The logs of the use of a search engine provide sufficient data to train a better ranker. However, it is well known that such implicit feedback reflects biases, and in particular a presentation bias that favors higher-ranked results. Unbiased Learning-to-Rank (ULTR) methods attempt to optimize performance by jointly modeling this bias along with the ranker so that the bias can be removed. Such methods have been shown to provide theoretical soundness, and promise superior performance and low deployment costs. However, existing ULTR methods don't recognize that query-document relevance is a confounder -- it affects both the likelihood of a result being clicked because of relevance and the likelihood of the result being ranked high by the base ranker. Moreover, the performance guarantees of existing ULTR methods assume the use of a weak ranker -- one that does a poor job of ranking documents based on relevance to a query. In practice, of course, commercial search engines use highly tuned rankers, and desire to improve upon them using the implicit judgments in search logs. This results in a significant correlation between position and relevance, which leads existing ULTR methods to overestimate click propensities in highly ranked results, reducing ULTR's effectiveness. This paper is the first to demonstrate the problem of propensity overestimation by ULTR algorithms, based on a causal analysis. We develop a new learning objective based on a backdoor adjustment. In addition, we introduce the Logging-Policy-aware Propensity (LPP) model that can jointly learn LPP and a more accurate ranker. We extensively test our approach on two public benchmark tasks and show that our proposal is effective, practical and significantly outperforms the state of the art.

Unbiased Learning to Rank(ULTR), i.e., learning to rank documents with biased user feedback data, is a well-known challenge in information retrieval. Existing methods in unbiased learning to rank typically rely on click modeling or inverse propensity weighting(IPW). Unfortunately, search engines face the issue of a severe long-tail query distribution, which neither click modeling nor IPW handles well. Click modeling usually requires that the same query-document pair appears multiple times for reliable inference, which makes it fall short for tail queries; IPW suffers from high variance since it is highly sensitive to small propensity score values. Therefore, a general debiasing framework that works well under tail queries is sorely needed. To address this problem, we propose a model-based unbiased learning-to-rank framework. Specifically, we develop a general context-aware user simulator to generate pseudo clicks for unobserved ranked lists to train rankers, which addresses the data sparsity problem. In addition, considering the discrepancy between pseudo clicks and actual clicks, we take the observation of a ranked list as the treatment variable and further incorporate inverse propensity weighting with pseudo labels in a doubly robust way. The derived bias and variance indicate that the proposed model-based method is more robust than existing methods. Extensive experiments on benchmark datasets, including simulated datasets and real click logs, demonstrate that the proposed model-based method consistently outperforms state-of-the-art methods in various scenarios. The code is available at https://github.com/rowedenny/MULTR.

A table is composed of data values that are organized in rows and columns providing implicit structural information. A table is usually accompanied by secondary information such as the caption, page title, etc., that form the textual information. Understanding the connection between the textual and structural information is an important, yet neglected aspect in table retrieval, as previous methods treat each source of information independently. In this paper, we propose StruBERT, a structure-aware BERT model that fuses the textual and structural information of a data table to produce context-aware representations for both textual and tabular content of a data table. We introduce the concept of horizontal self-attention, which extends the idea of vertical self-attention introduced in TaBERT and allows us to treat both dimensions of a table equally. StruBERT features are integrated in a new end-to-end neural ranking model to solve three table-related downstream tasks: keyword- and content-based table retrieval, and table similarity. We evaluate our approach using three datasets, and we demonstrate substantial improvements in terms of retrieval and classification metrics over state-of-the-art methods.

Table search aims to retrieve a list of tables given a user's query. Previous methods only consider the textual information of tables and the structural information is rarely used. In this paper, we propose to model the complex relations in the table corpus as one or more graphs and then utilize graph neural networks to learn representations of queries and tables. We show that the text-based table retrieval methods can be further improved by graph-based predictions which fuse multiple field-level information.

We describe the development, characteristics and availability of a test collection for the task of Web table retrieval, which uses a large-scale Web Table Corpora extracted from the Common Crawl. Since a Web table usually has rich context information such as the page title and surrounding paragraphs, we not only provide relevance judgments of query-table pairs, but also the relevance judgments of query-table context pairs with respect to a query, which are ignored by previous test collections. To facilitate future research with this benchmark, we provide details about how the dataset is pre-processed and also baseline results from both traditional and recently proposed table retrieval methods. Our experimental results show that proper usage of context labels can benefit previous table retrieval methods.

