Abstract In this work, we address the problem of providing job recommendations in an online session setting, in which we do not have full user histories. We propose a recommendation approach, which uses different autoencoder architectures to encode sessions from the job domain. The inferred latent session representations are then used in a k-nearest neighbor manner to recommend jobs within a session. We evaluate our approach on three datasets, (1) a proprietary dataset we gathered from the Austrian student job portal Studo Jobs, (2) a dataset released by XING after the RecSys 2017 Challenge and (3) anonymized job applications released by CareerBuilder in 2012. Our results show that autoencoders provide relevant job recommendations as well as maintain a high coverage and, at the same time, can outperform state-of-the-art session-based recommendation techniques in terms of system-based and session-based novelty.

Twitter (and similar microblogging services) has become a central nexus for discussion of the topics of the day. Twitter data contains rich content and structured information on users' topics of interest and behavior patterns. Correctly analyzing and modeling Twitter data enables the prediction of the user behavior and preference in a variety of practical applications, such as tweet recommendation and followee recommendation. Although a number of models have been developed on Twitter data in prior work, most of these only model the tweets from users, while neglecting their valuable retweet information in the data. Models would enhance their predictive power by incorporating users' retweet content as well as their retweet behavior. In this paper, we propose two novel Bayesian nonparametric models, URM and UCM, on retweet data. Both of them are able to integrate the analysis of tweet text and users' retweet behavior in the same probabilistic framework. Moreover, they both jointly model users' interest in tweet and retweet. As nonparametric models, URM and UCM can automatically determine the parameters of the models based on input data, avoiding arbitrary parameter settings. Extensive experiments on real-world Twitter data show that both URM and UCM are superior to all the baselines, while UCM further outperforms URM, confirming the appropriateness of our models in retweet modeling.

Over the past few years, major web search engines have introduced knowledge bases to offer popular facts about people, places, and things on the entity pane next to regular search results. In addition to information about the entity searched by the user, the entity pane often provides a ranked list of related entities. To keep users engaged, it is important to develop a recommendation model that tailors the related entities to individual user interests. We propose a probabilistic Three-way Entity Model (TEM) that provides personalized recommendation of related entities using three data sources: knowledge base, search click log, and entity pane log. Specifically, TEM is capable of extracting hidden structures and capturing underlying correlations among users, main entities, and related entities. Moreover, the TEM model can also exploit the click signals derived from the entity pane log. We further provide an inference technique to learn the parameters in TEM, and propose a principled preference learning method specifically designed for ranking related entities. Extensive experiments with two real-world datasets show that TEM with our probabilistic framework significantly outperforms a state of the art baseline, confirming the effectiveness of TEM and our probabilistic framework in related entity recommendation.

Categorical (topic) similarity between a web page and an advertisement (ad) text has long been used for contextual advertising. In this paper, we explore the use of the categorical similarity score, referred to as Category Match Score (CMS), in the context of search advertising. In particular, we explore the effect of CMS on various ad-effectiveness prediction tasks, including user-judgment prediction, ad click-through-rate prediction (CTR), and revenue-per-impression prediction. Our extensive experiments on two editorial datasets and one live traffic dataset demonstrate that CMS is one of the strongest features in the judgment prediction task and that CMS-based filtering is very effective in improving revenue per impression as well as CTR. We believe that our analyses can be extremely effective in helping web service providers serve more relevant and profitable ads to users.

With the rapid growth of Web 2.0, a variety of content sharing services, such as Flickr, YouTube, Blogger, and TripAdvisor etc, have become extremely popular over the last decade. On these websites, users have created and shared with each other various kinds of resources, such as photos, video, and travel blogs. The sheer amount of user-generated content varies greatly in quality, which calls for a principled method to identify a set of authorities, who created high-quality resources, from a massive number of contributors of content. Since most previous studies only infer global authoritativeness of a user, there is no way to differentiate the authoritativeness in different aspects of life (topics).

Social influence analysis on microblog networks, such as Twitter, has been playing a crucial role in online advertising and brand management. While most previous influence analysis schemes rely only on the links between users to find key influencers, they omit the important text content created by the users. As a result, there is no way to differentiate the social influence in different aspects of life (topics). Although a few prior works do support topic-specific influence analysis, they either separate the analysis of content from the analysis of network structure, or assume that content is the only cause of links, which is clearly an inappropriate assumption for microblog networks.

