Personalization is well studied in search and recommendation, but personalized question answering remains underexplored due to challenges in inferring preferences from long, noisy, implicit contexts and generating responses that are both accurate and aligned with user expectations. To address this, we propose Pathways of Thoughts (PoT), an inference-stage method that applies to any large language model (LLM) without task-specific fine-tuning. PoT models the thinking as an iterative decision process, where the model dynamically selects among cognitive operations such as reasoning, revision, personalization, and clarification. This enables exploration of multiple reasoning trajectories, producing diverse candidate responses that capture different perspectives. PoT then aggregates and reweights these candidates according to inferred user preferences, yielding a final personalized response that benefits from the complementary strengths of diverse reasoning paths. Experiments on the LaMP-QA benchmark show that PoT consistently outperforms competitive baselines, achieving up to a 10.8% relative improvement. Human evaluation further validates these improvements, with annotators preferring PoT in 66% of cases compared to the best-performing baseline and reporting ties in 15% of cases.

Facilitated by large language models (LLMs), personalized text generation has become a rapidly growing research direction. Most existing studies focus on designing specialized models for a particular domain, or they require fine-tuning the LLMs to generate personalized text. We consider a typical scenario in which the large language model, which generates personalized output, is frozen and can only be accessed through APIs. Under this constraint, all one can do is to improve the input text (i.e., text prompts) sent to the LLM, a procedure that is usually done manually. In this paper, we propose a novel method to automatically revise prompts for personalized text generation. The proposed method takes the initial prompts generated by a state-of-the-art, multistage framework for personalized generation and rewrites a few critical components that summarize and synthesize the personal context. The prompt rewriter employs a training paradigm that chains together supervised learning (SL) and reinforcement learning (RL), where SL reduces the search space of RL and RL facilitates end-to-end training of the rewriter. Using datasets from three representative domains, we demonstrate that the rewritten prompts outperform both the original prompts and the prompts optimized via supervised learning or reinforcement learning alone. In-depth analysis of the rewritten prompts shows that they are not only human readable, but also able to guide manual revision of prompts when there is limited resource to employ reinforcement learning to train the prompt rewriter, or when it is costly to deploy an automatic prompt rewriter for inference.

A recent line of work in first-stage Neural Information Retrieval has focused on learning sparse lexical representations instead of dense embeddings. One such work is SPLADE, which has been shown to lead to state-of-the-art results in both the in-domain and zero-shot settings, can leverage inverted indices for efficient retrieval, and offers enhanced interpretability. However, existing SPLADE models are fundamentally limited to learning a sparse representation based on the native BERT WordPiece vocabulary.

Bag-of-words based lexical retrieval systems are still the most commonly used methods for real-world search applications. Recently deep learning methods have shown promising results to improve this retrieval performance but are expensive to run in an online fashion, non-trivial to integrate into existing production systems, and might not generalize well in out-of-domain retrieval scenarios. Instead, we build on top of lexical retrievers by proposing a Term Weighting BERT (TW-BERT) model. TW-BERT learns to predict the weight for individual n-gram (e.g., uni-grams and bi-grams) query input terms. These inferred weights and terms can be used directly by a retrieval system to perform a query search. To optimize these term weights, TW-BERT incorporates the scoring function used by the search engine, such as BM25, to score query-document pairs. Given sample query-document pairs we can compute a ranking loss over these matching scores, optimizing the learned query term weights in an end-to-end fashion. Aligning TW-BERT with search engine scorers minimizes the changes needed to integrate it into existing production applications, whereas existing deep learning based search methods would require further infrastructure optimization and hardware requirements. The learned weights can be easily utilized by standard lexical retrievers and by other retrieval techniques such as query expansion. We show that TW-BERT improves retrieval performance over strong term weighting baselines within MSMARCO and in out-of-domain retrieval on TREC datasets.

In dense retrieval, prior work has largely improved retrieval effectiveness using multi-vector dense representations, exemplified by ColBERT. In sparse retrieval, more recent work, such as SPLADE, demonstrated that one can also learn sparse lexical representations to achieve comparable effectiveness while enjoying better interpretability. In this work, we combine the strengths of both the sparse and dense representations for first-stage retrieval. Specifically, we propose SparseEmbed - a novel retrieval model that learns sparse lexical representations with contextual embeddings. Compared with SPLADE, our model leverages the contextual embeddings to improve model expressiveness. Compared with ColBERT, our sparse representations are trained end-to-end to optimize both efficiency and effectiveness.