Abstract Ranking models are the main components of information retrieval systems. Several approaches to ranking are based on traditional machine learning algorithms using a set of hand-crafted features. Recently, researchers have leveraged deep learning models in information retrieval. These models are trained end-to-end to extract features from the raw data for ranking tasks, so that they overcome the limitations of hand-crafted features. A variety of deep learning models have been proposed, and each model presents a set of neural network components to extract features that are used for ranking. In this paper, we compare the proposed models in the literature along different dimensions in order to understand the major contributions and limitations of each model. In our discussion of the literature, we analyze the promising neural components, and propose future research directions. We also show the analogy between document retrieval and other retrieval tasks where the items to be ranked are structured documents, answers, images and videos.

While scientific collaboration is critical for a scholar, some collaborators can be more significant than others, e.g., lifetime collaborators. It has been shown that lifetime collaborators are more influential on a scholar’s academic performance. However, little research has been done on investigating predicting such special relationships in academic networks. To this end, we propose Scholar2vec, a novel neural network embedding for representing scholar profiles. First, our approach creates scholars’ research interest vector from textual information, such as demographics, research, and influence. After bridging research interests with a collaboration network, vector representations of scholars can be gained with graph learning. Meanwhile, since scholars are occupied with various attributes, we propose to incorporate four types of scholar attributes for learning scholar vectors. Finally, the early-stage similarity sequence based on Scholar2vec is used to predict lifetime collaborators with machine learning methods. Extensive experiments on two real-world datasets show that Scholar2vec outperforms state-of-the-art methods in lifetime collaborator prediction. Our work presents a new way to measure the similarity between two scholars by vector representation, which tackles the knowledge between network embedding and academic relationship mining.

Large collections of datasets are being published on the Web at an increasing rate. This poses a problem to researchers and data journalists who must sift through these large quantities of data to find datasets that meet their needs. Our solution to this problem is cell-centric indexing, a novel approach which considers the individual cell of a dataset to be the fundamental unit of search, indexing the corresponding metadata to each individual cell. This facilitates a new style of user interface that allows users to explore the collection via histograms that show the distributions of various terms organized by how they are used in the dataset.

We address the problem of ad hoc table retrieval via a new neural architecture that incorporates both semantic and relevance matching. Understanding the connection between the structured form of a table and query tokens is an important yet neglected problem in information retrieval. We use a learning- to-rank approach to train a system to capture semantic and relevance signals within interactions between the structured form of candidate tables and query tokens. Convolutional filters that extract contextual features from query/table interactions are combined with a feature vector based on the distributions of term similarity between queries and tables. We propose using row and column summaries to incorporate table content into our new neural model. We evaluate our approach using two datasets, and we demonstrate substantial improvements in terms of retrieval metrics over state-of-the-art methods in table retrieval and document retrieval, and neural architectures from sentence, document, and table type classification adapted to the table retrieval task. Our ablation study supports the importance of both semantic and relevance matching in the table retrieval.

Ad hoc table retrieval is the problem of identifying the most relevant datasets to a user's query. We present an approach to the problem that builds a knowledge graph by combining information about the collection of tables with external sources such as WordNet and pretrained Glove embeddings. We apply multi-relational graph convolutional networks to learn embeddings for the knowledge graph nodes and utilize three different methods to create vectors representing the tables and queries from these embeddings. We create a novel learning-to-rank neural architecture that incorporates the multiple embeddings in order to improve table retrieval results. We evaluate our approach using two large collections of tables from public WikiTables and Web tables data, demonstrating substantial improvements over state-of-the-art methods in table retrieval.

Pretrained contextualized language models such as BERT have achieved impressive results on various natural language processing benchmarks. Benefiting from multiple pretraining tasks and large scale training corpora, pretrained models can capture complex syntactic word relations. In this paper, we use the deep contextualized language model BERT for the task of ad hoc table retrieval. We investigate how to encode table content considering the table structure and input length limit of BERT. We also propose an approach that incorporates features from prior literature on table retrieval and jointly trains them with BERT. In experiments on public datasets, we show that our best approach can outperform the previous state-of-the-art method and BERT baselines with a large margin under different evaluation metrics.