We have been witnessing an increasing number of social tagging systems on the web. Tags help users understand a resource readily and accurately. In a social tagging system, however, there are typically a fairly large number of resources each associated with a long list of tags. When browsing resources, users are reluctant to read these tags one by one. Instead, users prefer a shorter list of tags as a compact description of a resource. Such a tag description facilitates users to understand the resource accurately and effortlessly. This calls for a generator for a tag description, which selects a set of high-quality tags for a given resource. The tag description condenses the original tag list by retaining the most important tags of the long list.

Topic models are used to group words in a text dataset into a set of relevant topics. Unfortunately, when a few words frequently appear in a dataset, the topic groups identified by topic models become noisy because these frequent words repeatedly appear in "irrelevant" topic groups. This noise has not been a serious problem in a text dataset because the frequent words (e.g., the and is) do not have much meaning and have been simply removed before a topic model analysis. However, in a social network dataset we are interested in, they correspond to popular persons (e.g., Barack Obama and Justin Bieber) and cannot be simply removed because most people are interested in them.

In this paper, we discuss how we can extend probabilistic topic models to analyze the relationship graph of popular social-network data, so that we can group or label the edges and nodes in the graph based on their topic similarity. In particular, we first apply the well-known Latent Dirichlet Allocation (LDA) model and its existing variants to the graph-labeling task and argue that the existing models do not handle popular nodes (nodes with many incoming edges) in the graph very well. We then propose possible extensions to this model to deal with popular nodes. Our experiments show that the proposed extensions are very effective in labeling popular nodes, showing significant improvements over the existing methods. Our proposed methods can be used for providing, for instance, more relevant friend recommendations within a social network.

This paper presents our developed general open source for ontology-based information retrieval to answer queries that involve named entities with their ontological features, namely, aliases, classes, and identifiers. We propose a novel approach for semantic search engines that exploit the ontology features of named entities in proximity search and develop an algorithm for computing dynamic distances between named entities and keywords in queries and documents. In particular, it deals with phrase and proximity queries for which the token-based lengths and positions of the queried named entities in a document may vary. The result provides a platform and library for implementing semantic search engines.

Twitter, a micro-blogging platform with an estimated 20 million unique monthly visitors and over 100 million registered users, offers an abundance of rich, structured data at a rate exceeding 600 tweets per second. Recent efforts to leverage this social data to rank users by quality and topical relevance have largely focused on the "follow" relationship. Twitter's data offers additional implicit relationships between users, however, such as "retweets" and "mentions". In this paper we investigate the semantics of the follow and retweet relationships. Specifically, we show that the transitivity of topical relevance is better preserved over retweet links, and that retweeting a user is a significantly stronger indicator of topical interest than following him. We demonstrate these properties by ranking users with two variants of the PageRank algorithm; one based on the follows sub-graph and one based on the implicit retweet sub-graph. We perform a user study to assess the topical relevance of the resulting top-ranked users.

Ambiguous queries constitute a significant fraction of search instances and pose real challenges to web search engines. With current approaches the top results for these queries tend to be homogeneous, making it difficult for users interested in less popular aspects to find relevant documents. While existing research in search diversification offers several solutions for introducing variety into the results, the majority of such work is predicated, implicitly or otherwise, on the assumption that a single relevant document will fulfill a user's information need, making them inadequate for many informational queries. In this paper we present a search-diversification algorithm particularly suitable for informational queries by explicitly modeling that the user may need more than one page to satisfy their need. This modeling enables our algorithm to make a well-informed tradeoff between a user's desire for multiple relevant documents, probabilistic information about an average user's interest in the subtopics of a multifaceted query, and uncertainty in classifying documents into those subtopics. We evaluate the effectiveness of our algorithm against commercial search engine results and other modern ranking strategies, demonstrating notable improvement in multiple document scenarios.

