such as when to stop ranking—i.e. its minimum effective size—remain unstudied. Similarly, effective listwise learning to rank techniques minimise a loss function corresponding to a standard information retrieval evaluation measure. However, the appropriate choice of how to calculate the loss function—i.e. the choice of the learning evaluation measure and the rank depth at which this measure should be calculated—are as yet unclear. In this paper, we address all of these issues by formulating various hypotheses and research questions, before performing exhaustive experiments using multiple learning to rank techniques and different types of information needs on the ClueWeb09 and LETOR corpora. Among many conclusions, we find, for instance, that the smallest effective sample for a given query set is dependent on the type of information need of the queries, the document representation used during sampling and the test evaluation measure. As the sample size is varied, the selected features markedly change—for instance, we find that the link analysis features are favoured for smaller document samples. Moreover, despite reflecting a more realistic user model, the recently proposed ERR measure is not as effective as the traditional NDCG as a learning loss function. Overall, our comprehensive experiments provide the first empirical derivation of best practices for learning to rank deployments.     

A study of medical and health queries to web search engines

This paper reports findings from an analysis of medical or health queries to different web search engines. We report results: (i). comparing samples of 10000 web queries taken randomly from 1.2 million query logs from the AlltheWeb.com and Excite.com commercial web search engines in 2001 for medical or health queries, (ii). comparing the 2001 findings from Excite and AlltheWeb.com users with results from a previous analysis of medical and health related queries from the Excite Web search engine for 1997 and 1999, and (iii). medical or health advice-seeking queries beginning with the word 'should'. Findings suggest: (i). a small percentage of web queries are medical or health related, (ii). the top five categories of medical or health queries were: general health, weight issues, reproductive health and puberty, pregnancy/obstetrics, and human relationships, and (iii). over time, the medical and health queries may have declined as a proportion of all web queries, as the use of specialized medical/health websites and e-commerce-related queries has increased. Findings provide insights into medical and health-related web querying and suggests some implications for the use of the general web search engines when seeking medical/health information.     

Quantum algorithms: A?survey of some recent results

Quantum algorithms are a field of growing interest within the theoretical computer science as well as the physics community. Surprisingly, although the number of researchers working on the subject is ever-increasing, the number of quantum algorithms found so far is quite small. In fact, the task of designing new quantum algorithms has been proven to be extremely difficult. In this paper we give an overview of the known quantum algorithms and briefly describe the underlying ideas. Roughly, the algorithms presented are divided into hidden subgroup type algorithms and in amplitude amplification type algorithms. While the former deal with problems of group-theoretical nature and have the promise to yield strong separations of classical and quantum algorithms, the latter have been proved to be a prolific source of algorithms in which a polynomial speed-up as compared to classical algorithms can be achieved. We also discuss quantum algorithms which do not fall under these two categories and give a survey of techniques of general interest in quantum computing such as adiabatic computing, lower bounds for quantum algorithms, and quantum interactive proofs.     

Quantum algorithms: A?survey of some recent results

Quantum algorithms are a field of growing interest within the theoretical computer science as well as the physics community. Surprisingly, although the number of researchers working on the subject is ever-increasing, the number of quantum algorithms found so far is quite small. In fact, the task of designing new quantum algorithms has been proven to be extremely difficult. In this paper we give an overview of the known quantum algorithms and briefly describe the underlying ideas. Roughly, the algorithms presented are divided into hidden subgroup type algorithms and in amplitude amplification type algorithms. While the former deal with problems of group-theoretical nature and have the promise to yield strong separations of classical and quantum algorithms, the latter have been proved to be a prolific source of algorithms in which a polynomial speed-up as compared to classical algorithms can be achieved. We also discuss quantum algorithms which do not fall under these two categories and give a survey of techniques of general interest in quantum computing such as adiabatic computing, lower bounds for quantum algorithms, and quantum interactive proofs.     

Enhancing web search in the medical domain via query clarification

The majority of Internet users search for medical information online; however, many do not have an adequate medical vocabulary. Users might have difficulties finding the most authoritative and useful information because they are unfamiliar with the appropriate medical expressions describing their condition; consequently, they are unable to adequately satisfy their information need. We investigate the utility of bridging the gap between layperson and expert vocabularies; our approach adds the most appropriate expert expression to queries submitted by users, a task we call query clarification. We evaluated the impact of query clarification. Using three different synonym mappings and conducting two task-based retrieval studies, users were asked to answer medically-related questions using interleaved results from a major search engine. Our results show that the proposed system was preferred by users and helped them answer medical concerns correctly more often, with up to a 7 % increase in correct answers over an unmodified query. Finally, we introduce a supervised classifier to select the most appropriate synonym mapping for each query, which further increased the fraction of correct answers (12 %).     

基于语义网的智能搜索引擎在数字图书馆中的应用

针对数字图书馆现有搜索引擎检索所出现的检索结果无论是在召回率还是在精确度上都不能令人满意的问题,运用语义网理论和智能搜索引擎相关性理论,指出查全率和查准率不高的原因,并提出一种基于语义网的智能搜索引擎来改善查全率和查准率。     

论网络环境中科技编辑对学术成果的价值判断

网络时代学术期刊的社会功能已发生价值位移,从注重科研成果的首次发布转变为注重科研成果的社会认同.科技编辑应该针对不同学科的发展状况,判断该学科科学规范的实际价值,在鉴定学术成果过程中对常规研究和创新研究作出不同的价值评定.科技编辑还应对学术成果的表达方式有所创新.     

Accounting for the knowledge gained during a web search: An empirical study on learning transfer indicators

Searches with learning intent typically require the users to interact with the searching environment and perform knowledge acquisition features such as scan, read, and process the online content to fulfill their information needs. To capture indicators from searching behaviors that could account for the knowledge gained during a Web search, a qualitative study was performed using the Concurrent Think-Aloud protocol to observe the mechanisms of transfer and map knowledge flows during 78 search sessions. Findings indicate evidence of transfer of learning in the form of sixteen online information searching strategy indicators. This research aids the understanding of how knowledge is gained during search sessions and how to identify behaviors that could indicate that learning has occurred, which could be used to represent knowledge gain on Web search engines. In this way, it can aid search engines to become not only better tools of searching, but also tools of learning.     

Use of a search summary table to improve systematic review search methods,results, and efficiency

Background:Systematic reviews are comprehensive, robust, inclusive, transparent, and reproducible when bringing together the evidence to answer a research question. Various guidelines provide recommendations on the expertise required to conduct a systematic review, where and how to search for literature, and what should be reported in the published review. However, the finer details of the search results are not typically reported to allow the search methods or search efficiency to be evaluated.Case Presentation:This case study presents a search summary table, containing the details of which databases were searched, which supplementary search methods were used, and where the included articles were found. It was developed and published alongside a recent systematic review. This simple format can be used in future systematic reviews to improve search results reporting.Conclusions:Publishing a search summary table in all systematic reviews would add to the growing evidence base about information retrieval, which would help in determining which databases to search for which type of review (in terms of either topic or scope), what supplementary search methods are most effective, what type of literature is being included, and where it is found. It would also provide evidence for future searching and search methods research.