Table retrieval is the task of extracting the most relevant tables to answer a user's query. Table retrieval is an important task because many domains have tables that contain useful information in a structured form. Given a user's query, the goal is to obtain a relevance ranking for query-table pairs, such that higher ranked tables should be more relevant to the query. In this paper, we present a context-aware table retrieval method that is based on a novel embedding for attribute tokens. We find that differentiated types of contexts are useful in building word embeddings. We also find that including a specialized representation of numerical cell values in our model improves table retrieval performance. We use the trained model to predict different contexts of every table. We show that the predicted contexts are useful in ranking tables against a query using a multi-field ranking approach. We evaluate our approach using public WikiTables data, and we demonstrate improvements in terms of NDCG over unsupervised baseline methods in the table retrieval task.

IP geolocation databases map IP addresses to their geographical locations. These databases are used in a variety of online services to serve local content to users. Here we present methods for extracting locations from the reverse DNS hostnames assigned to IP addresses. We summarize a machine learning based approach which, given a hostname, aims to extract and rank potential location candidates, which can then potentially be fused with other geolocation signals. We show that this approach significantly outperforms a state-of-the-art academic baseline, and it is competitive and complementary to commercial baselines. Since extracting locations from more than a billion reverse DNS hostnames at once poses a significant computational challenge, we develop a distributed version of our algorithm. We perform experiments on a cluster of 2,000 machines to demonstrate that our distributed implementation can scale. We show that compared to the single machine version, our distributed approach can achieve a speedup of more than 150×.

No abstract available.

IP geolocation databases map IP addresses to their geographical locations. These databases are important for several applications such as local search engine relevance, credit card fraud protection, geotargetted advertising, and online content delivery. While they are the most popular method of geolocation, they can have low accuracy at the city level. In this paper we evaluate and improve IP geolocation databases using data collected from search engine logs. We generate a large ground-truth dataset using real time global positioning data extracted from search engine logs. We show that incorrect geolocation information can have a negative impact on implicit user metrics. Using the dataset we measure the accuracy of three state-of-the-art commercial IP geolocation databases. We then introduce a technique to improve existing geolocation databases by mining explicit locations from query logs. We show significant accuracy gains in 44 to 49 out of the top 50 countries, depending on the IP geolocation database. Finally, we validate the approach with a large scale A/B experiment that shows improvements in several user metrics.

In this paper, we target at four specific recommendation tasks in the academic environment: the recommendation for author coauthorships, paper citation recommendation for authors, paper citation recommendation for papers, and publishing venue recommendation for author-paper pairs. Different from previous work which tackles each of these tasks separately while neglecting their mutual effect and connection, we propose a joint multi-relational model that can exploit the latent correlation between relations and solve several tasks in a unified way. Moreover, for better ranking purpose, we extend the work maximizing MAP over one single tensor, and make it applicable to maximize MAP over multiple matrices and tensors. Experiments conducted over two real world data sets demonstrate the effectiveness of our model: 1) improved performance can be achieved with joint modeling over multiple relations; 2) our model can outperform three state-of-the art algorithms for several tasks.

In modern commercial search engines, the pay-per-click (PPC) advertising model is widely used in sponsored search. The search engines try to deliver ads which can produce greater click yields (the total number of clicks for the list of ads per impression). Therefore, predicting user clicks plays a critical role in sponsored search. The current ad-delivery strategy is a two-step approach which first predicts individual ad CTR for the given query and then selects the ads with higher predicted CTR. However, this strategy is naturally suboptimal and correlation between ads is often ignored under this strategy. The learning problem is focused on predicting individual performance rather than group performance which is the more important measurement.

Sponsored search is the primary business for today's commercial search engines. Accurate prediction of the Click-Through Rate (CTR) for ads is key to displaying relevant ads to users. In this paper, we systematically study the two kinds of contextual factors influencing the CTR: 1) In micro factors, we focus on the factors for mainline ads, including ad depth, query diversity, ad interaction. 2) In macro factors, we try to understand the correlations of clicks between organic search and sponsored search. Based on this data analysis, we propose novel click models which harvest these new explored factors. To the best of our knowledge, this is the first paper to examine and model the effects of the above contextual factors in sponsored search. Extensive experiments on large-scale real-world datasets show that by incorporating these contextual factors, our novel click models can outperform state-of-the-art methods.