With the proliferation of online distribution methods for videos, content owners require easier and more effective methods for monetization through advertising. Matching advertisements with related content has a significant impact on the effectiveness of the ads, but current methods for selecting relevant advertising keywords for videos are limited by reliance on manually supplied metadata. In this paper we study the feasibility of using text available from video content to obtain high quality keywords suitable for matching advertisements. In particular, we tap into three sources of text for ad keyword generation: production scripts, closed captioning tracks, and speech-to-text transcripts. We address several challenges associated with using such data. To overcome the high error rates prevalent in automatic speech recognition and the lack of an explicit structure to provide hints about which keywords are most relevant, we use statistical and generative methods to identify dominant terms in the source text. To overcome the sparsity of the data and resulting vocabulary mismatches between source text and the advertiser's chosen keywords, these terms are then expanded into a set of related keywords using related term mining methods. Our evaluations present a comprehensive analysis of the relative performance for these methods across a range of videos, including professionally produced films and popular videos from YouTube.

In this paper, we propose the Hybrid-Layer Index (simply, the HL-index) that is designed to answer top-k queries efficiently when the queries are expressed on any arbitrary subset of attributes in the database. Compared to existing approaches, the HL-index significantly reduces the number of tuples accessed during query processing by pruning unnecessary tuples based on two criteria, i.e., it filters out tuples both (1) globally based on the combination of all attribute values of the tuples like in the layer-based approach (simply, layer-level filtering) and (2) based on individual attribute values used for ranking the tuples like in the list-based approach (simply, list-level filtering). Specifically, the HL-index exploits the synergic effect of integrating the layer-level filtering method and the list-level filtering method. Details and extensive experiments are available in the full paper.

The high rate of change and the unprecedented scale of the Web pose enormous challenges to search engines who wish to provide the most up-to-date and highly relevant information to its users. The VLDB 2000 paper "The Evolution of the Web and Implications for an Incremental Crawler" tried to address part of this challenge by collecting and analyzing the Web history data and by describing the architecture and the associated algorithms for an incremental Web crawler that can provide more up-to-date data to users in a timely manner. Experiments and theoretical analysis showed --- surprisingly at the time --- that a policy that allocates more resources to more frequently changing items does not necessarily lead to better performance. In this paper, we discuss what has happened in the 10 years since and talk about the challenges that lie head.

Sensor networks have been widely used to collect data about the environment. When analyzing data from these systems, people tend to ask exploratory questions---they want to find subsets of data, namely signal, reflecting some characteristics of the environment. In this paper, we study the problem of searching for drops in sensor data. Specifically, the search is to find periods in history when a certain amount of drop over a threshold occurs in data within a time span. We propose a framework, SegDiff, for extracting features, compressing them, and transforming the search into standard database queries. Approximate results are returned from the framework with the guarantee that no true events are missed and false positives are within a user specified tolerance. The framework efficiently utilizes space and provides fast response to users' search. Experimental results with real world data demonstrate the efficiency of our framework with respect to feature size and search time.

There is an exploding amount of user-generated content on theWeb due to the emergence of "Web 2.0" services, such as Blogger,MySpace, Flickr, and del.icio.us. The participation of a large number of users in sharing their opinion on the Web has inspired researchers to build an effective "information filter" by aggregating these independent opinions. However, given the diverse groups of users on the Web nowadays, the global aggregation of the information may not be of much interest to different groups of users. In this paper, we explore the possibility of computing personalized aggregation over the opinions expressed on the Web based on a user's indication of trust over the information sources. The hope is that by employing such "personalized" aggregation, we can make the recommendation more likely to be interesting to the users. We address the challenging scalability issues by proposing an efficient method, that utilizes two core techniques: Non-Negative Matrix Factorization and Threshold Algorithm, to compute personalized aggregations when there are potentially millions of users and millions of sources within a system. We show that, through experiments on real-life dataset, our personalized aggregation approach indeed makes a significant difference in the items that are recommended and it reduces the query computational cost significantly, often more than 75%, while the result of personalized aggregation is kept accurate enough.

Personalization of web search results as a technique for improving user satisfaction has received notable attention in the research community over the past decade. Much of this work focuses on modeling and establishing a profile for each user to aid in personalization. Our work takes a more query-centric approach. In this paper, we present a method for efficient, automatic identification of a class of queries we define as localizable from a web search engine query log. We determine a set of relevant features and use conventional machine learning techniques to classify queries. Our experiments find that our technique is able to identify localizable queries with 94% accuracy.

The Web search engines maintain large-scale inverted indexes which are queried thousands of times per second by users eager for information. In order to cope with the vast amounts of query loads, search engines prune their index to keep documents that are likely to be returned as top results, and use this pruned index to compute the first batches of results. While this approach can improve performance by reducing the size of the index, if we compute the top results only from the pruned index we may notice a significant degradation in the result quality: if a document should be in the top results but was not included in the pruned index, it will be placed behind the results computed from the pruned index. Given the fierce competition in the online search market, this phenomenon is clearly undesirable.

