Are Topics Interesting or Not? An LDA-based Topic-graph Probabilistic Model for Web Search Personalization

In this article, we propose a Latent Dirichlet Allocation– (LDA) based topic-graph probabilistic personalization model for Web search. This model represents a user graph in a latent topic graph and simultaneously estimates the probabilities that the user is interested in the topics, as well as the probabilities that the user is not interested in the topics. For a given query issued by the user, the webpages that have higher relevancy to the interested topics are promoted, and the webpages more relevant to the non-interesting topics are penalized. In particular, we simulate a user’s search intent by building two profiles: A positive user profile for the probabilities of the user is interested in the topics and a corresponding negative user profile for the probabilities of being not interested in the the topics. The profiles are estimated based on the user’s search logs. A clicked webpage is assumed to include interesting topics. A skipped (viewed but not clicked) webpage is assumed to cover some non-interesting topics to the user. Such estimations are performed in the latent topic space generated by LDA. Moreover, a new approach is proposed to estimate the correlation between a given query and the user’s search history so as to determine how much personalization should be considered for the query. We compare our proposed models with several strong baselines including state-of-the-art personalization approaches. Experiments conducted on a large-scale real user search log collection illustrate the effectiveness of the proposed models.

Relevance Assessments for Web Search Evaluation: Should We Randomise or Prioritise the Pooled Documents?

In the context of depth- k pooling for constructing web search test collections, we compare two approaches to ordering pooled documents for relevance assessors: The prioritisation strategy (PRI) used widely at NTCIR, and the simple randomisation strategy (RND). In order to address research questions regarding PRI and RND, we have constructed and released the WWW3E8 dataset, which contains eight independent relevance labels for 32,375 topic-document pairs, i.e., a total of 259,000 labels. Four of the eight relevance labels were obtained from PRI-based pools; the other four were obtained from RND-based pools. Using WWW3E8, we compare PRI and RND in terms of inter-assessor agreement, system ranking agreement, and robustness to new systems that did not contribute to the pools. We also utilise an assessor activity log we obtained as a byproduct of WWW3E8 to compare the two strategies in terms of assessment efficiency. Our main findings are: (a) The presentation order has no substantial impact on assessment efficiency; (b) While the presentation order substantially affects which documents are judged (highly) relevant, the difference between the inter-assessor agreement under the PRI condition and that under the RND condition is of no practical significance; (c) Different system rankings under the PRI condition are substantially more similar to one another than those under the RND condition; and (d) PRI-based relevance assessment files (qrels) are substantially and statistically significantly more robust to new systems than RND-based ones. Finding (d) suggests that PRI helps the assessors identify relevant documents that affect the evaluation of many existing systems, including those that did not contribute to the pools. Hence, if researchers need to evaluate their current IR systems using legacy IR test collections, we recommend the use of those constructed using the PRI approach unless they have a good reason to believe that their systems retrieve relevant documents that are vastly different from the pooled documents. While this robustness of PRI may also mean that the PRI-based pools are biased against future systems that retrieve highly novel relevant documents, one should note that there is no evidence that RND is any better in this respect.

Scaling High-Quality Pairwise Link-Based Similarity Retrieval on Billion-Edge Graphs