Prior work in Dense Retrieval usually encodes queries and documents using single-vector representations (also called embeddings) and performs retrieval in the embedding space using approximate nearest neighbor search. This paradigm enables efficient semantic retrieval. However, the single-vector representations can be ineffective at capturing different aspects of the queries and documents in relevance matching, especially for some vertical domains. For example, in e-commerce search, these aspects could be category, brand and color. Given a query ''white nike socks", a Dense Retrieval model may mistakenly retrieve some ''white adidas socks" while missing out the intended brand. We propose to explicitly represent multiple aspects using one embedding per aspect. We introduce an aspect prediction task to teach the model to capture aspect information with particular aspect embeddings. We design a lightweight network to fuse the aspect embeddings for representing queries and documents. Our evaluation using an e-commerce dataset shows impressive improvements over strong Dense Retrieval baselines. We also discover that the proposed aspect embeddings can enhance the interpretability of Dense Retrieval models as a byproduct.

Privacy preservation remains a key challenge in data mining and Natural Language Understanding (NLU). Previous research shows that the input text or even text embeddings can leak private information. This concern motivates our research on effective privacy preservation approaches for pretrained Language Models (LMs). We investigate the privacy and utility implications of applying dχ-privacy, a variant of Local Differential Privacy, to BERT fine-tuning in NLU applications. More importantly, we further propose privacy-adaptive LM pretraining methods and show that our approach can boost the utility of BERT dramatically while retaining the same level of privacy protection. We also quantify the level of privacy preservation and provide guidance on privacy configuration. Our experiments and findings lay the groundwork for future explorations of privacy-preserving NLU with pretrained LMs.

Cloud-based file storage platforms such as Google Drive are widely used as a means for storing, editing and sharing personal and organizational documents. In this paper, we improve search ranking quality for cloud storage platforms by utilizing user activity logs. Different from search logs, activity logs capture general document usage activity beyond search, such as opening, editing and sharing documents. We propose to automatically learn text embeddings that are effective for search ranking from activity logs. We develop a novel co-access signal, i.e., whether two documents were accessed by a user around the same time, to train deep semantic matching models that are useful for improving the search ranking quality. We confirm that activity-trained semantic matching models can improve ranking by conducting extensive offline experimentation using Google Drive search and activity logs. To the best of our knowledge, this is the first work to examine the benefits of leveraging document usage activity at large scale for cloud storage search; as such it can shed light on using such activity in scenarios where direct collection of search-specific interactions (e.g., query and click logs) may be expensive or infeasible.

Google Drive is widely used for managing personal and work-related documents in the cloud. To help users organize their documents in Google Drive, we develop a new feature to allow users to create a set of working files for ongoing easy access, called workspace. A workspace is a cluster of documents, but unlike a typical document cluster, it contains documents that are not only topically coherent, but are also useful in the ongoing user tasks.

Faceted search has been used successfully for many vertical applications such as e-commerce and digital libraries. However, it remains challenging to extend faceted search to the open-domain web due to the large and heterogeneous nature of the web. Recent work proposed an alternative solution that extracts facets for queries from their web search results, but neglected the precision-oriented perspective of the task -- users are likely to care more about precision of presented facets than recall.

While many studies have been conducted on query understanding, there is limited understanding on why users start searches and how to predict search intent. In this paper, we propose to study this important but less explored problem. Our key intuition is that searches are triggered by different pre-search contexts, but the triggering relations are often hidden. For example, a user may search "bitcoin" because of a news article or an email the user just read, but the system does not know which of the pre-search contexts (the news article or the email) is the triggering source. Following this intuition, we conduct an in-depth analysis of pre-search context on a large-scale user log, which not only verifies the hidden triggering relations in the real world but also identifies a set of important characteristics of pre-search context and their triggered queries. Since the hidden triggering relations make it challenging to directly use pre-search context for intent prediction, we develop a mixture generative model to learn without any supervision how queries are triggered by different types of pre-search context. Further, we discuss how to apply our model to improve query prediction and query auto-completion. Our experiments on a large-scale of real-world data show that our model could accurately predict user search intent with pre-search context and improve upon the state-of-the-art methods significantly.