Recently, there has been a dramatic increase in the use of XML data to deliver information over the Web. Personal Weblogs, news Web sites, and discussion forums are now publishing RSS feeds for their subscribers to retrieve new postings. As the popularity of personal Weblogs and RSS feeds grows rapidly, RSS aggregation services and blog search engines have appeared, which try to provide a central access point for simpler access and discovery of new content from a large number of diverse RSS sources. In this paper, we study how the RSS aggregation services should monitor the data sources to retrieve new content quickly using minimal resources and to provide its subscribers with fast news alerts. We believe that the change characteristics of RSS sources and the general user access behavior pose distinct requirements that make this task significantly different from the traditional index refresh problem for Web search engines. Our studies on a collection of 10,000 RSS feeds reveal some general characteristics of the RSS feeds and show that, with proper resource allocation and scheduling, the RSS aggregator provides news alerts significantly faster than the best existing approach.

Large amounts of (often valuable) information are stored in web-accessible text databases. “Metasearchers” provide unified interfaces to query multiple such databases at once. For efficiency, metasearchers rely on succinct statistical summaries of the database contents to select the best databases for each query. So far, database selection research has largely assumed that databases are static, so the associated statistical summaries do not evolve over time. However, databases are rarely static and the statistical summaries that describe their contents need to be updated periodically to reflect content changes. In this article, we first report the results of a study showing how the content summaries of 152 real web databases evolved over a period of 52 weeks. Then, we show how to use “survival analysis” techniques in general, and Cox's proportional hazards regression in particular, to model database changes over time and predict when we should update each content summary. Finally, we exploit our change model to devise update schedules that keep the summaries up to date by contacting databases only when needed, and then we evaluate the quality of our schedules experimentally over real web databases.

We have witnessed the rapid growth of the Web-- It has not only "broadened" but also "deepened": While the "surface Web" has expanded from the 1999 estimate of 800 million to the recent 19.2 billion pages reported by Yahoo index, an equally or even more significant amount of information is hidden on the "deep Web," behind query forms, recently estimated at over 1.2 million, of online databases. Accessing the information on the Web thus requires not only search to locate pages of interests, from the surface Web, but also integration to aggregate data from alternative or complementary sources, from the deep Web. Although the opportunities are unprecedented, the challenges are also immense: On the one hand, for the surface Web, while search seems to have evolved into a standard technology, its maturity and pervasiveness have also invited the attack of spam and the demand of personalization. On the other hand, for the deep Web, while the proliferation of structured sources has promised unlimited possibilities for more precise and aggregated access, it has also presented new challenges for realizing large scale and dynamic information integration. These issues are in essence related to data management, in a large scale, and thus present novel problems and interesting opportunities for our research community. This tutorial will discuss the new access scenarios and research problems in Web information access: from search of the surface Web to integration of the deep Web.

One hundred users, one hundred needs. As more and more topics are being discussed on the web and our vocabulary remains relatively stable, it is increasingly difficult to let the search engine know what we want. Coping with ambiguous queries has long been an important part of the research on Information Retrieval, but still remains a challenging task. Personalized search has recently got significant attention in addressing this challenge in the web search community, based on the premise that a user's general preference may help the search engine disambiguate the true intention of a query. However, studies have shown that users are reluctant to provide any explicit input on their personal preference. In this paper, we study how a search engine can learn a user's preference automatically based on her past click history and how it can use the user preference to personalize search results. Our experiments show that users' preferences can be learned accurately even from little click-history data and personalized search based on user preference yields significant improvements over the best existing ranking mechanism in the literature.

We describe the design and performance of WebBase, a tool for Web research. The system includes a highly customizable crawler, a repository for collected Web pages, an indexer for both text and link-related page features, and a high-speed content distribution facility. The distribution module enables researchers world-wide to retrieve pages from WebBase, and stream them across the Internet at high speed. The advantage for the researchers is that they need not all crawl the Web before beginning their research. WebBase has been used by scores of research and teaching organizations world-wide, mostly for investigations into Web topology and linguistic content analysis. After describing the system's architecture, we explain our engineering decisions for each of the WebBase components, and present respective performance measurements.