We propose a novel probabilistic topic model that jointly models authors, documents, cited authors, and venues simultaneously in one integrated framework, as compared to previous work which embeds fewer components. This model is designed for three typical applications in academic network analysis: the problems of expert ranking, cited author prediction and venue prediction. Experiments based on two real world data sets demonstrate the model to be effective, and it outperforms several state-of-the-art algorithms in all three applications.

Users of popular services like Twitter and Facebook are often simultaneously overwhelmed with the amount of information delivered via their social connections and miss out on much content that they might have liked to see, even though it was distributed outside of their social circle. Both issues serve as difficulties to the users and drawbacks to the services.

Collaborative tagging systems are now deployed extensively to help users share and organize resources. Tag prediction and recommendation can simplify and streamline the user experience, and by modeling user preferences, predictive accuracy can be significantly improved. However, previous methods typically model user behavior based only on a log of prior tags, neglecting other behaviors and information in social tagging systems, e.g., commenting on items and connecting with other users. On the other hand, little is known about the connection and correlations among these behaviors and contexts in social tagging systems.

As online social media further integrates deeper into our lives, we spend more time consuming social update streams that come from our online connections. Although social update streams provide a tremendous opportunity for us to access information on-the-fly, we often complain about its relevance. Some of us are flooded with a steady stream of information and simply cannot process it in full. Ranking the incoming content becomes the only solution for the overwhelmed users. For some others, in contrast, the incoming information stream is pretty weak, and they have to actively search for relevant information which is quite tedious. For these users, augmenting their incoming content flow with relevant information from outside their first-degree network would be a viable solution. In that case, the problem of relevance becomes even more prominent. In this paper, we start an open discussion on how to build effective systems for ranking social updates from a unique perspective of LinkedIn -- the largest professional network in the world. More specifically, we address this problem as an intersection of learning to rank, collaborative filtering, and clickthrough modeling, while leveraging ideas from information retrieval and recommender systems. We propose a novel probabilistic latent factor model with regressions on explicit features and compare it with a number of non-trivial baselines. In addition to demonstrating superior performance of our model, we shed some light on the nature of social updates on LinkedIn and how users interact with them, which might be applicable to social update streams in general.

Hierarchical classification has been shown to have superior performance than flat classification. It is typically performed on hierarchies created by and for humans rather than for classification performance. As a result, classification based on such hierarchies often yields suboptimal results. In this paper, we propose a novel genetic algorithm-based method on hierarchy adaptation for improved classification. Our approach customizes the typical GA to optimize classification hierarchies. In several text classification tasks, our approach produced hierarchies that significantly improved upon the accuracy of the original hierarchy as well as hierarchies generated by state-of-the-art methods.

With hundreds of millions of participants, social media services have become commonplace. Unlike a traditional social network service, a microblogging network like Twitter is a hybrid network, combining aspects of both social networks and information networks. Understanding the structure of such hybrid networks and predicting new links are important for many tasks such as friend recommendation, community detection, and modeling network growth. We note that the link prediction problem in a hybrid network is different from previously studied networks. Unlike the information networks and traditional online social networks, the structures in a hybrid network are more complicated and informative. We compare most popular and recent methods and principles for link prediction and recommendation. Finally we propose a novel structure-based personalized link prediction model and compare its predictive performance against many fundamental and popular link prediction methods on real-world data from the Twitter microblogging network. Our experiments on both static and dynamic data sets show that our methods noticeably outperform the state-of-the-art.

Text corpora with documents from a range of time epochs are natural and ubiquitous in many fields, such as research papers, newspaper articles and a variety of types of recently emerged social media. People not only would like to know what kind of topics can be found from these data sources but also wish to understand the temporal dynamics of these topics and predict certain properties of terms or documents in the future. Topic models are usually utilized to find latent topics from text collections, and recently have been applied to temporal text corpora. However, most proposed models are general purpose models to which no real tasks are explicitly associated. Therefore, current models may be difficult to apply in real-world applications, such as the problems of tracking trends and predicting popularity of keywords. In this paper, we introduce a real-world task, tracking trends of terms, to which temporal topic models can be applied. Rather than building a general-purpose model, we propose a new type of topic model that incorporates the volume of terms into the temporal dynamics of topics and optimizes estimates of term volumes. In existing models, trends are either latent variables or not considered at all which limits the potential for practical use of trend information. In contrast, we combine state-space models with term volumes with a supervised learning model, enabling us to effectively predict the volume in the future, even without new documents. In addition, it is straightforward to obtain the volume of latent topics as a by-product of our model, demonstrating the superiority of utilizing temporal topic models over traditional time-series tools (e.g., autoregressive models) to tackle this kind of problem. The proposed model can be further extended with arbitrary word-level features which are evolving over time. We present the results of applying the model to two datasets with long time periods and show its effectiveness over non-trivial baselines.

