Personalized retrieval seeks to retrieve items relevant to a user event (e.g. a page visit or a query) that are adapted to the user's personal preferences. For example, two users who happen to perform the same event such as visiting the same product page or asking the same query should receive potentially distinct recommendations adapted to their individual tastes. Personalization is seldom attempted over catalogs of millions of items since the cost of existing personalization routines scale linearly in the number of candidate items. For example, performing two-sided personalized retrieval (with both event and item embeddings personalized to the user) incurs prohibitive storage and compute costs. Instead, it is common to use non-personalized retrieval to obtain a small shortlist of items over which personalized re-ranking can be done quickly. Despite being scalable, this strategy risks losing items uniquely relevant to a user that fail to get shortlisted during non-personalized retrieval. This paper bridges this gap by developing the XPERT algorithm that identifies a form of two-sided personalization that can be scalably implemented over millions of items and hundreds of millions of users. Key to overcoming the computational challenges of personalized retrieval is a novel concept of morph operators that can be used with arbitrary encoder architectures, completely avoids the steep memory overheads of two-sided personalization, provides millisecond-time inference and offers multi-intent retrieval. On multiple public and proprietary datasets, XPERT offered upto 5% superior recall and AUC than state-of-the-art techniques. Code for XPERT is available at https://github.com/personalizedretrieval/xpert.

Over the past decade, Determinantal Point Processes (DPPs) have proven to be a mathematically elegant framework for modeling diversity. Given a set of items N, DPPs define a probability distribution over subsets of N, with sets of larger diversity having greater probability. Recently, DPPs have achieved success in the domain of recommendation systems, as a method to enforce diversity of recommendations in addition to relevance. In large-scale recommendation applications however, the input typically comes in the form of a stream too large to fit into main memory. However, the natural greedy algorithm for DPP-based recommendations is memory intensive, and cannot be used in a streaming setting.

Recent studies have shown that both accuracy and explainability are important for recommendation. In this paper, we introduce explainable conversational recommendation, which enables incremental improvement of both recommendation accuracy and explanation quality through multi-turn user-model conversation. We show how the problem can be formulated, and design an incremental multi-task learning framework that enables tight collaboration between recommendation prediction, explanation generation, and user feedback integration. We also propose a multi-view feedback integration method to enable effective incremental model update. Empirical results demonstrate that our model not only consistently improves the recommendation accuracy but also generates explanations that fit user interests reflected in the feedbacks.

In Contextual advertising, textual ads relevant to the content in a webpage are embedded in the page. Content keywords are extracted offline by crawling webpages and then stored in an index for fast serving. Given a page, ad selection involves index lookup, computing similarity between the keywords of the page and those of candidate ads and returning the top-k scoring ads. In this approach, ad relevance can suffer in two scenarios. First, since page-ad similarity is computed using keywords extracted only from that particular page, a few non pertinent keywords can skew ad selection. Second, requesting page may not be present in the index but we still need to serve relevant ads.

With evolving Web, short length parallel corpora is becoming very common and some of these include user queries, web snippets etc. This paper concerns situations where short length parallel corpora has to be analyzed in order to find meaningful unit-alignment. This is similar to dealing with parallel corpora where a sentence level alignment of translations is required, but differs in that the alignment is to be inferred at unit (word or phrase) level. A Conditional Random Field (CRF) based approach is proposed to discover this unit alignment. Given pairs of semantically or syntactically similar entities, the problem is formulated as that of mutual segmentation and sequence alignment problem. The mutual segmentation refers to the process of segmenting the first entity based on units (or labels) in the second entity and vice-versa. The process of optimizing this mutual segmentation also results in optimal unit alignment. Since our training data is not segmented and unit-aligned, we modify the CRF objective function to accommodate unsupervised data and iterative learning. We have applied this framework to Web Search domain and specifically for query reformulation task. Finally, our experiments suggest that the proposed approach indeed results in meaningful alternatives of the original query.

In Contextual advertising, textual ads relevant to the content in a webpage are embedded in the page. Content keywords are extracted offline by crawling webpages and then stored in an index for fast serving. Given a page, ad selection involves index lookup, computing similarity between the keywords of the page and those of candidate ads and returning the top-k scoring ads. In this approach, there is a tradeoff between relevance and index size where better relevance can be achieved if there are no limits on the index size. However, the assumption of unlimited index size is not practical due to the large number of pages on the Web and stringent requirements on the serving latency. Secondly, page visits on the web follows power-law distribution where a significant proportion of the pages are visited infrequently, also called the tail pages. Indexing tail pages is not efficient given that these pages are accessed very infrequently.

This paper introduces kernels on attributed pointsets, which are sets of vectors embedded in an euclidean space. The embedding gives the notion of neighborhood, which is used to define positive semidefinite kernels on pointsets. Two novel kernels on neighborhoods are proposed, one evaluating the attribute similarity and the other evaluating shape similarity. Shape similarity function is motivated from spectral graph matching techniques. The kernels are tested on three real life applications: face recognition, photo album tagging, and shot annotation in video sequences, with encouraging results.