Large amounts of (often valuable) information are stored in Web-accessible text databases. "Metasearchers" provide unified interfaces to query multiple such databases at once. For efficiency, metasearchers rely on succinct statistical summaries of the database contents to select the best databases for each query. So far, database selection research has largely assumed that databases are static, so the associated statistical summaries do not need to change over time. However, databases are rarely static and the statistical summaries that describe their contents need to be updated periodically to reflect content changes. In this paper, we first report the results of a study showing how the content summaries of 152 real Web databases evolved over a period of 52 weeks. Then, we show how to use "survival analysis" techniques in general, and Cox's proportional hazards regression in particular, to model database changes over time and predict when we should update each content summary. Finally, we exploit our change model to devise update schedules that keep the summaries up to date by contacting databases only when needed, and then we evaluate the quality of our schedules experimentally over real Web databases.

In a number of recent studies [4, 8] researchers have found that because search engines repeatedly return currently popular pages at the top of search results, popular pages tend to get even more popular, while unpopular pages get ignored by an average user. This "rich-get-richer" phenomenon is particularly problematic for new and high-quality pages because they may never get a chance to get users' attention, decreasing the overall quality of search results in the long run. In this paper, we propose a new ranking function, called page quality that can alleviate the problem of popularity-based ranking. We first present a formal framework to study the search engine bias by discussing what is an "ideal" way to measure the intrinsic quality of a page. We then compare how PageRank, the current ranking metric used by major search engines, differs from this ideal quality metric. This framework will help us investigate the search engine bias in more concrete terms and provide clear understanding why PageRank is effective in many cases and exactly when it is problematic. We then propose a practical way to estimate the intrinsic page quality to avoid the inherent bias of PageRank. We derive our proposed quality estimator through a careful analysis of a reasonable web user model, and we present experimental results that show the potential of our proposed estimator. We believe that our quality estimator has the potential to alleviate the rich-get-richer phenomenon and help new and high-quality pages get the attention that they deserve.

There has been recent interests in studying the "goal" behind a user's Web query, so that this goal can be used to improve the quality of a search engine's results. Previous studies have mainly focused on using manual query-log investigation to identify Web query goals. In this paper we study whether and how we can automate this goal-identification process. We first present our results from a human subject study that strongly indicate the feasibility of automatic query-goal identification. We then propose two types of features for the goal-identification task: user-click behavior and anchor-link distribution. Our experimental evaluation shows that by combining these features we can correctly identify the goals for 90% of the queries studied.

An ever-increasing amount of valuable information is stored in Web databases, "hidden" behind search interfaces. To save the user's effort in manually exploring each database, metasearchers automatically select the most relevant databases to a user's query. In this paper, we focus on one of the technical challenges in metasearching, namely database selection. Past research uses a precollected summary of each database to estimate its "relevancy" to the query, and in many cases make incorrect database selection. In this paper, we propose two techniques: probabilistic relevancy modelling and adaptive probing. First, we model the relevancy of each database to a given query as a probabilistic distribution, derived by sampling that database. Using the probabilistic model, the user can explicitly specify a desired level of certainty for database selection. The adaptive probing technique decides which and how many databases to contact in order to satisfy the user's requirement. Our experiments on real hidden-Web databases indicate that our approach significantly improves the accuracy of database selection at the cost of a small number of database probing.

Recent studies show that a majority of Web page accesses are referred by search engines. In this paper we study the widespread use of Web search engines and its impact on the ecology of the Web. In particular, we study how much impact search engines have on the popularity evolution of Web pages. For example, given that search engines return currently popular" pages at the top of search results, are we somehow penalizing newly created pages that are not very well known yet? Are popular pages getting even more popular and new pages completely ignored? We first show that this unfortunate trend indeed exists on the Web through an experimental study based on real Web data. We then analytically estimate how much longer it takes for a new page to attract a large number of Web users when search engines return only popular pages at the top of search results. Our result shows that search engines can have an immensely worrisome impact on the discovery of new Web pages.