Freshness of results is important in modern web search. Failing to recognize the temporal aspect of a query can negatively affect the user experience, and make the search engine appear stale. While freshness and relevance can be closely related for some topics (e.g., news queries), they are more independent in others (e.g., time insensitive queries). Therefore, optimizing one criterion does not necessarily improve the other, and can even do harm in some cases.

Supervised learning to rank algorithms typically optimize for high relevance and ignore other facets of search quality, such as freshness and diversity. Prior work on multi-objective ranking trained rankers focused on using hybrid labels that combine overall quality of documents, and implicitly incorporate multiple criteria into quantifying ranking risks. However, these hybrid scores are usually generated based on heuristics without considering potential correlations between individual facets (e.g., freshness versus relevance). In this poster, we empirically demonstrate that the correlation between objective facets in multi-criteria ranking optimization may significantly influence the effectiveness of trained rankers with respect to each objective.

Each year many ACM SIG communities will recognize an outstanding researcher through an award in honor of his or her profound impact and numerous research contributions. This work is the first to investigate an automated mechanism to help in selecting future award winners. We approach the problem as a researchers' expertise ranking problem, and propose a temporal probabilistic ranking model which combines content with citation network analysis. Experimental results based on real-world citation data and historical awardees indicate that some kinds of SIG awards are well-modeled by this approach.

Unlike a traditional social network service, a microblogging network like Twitter is a hybrid network, combining aspects of both social networks and information networks. Understanding the structure of such hybrid networks and to predict new links are important for many tasks such as friend recommendation, community detection, and network growth models. In this paper, by analyzing data collected over time, we find that 90% of new links are to people just two hops away and dynamics of friend acquisition are also related to users' account age. Finally, we compare two popular sampling methods which are widely used for network analysis and find that ForestFire does not preserve properties required for the link prediction task.

A taxonomy organizes concepts or topics in a hierarchical structure and can be created manually or via automated systems. A major drawback of taxonomies is that they require users to have the same view of the topics as the taxonomy creator. Users who do not share that mental taxonomy are likely to have difficulty in finding the desired topic. In this paper, we propose a new approach to taxonomy expansion which is able to provide more flexible views. Based on an existing taxonomy, our algorithm finds possible alternative paths and generates an expanded taxonomy with flexibility in user browsing choices. In experiments on the dmoz Open Directory Project, the rebuilt taxonomies provide more alternative paths and shorter paths to information. User studies show that our expanded taxonomies are preferred compared to the original

Social tagging systems have become increasingly popular for sharing and organizing web resources. Tag prediction is a common feature of social tagging systems. Social tagging by nature is an incremental process, meaning that once a user has saved a web page with tags, the tagging system can provide more accurate predictions for the user, based on user's incremental behaviors. However, existing tag prediction methods do not consider this important factor, in which their training and test datasets are either split by a fixed time stamp or randomly sampled from a larger corpus. In our temporal experiments, we perform a time-sensitive sampling on an existing public dataset, resulting in a new scenario which is much closer to "real-world".

The collective contributions of billions of users across the globe each day result in an ever-changing web. In verticals like news and real-time search, recency is an obvious significant factor for ranking. However, traditional link-based web ranking algorithms typically run on a single web snapshot without concern for user activities associated with the dynamics of web pages and links. Therefore, a stale page popular many years ago may still achieve a high authority score due to its accumulated in-links. To remedy this situation, we propose a temporal web link-based ranking scheme, which incorporates features from historical author activities. We quantify web page freshness over time from page and in-link activity, and design a web surfer model that incorporates web freshness, based on a temporal web graph composed of multiple web snapshots at different time points. It includes authority propagation among snapshots, enabling link structures at distinct time points to influence each other when estimating web page authority. Experiments on a real-world archival web corpus show our approach improves upon PageRank in both relevance and freshness of the search results.