SimRank is an attractive link-based similarity measure used in fertile fields of Web search and sociometry. However, the existing deterministic method by Kusumoto et al. [ 24 ] for retrieving SimRank does not always produce high-quality similarity results, as it fails to accurately obtain diagonal correction matrix D . Moreover, SimRank has a “connectivity trait” problem: increasing the number of paths between a pair of nodes would decrease its similarity score. The best-known remedy, SimRank++ [ 1 ], cannot completely fix this problem, since its score would still be zero if there are no common in-neighbors between two nodes. In this article, we study fast high-quality link-based similarity search on billion-scale graphs. (1) We first devise a “varied- D ” method to accurately compute SimRank in linear memory. We also aggregate duplicate computations, which reduces the time of [ 24 ] from quadratic to linear in the number of iterations. (2) We propose a novel “cosine-based” SimRank model to circumvent the “connectivity trait” problem. (3) To substantially speed up the partial-pairs “cosine-based” SimRank search on large graphs, we devise an efficient dimensionality reduction algorithm, PSR # , with guaranteed accuracy. (4) We give mathematical insights to the semantic difference between SimRank and its variant, and correct an argument in [ 24 ] that “if D is replaced by a scaled identity matrix (1-Ɣ)I, their top-K rankings will not be affected much”. (5) We propose a novel method that can accurately convert from Li et al. SimRank ~{S} to Jeh and Widom’s SimRank S . (6) We propose GSR # , a generalisation of our “cosine-based” SimRank model, to quantify pairwise similarities across two distinct graphs, unlike SimRank that would assess nodes across two graphs as completely dissimilar. Extensive experiments on various datasets demonstrate the superiority of our proposed approaches in terms of high search quality, computational efficiency, accuracy, and scalability on billion-edge graphs.

Does More Context Help? Effects of Context Window and Application Source on Retrieval Performance

We study the effect of contextual information obtained from a user’s digital trace on Web search performance. Contextual information is modeled using Dirichlet–Hawkes processes (DHP) and used in augmenting Web search queries. The context is captured by monitoring all naturally occurring user behavior using continuous 24/7 recordings of the screen and associating the context with the queries issued by the users. We report a field study in which 13 participants installed a screen recording and digital activity monitoring system on their laptops for 14 days, resulting in data on all Web search queries and the associated context data. A query augmentation (QAug) model was built to expand the original query with semantically related terms. The effects of context window and source were determined by training context models with temporally varying context windows and varying application sources. The context models were then utilized to re-rank the QAug model. We evaluate the context models by using the Web document rankings of the original query as a control condition compared against various experimental conditions: (1) a search context condition in which the context was sourced from search history; (2) a non-search context condition in which the context was sourced from all interactions excluding search history; (3) a comprehensive context condition in which the context was sourced from both search and non-search histories; and (4) an application-specific condition in which the context was sourced from interaction histories captured on a specific application type. Our results indicated that incorporating more contextual information significantly improved Web search rankings as measured by the positions of the documents on which users clicked in the search result pages. The effects and importance of different context windows and application sources, along with different query types are analyzed, and their impact on Web search performance is discussed.

Beyond Relevance Ranking: A General Graph Matching Framework for Utility-Oriented Learning to Rank

Learning to rank from logged user feedback, such as clicks or purchases, is a central component of many real-world information systems. Different from human-annotated relevance labels, the user feedback is always noisy and biased. Many existing learning to rank methods infer the underlying relevance of query–item pairs based on different assumptions of examination, and still optimize a relevance based objective. Such methods rely heavily on the correct estimation of examination, which is often difficult to achieve in practice. In this work, we propose a general framework U-rank+ for learning to rank with logged user feedback from the perspective of graph matching. We systematically analyze the biases in user feedback, including examination bias and selection bias. Then, we take both biases into consideration for unbiased utility estimation that directly based on user feedback, instead of relevance. In order to maximize the estimated utility in an efficient manner, we design two different solvers based on Sinkhorn and LambdaLoss for U-rank+ . The former is based on a standard graph matching algorithm, and the latter is inspired by the traditional method of learning to rank. Both of the algorithms have good theoretical properties to optimize the unbiased utility objective while the latter is proved to be empirically more effective and efficient in practice. Our framework U-rank+ can deal with a general utility function and can be used in a widespread of applications including web search, recommendation, and online advertising. Semi-synthetic experiments on three benchmark learning to rank datasets demonstrate the effectiveness of U-rank+ . Furthermore, our proposed framework has been deployed on two different scenarios of a mainstream App store, where the online A/B testing shows that U-rank+ achieves an average improvement of 19.2% on click-through rate and 20.8% improvement on conversion rate in recommendation scenario, and 5.12% on platform revenue in online advertising scenario over the production baselines.