We seek to gain improved insight into how Web search engines shouldcope with the evolving Web, in an attempt to provide users with themost up-to-date results possible. For this purpose we collectedweekly snapshots of some 150 Web sites over the course of one year,and measured the evolution of content and link structure. Our measurements focus on aspects of potential interest to search engine designers: the evolution of link structure over time, the rate ofcreation of new pages and new distinct content on the Web, and the rate of change of the content of existing pages under search-centric measures of degree of change.Our findings indicate a rapid turnover rate of Web pages, i.e.,high rates of birth and death, coupled with an even higher rate ofturnover in the hyperlinks that connect them. For pages that persistover time we found that, perhaps surprisingly, the degree of contentshift as measured using TF.IDF cosine distance does not appear to beconsistently correlated with the frequency of contentupdating. Despite this apparent non-correlation, the rate of content shift of a given page is likely to remain consistent over time. That is, pages that change a great deal in one week will likely change by a similarly large degree in the following week. Conversely, pages that experience little change will continue to experience little change. We conclude the paper with a discussion of the potential implications ofour results for the design of effective Web search engines.

In this article, we study how we can maintain local copies of remote data sources "fresh," when the source data is updated autonomously and independently. In particular, we study the problem of Web crawlers that maintain local copies of remote Web pages for Web search engines. In this context, remote data sources (Websites) do not notify the copies (Web crawlers) of new changes, so we need to periodically poll the sources to maintain the copies up-to-date. Since polling the sources takes significant time and resources, it is very difficult to keep the copies completely up-to-date.This article proposes various refresh policies and studies their effectiveness. We first formalize the notion of "freshness" of copied data by defining two freshness metrics, and we propose a Poisson process as the change model of data sources. Based on this framework, we examine the effectiveness of the proposed refresh policies analytically and experimentally. We show that a Poisson process is a good model to describe the changes of Web pages and we also show that our proposed refresh policies improve the "freshness" of data very significantly. In certain cases, we got orders of magnitude improvement from existing policies.

Many online data sources are updated autonomously and independently. In this article, we make the case for estimating the change frequency of data to improve Web crawlers, Web caches and to help data mining. We first identify various scenarios, where different applications have different requirements on the accuracy of the estimated frequency. Then we develop several "frequency estimators" for the identified scenarios, showing analytically and experimentally how precise they are. In many cases, our proposed estimators predict change frequencies much more accurately and improve the effectiveness of applications. For example, a Web crawler could achieve 35% improvement in "freshness" simply by adopting our proposed estimator.

In this paper; we describe the design, architecture, and lessons learned from the implementation of a fast regular-expression indexing engine FREE. FREE uses a prebuilt index to identify the text data units which may contain a matching string and only examines these further. In this way, FREE shows orders of magnitude performance improvement in certain cases over standard regular expression matching systems, such as lex, awk and grep.

In this paper we study how we can design an effective parallel crawler. As the size of the Web grows, it becomes imperative to parallelize a crawling process, in order to finish downloading pages in a reasonable amount of time. We first propose multiple architectures for a parallel crawler and identify fundamental issues related to parallel crawling. Based on this understanding, we then propose metrics to evaluate a parallel crawler, and compare the proposed architectures using 40 million pages collected from the Web. Our results clarify the relative merits of each architecture and provide a good guideline on when to adopt which architecture.

We offer an overview of current Web search engine design. After introducing a generic search engine architecture, we examine each engine component in turn. We cover crawling, local Web page storage, indexing, and the use of link analysis for boosting search performance. The most common design and implementation techniques for each of these components are presented. For this presentation we draw from the literature and from our own experimental search engine testbed. Emphasis is on introducing the fundamental concepts and the results of several performance analyses we conducted to compare different designs.

In this paper we study how to refresh a local copy of an autonomous data source to maintain the copy up-to-date. As the size of the data grows, it becomes more difficult to maintain the copy \ fresh, “making it crucial to synchronize the copy effectively. We define two freshness metrics, change models of the underlying data, and synchronization policies. We analytically study how effective the various policies are. We also experimentally verify our analysis, based on data collected from 270 web sites for more than 4 months, and we show that our new policy improves the \ freshness” very significantly compared to current policies in use.

In this paper we study in what order a crawler should visit the URLs it has seen, in order to obtain more “important” pages first. Obtaining important pages rapidly can be very useful when a crawler cannot visit the entire Web in a reasonable amount of time. We define several importance metrics, ordering schemes, and performance evaluation measures for this problem. We also experimentally evaluate the ordering schemes on the Stanford University Web. Our results show that a crawler with a good ordering scheme can obtain important pages significantly faster than one without.