Freshness has been increasingly realized by commercial search engines as an important criteria for measuring the quality of search results. However, most information retrieval methods focus on the relevance of page content to given queries without considering the recency issue. In this work, we mine page freshness from web user maintenance activities and incorporate this feature into web search. We first quantify how fresh the web is over time from two distinct perspectives--the page itself and its in-linked pages--and then exploit a temporal correlation between two types of freshness measures to quantify the confidence of page freshness. Results demonstrate page freshness can be better quantified when combining with temporal freshness correlation. Experiments on a real-world archival web corpus show that incorporating the combined page freshness into the searching process can improve ranking performance significantly on both relevance and freshness.

Many web links mislead human surfers and automated crawlers because they point to changed content, out-of-date information, or invalid URLs. It is a particular problem for large, well-known directories such as the dmoz Open Directory Project, which maintains links to representative and authoritative external web pages within their various topics. Therefore, such sites involve many editors to manually revisit and revise links that have become out-of-date. To remedy this situation, we propose the novel web mining task of identifying outdated links on the web. We build a general classification model, primarily using local and global temporal features extracted from historical content, topic, link and time-focused changes over time. We evaluate our system via five-fold cross-validation on more than fifteen thousand ODP external links selected from thirteen top-level categories. Our system can predict the actions of ODP editors more than 75% of the time. Our models and predictions could be useful for various applications that depend on analysis of web links, including ranking and crawling.

Discussion boards and online forums are important platforms for people to share information. Users post questions or problems onto discussion boards and rely on others to provide possible solutions and such question-related content sometimes even dominates the whole discussion board. However, to retrieve this kind of information automatically and effectively is still a non-trivial task. In addition, the existence of other types of information (e.g., announcements, plans, elaborations, etc.) makes it difficult to assume that every thread in a discussion board is about a question. We consider the problems of identifying question-related threads and their potential answers as classification tasks. Experimental results across multiple datasets demonstrate that our method can significantly improve the performance in both question detection and answer finding subtasks. We also do a careful comparison of how different types of features contribute to the final result and show that non-content features play a key role in improving overall performance. Finally, we show that a ranking scheme based on our classification approach can yield much better performance than prior published methods.

The AncestorRank algorithm calculates an authority score by using just one characteristic of the web graph-the number of ancestors per node. For scalability, we estimate the number of ancestors by using a probabilistic counting algorithm. We also consider the case in which ancestors which are closer to the node have more influence than those farther from the node. Thus we further apply a decay factor delta on the contributions from successively earlier ancestors. The resulting authority score is used in combination with a content-based ranking algorithm. Our experiments show that as long as delta is in the range of [0.1, 0.9], AncestorRank can greatly improve BM25 performance, and in our experiments is often better than PageRank.

Web pages, like people, are often known by others in a variety of contexts. When those contexts are sufficiently distinct, a page's importance may be better represented by multiple domains of authority, rather than by one that indiscriminately mixes reputations. In this work we determine domains of authority by examining the contexts in which a page is cited. However, we find that it is not enough to determine separate domains of authority; our model additionally determines the local flow of authority based upon the relative similarity of the source and target authority domains. In this way, we differentiate both incoming and outgoing hyperlinks by topicality and importance rather than treating them indiscriminately. We find that this approach compares favorably to other topical ranking methods on two real-world datasets and produces an approximately 10% improvement in precision and quality of the top ten results over PageRank.

Accurate web page classification often depends crucially on information gained from neighboring pages in the local web graph. Prior work has exploited the class labels of nearby pages to improve performance. In contrast, in this work we utilize a weighted combination of the contents of neighbors to generate a better virtual document for classification. In addition, we break pages into fields, finding that a weighted combination of text from the target and fields of neighboring pages is able to reduce classification error by more than a third. We demonstrate performance on a large dataset of pages from the Open Directory Project and validate the approach using pages from a crawl from the Stanford WebBase. Interestingly, we find no value in anchor text and unexpected value in page titles (and especially titles of parent pages) in the virtual document.

No abstract available.

A web page may be relevant to multiple topics; even when nominally on a single topic, the page may attract attention (and thus links) from multiple communities. Instead of indiscriminately summing the authority provided by all pages, we decompose a web page into separate subnodes with respect to each community pointing to it. By considering the relevance of these communities, we are able to better model the query-specific reputation for each potential result. We apply a total of 125 queries to the TREC .GOV dataset to demonstrate how the use of community relevance can improve ranking performance.