Query auto-completion (QAC) facilitates user query composition by suggesting queries given query prefix inputs. In 2014, global users of Yahoo! Search saved more than 50% keystrokes when submitting English queries by selecting suggestions of QAC. Users' preference of queries can be inferred during user-QAC interactions, such as dwelling on suggestion lists for a long time without selecting query suggestions ranked at the top. However, the wealth of such implicit negative feedback has not been exploited for designing QAC models. Most existing QAC models rank suggested queries for given prefixes based on certain relevance scores.

Faceted search helps users by offering drill-down options as a complement to the keyword input box, and it has been used successfully for many vertical applications, including e-commerce and digital libraries. However, this idea is not well explored for general web search, even though it holds great potential for assisting multi-faceted queries and exploratory search. In this paper, we explore this potential by extending faceted search into the open-domain web setting, which we call Faceted Web Search. To tackle the heterogeneous nature of the web, we propose to use query-dependent automatic facet generation, which generates facets for a query instead of the entire corpus. To incorporate user feedback on these query facets into document ranking, we investigate both Boolean filtering and soft ranking models. We evaluate Faceted Web Search systems by their utility in assisting users to clarify search intent and find subtopic information. We describe how to build reusable test collections for such tasks, and propose an evaluation method that considers both gain and cost for users. Our experiments testify to the potential of Faceted Web Search, and show Boolean filtering feedback models, which are widely used in conventional faceted search, are less effective than soft ranking models.

User behavior information has proved valuable for inferring document relevance, but its role in deducing relevance at the passage/section level is not well explored. In this paper, we study how user behavior information implies section relevance, and use this information to improve section ranking. More specifically, we focus on four types of user search behavior that occur while browsing a document -- dwell time, highlighting, copying and clicks at the section level. Experimental results based on a commercial query log show that user behavior information can significantly improve section ranking. While section-level click information is a very powerful signal of relevance, it depends on an interface supporting section-level links. We find comparable levels of gain using other behavior information that does not depend upon such an interface.

Web search queries are often ambiguous or multi-faceted, which makes a simple ranked list of results inadequate. To assist information finding for such faceted queries, we explore a technique that explicitly represents interesting facets of a query using groups of semantically related terms extracted from search results. As an example, for the query ``baggage allowance'', these groups might be different airlines, different flight types (domestic, international), or different travel classes (first, business, economy). We name these groups query facets and the terms in these groups facet terms. We develop a supervised approach based on a graphical model to recognize query facets from the noisy candidates found. The graphical model learns how likely a candidate term is to be a facet term as well as how likely two terms are to be grouped together in a query facet, and captures the dependencies between the two factors. We propose two algorithms for approximate inference on the graphical model since exact inference is intractable. Our evaluation combines recall and precision of the facet terms with the grouping quality. Experimental results on a sample of web queries show that the supervised method significantly outperforms existing approaches, which are mostly unsupervised, suggesting that query facet extraction can be effectively learned.

Combating Web spam has become one of the top challenges for Web search engines. State-of-the-art spam-detection techniques are usually designed for specific, known types of Web spam and are incapable of dealing with newly appearing spam types efficiently. With user-behavior analyses from Web access logs, a spam page-detection algorithm is proposed based on a learning scheme. The main contributions are the following. (1) User-visiting patterns of spam pages are studied, and a number of user-behavior features are proposed for separating Web spam pages from ordinary pages. (2) A novel spam-detection framework is proposed that can detect various kinds of Web spam, including newly appearing ones, with the help of the user-behavior analysis. Experiments on large-scale practical Web access log data show the effectiveness of the proposed features and the detection framework.

We consider the problem of detecting epidemic tendency by mining search logs. We propose an algorithm based on click-through information to select epidemic related queries/terms. We adopt linear regression to model epidemic occurrences and frequencies of epidemic related terms (ERTs) in search logs. The results show our algorithm is effective in finding ERTs which obtain a high correlation value with epidemic occurrences. We also find the proposed method performs better when combining different ERTs than using single ERT.