The Web today includes many pages intended to deceive search engines, and attain an unwarranted result ranking. Since the links among web pages are used to calculate authority, ranking systems would benefit from knowing which pages contain content to be trusted and which do not. We propose and compare various trust propagation methods to estimate the trustworthiness of each page. We find that a non-trust-preserving propagation method is able to achieve close to a fifty percent improvement over TrustRank in separating spam from non-spam pages.

This work proposes a novel cautious surfer to incorporate trust into the process of calculating authority for web pages. We evaluate a total of sixty queries over two large, real-world datasets to demonstrate that incorporating trust can improve PageRank's performance.

Web page classification is important to many tasks in information retrieval and web mining. However, applying traditional textual classifiers on web data often produces unsatisfying results. Fortunately, hyperlink information provides important clues to the categorization of a web page. In this paper, an improved method is proposed to enhance web page classification by utilizing the class information from neighboring pages in the link graph. The categories represented by four kinds of neighbors (parents, children, siblings and spouses) are combined to help with the page in question. In experiments to study the effect of these factors on our algorithm, we find that the method proposed is able to boost the classification accuracy of common textual classifiers from around 70% to more than 90% on a large dataset of pages from the Open Directory Project, and outperforms existing algorithms. Unlike prior techniques, our approach utilizes same-host links and can improve classification accuracy even when neighboring pages are unlabeled. Finally, while all neighbor types can contribute, sibling pages are found to be the most important.

Traditional web link-based ranking schemes use a single score to measure a page's authority without concern of the community from which that authority is derived. As a result, a resource that is highly popular for one topic may dominate the results of another topic in which it is less authoritative. To address this problem, we suggest calculating a score vector for each page to distinguish the contribution from different topics, using a random walk model that probabilistically combines page topic distribution and link structure. We show how to incorporate the topical model within both PageRank and HITS without affecting the overall property and still render insight into topic-level transition. Experiments on multiple datasets indicate that our technique outperforms other ranking approaches that incorporate textual analysis.

By supplying different versions of a web page to search engines and to browsers, a content provider attempts to cloak the real content from the view of the search engine. Semantic cloaking refers to differences in meaning between pages which have the effect of deceiving search engine ranking algorithms. In this paper, we propose an automated two-step method to detect semantic cloaking pages based on different copies of the same page downloaded by a web crawler and a web browser. The first step is a filtering step, which generates a candidate list of semantic cloaking pages. In the second step, a classifier is used to detect semantic cloaking pages from the candidates generated by the filtering step. Experiments on manually labeled data sets show that we can generate a classifier with a precision of 93% and a recall of 85%. We apply our approach to links from the dmoz Open Directory Project and estimate that more than 50,000 of these pages employ semantic cloaking.

Web spam is behavior that attempts to deceive search engine ranking algorithms. TrustRank is a recent algorithm that can combat web spam. However, TrustRank is vulnerable in the sense that the seed set used by TrustRank may not be sufficiently representative to cover well the different topics on the Web. Also, for a given seed set, TrustRank has a bias towards larger communities. We propose the use of topical information to partition the seed set and calculate trust scores for each topic separately to address the above issues. A combination of these trust scores for a page is used to determine its ranking. Experimental results on two large datasets show that our Topical TrustRank has a better performance than TrustRank in demoting spam sites or pages. Compared to TrustRank, our best technique can decrease spam from the top ranked sites by as much as 43.1%.

This paper presents a simple yet profound idea. By thinking about the relationships between and within terms and documents, we can generate a richer representation that encompasses aspects of Web link analysis as well as text analysis techniques from information retrieval. This paper shows one path to this unified representation, and demonstrates the use of eigenvector calculations from Web link analysis by stepping through a simple example.

Most web pages are linked to others with related content. This idea, combined with another that says that text in, and possibly around, HTML anchors describe the pages to which they point, is the foundation for a usable World-Wide Web. In this paper, we examine to what extent these ideas hold by empirically testing whether topical locality mirrors spatial locality of pages on the Web. In particular, we find that the likelihood of linked pages having similar textual content to be high; the similarity of sibling pages increases when the links from the parent are close together; titles, descriptions, and anchor text represent at least part of the target page; and that anchor text may be a useful discriminator among unseen child pages. These results show the foundations necessary for the success of many web systems, including search engines, focused crawlers, linkage analyzers, and intelligent web agents.