Searcher struggle is important feedback to Web search engines. Existing Web search struggle detection methods rely on effort-based features to identify the struggling moments. Their underlying assumption is that the more effort a user spends, the more struggling the user may be. However, studies have shown that this simple association might be incorrect. This paper proposes a new feature modulation method for struggle detection and refers to the Reversal Theory in psychology. Reversal Theory points out that instead of having a static personality trait, people constantly switch between opposite psychological states, complicating the relationship between the efforts they spend and the level of frustration they feel. Supported by the theory, our method modulates the effort-based features based on Reversal Theory’s bi-modal arousal model. After modification, the users’ effort level is better aligned with their struggling experience. Evaluations on Pinterest search logs confirm that the proposed method can statistically significantly improve searcher struggle detection methods.

Pinterest as a popular image search platform has been widely adopted by users. Every day, people come to Pinterest searching for fashion- and home decor-related content. In these domains, users exhibit stable personal tastes. In this paper, we propose a novel search algorithm which can infer user tastes from their past engagement history and tailor the search results to fit their preferences. The online and offline experiments show that our method can efficiently improve user experience and increase user engagements.

Nowadays searching for complicated information needs becomes more and more common. These complicated needs usually require the users to reform different queries and conduct multiple retrievals in a search session. There are a lot of technologies are developed to help session searches. Riccho, pseudo relevance feedback, and etc. can help finding relevant documents. xQuAD, RxQuAD, and etc. can help the user to explore. However none of these approaches alone works well in session searches, because they don't treat a search session as a whole. They can't answer questions like when to explore and when to exploit. In this work, we model session searches as Partially Observable Markov Decision Processes (POMDP). We model user's implicit feedbacks, such as query reformulation and user clickthrough data into the POMDP framework. Further we extend the forms of user feedbacks. We implement a new search interface which allows us to capture more explicit feedbacks from users, such as passage level relevance judgments, irrelevant judgments, duplicate judgment, and etc. We propose algorithms to effectively model these feedback signals into the POMDP framework and improve session search performance. Our algorithm is able to automatically balance users' needs of exploration and exploitation.

In this demo paper, we introduce a new search engine that supports Information Retrieval (IR) in a dynamic setting. A dynamic search engine distinguishes itself by handling rich interactions and temporal dependency among the queries in a session or for a task. The proposed search engine is called Dumpling, named after the development team's favorite food. It implements state-of-the-art dynamic search algorithms and provides: (i) a dynamic search toolkit by integrating the Query Change Retrieval Model (QCM) and the Win-win search algorithm; (ii) a user-friendly interface supporting side-by-side comparison of search results given by a state-of-the-art static search algorithm and the proposed dynamic search algorithms; (iii) and APIs for developers to apply the dynamic search algorithms to index and search over custom datasets. Dumpling is developed under the umbrella of a bigger project in the DARPA Memex program to crawl and search the dark web to support law enforcement and national security.

Session search is a complex search task that involves multiple search iterations triggered by query reformulations. We observe a Markov chain in session search: user's judgment of retrieved documents in the previous search iteration affects user's actions in the next iteration. We thus propose to model session search as a dual-agent stochastic game: the user agent and the search engine agent work together to jointly maximize their long term rewards. The framework, which we term "win-win search", is based on Partially Observable Markov Decision Process. We mathematically model dynamics in session search, including decision states, query changes, clicks, and rewards, as a cooperative game between the user and the search engine. The experiments on TREC 2012 and 2013 Session datasets show a statistically significant improvement over the state-of-the-art interactive search and session search algorithms.

Log-based document re-ranking is a special form of session search. The task re-ranks documents from Search Engine Results Page (SERP) according to the search logs, in which both the search activities from other users and personalized query log for a user are available. The purpose of re-ranking is to provide the user with a new and better ordering of the initial retrieved documents. We test the system on the WSCD 2014 dataset, in which the actual content of the queries and documents are not available due to privacy concerns. The challenge is to perform effective re-ranking purely based on user behaviors, such as clicks and query reformulations rather than document content. In this paper, we propose to model log-based document re-ranking as a Partially Observable Markov Decision Process (POMDP). Experiments on the document re-ranking task show that our approach is effective and outperforms the baseline rankings provided by a commercial search engine.

Professional search activities such as patent and legal search are often time sensitive and consist of rich information needs with multiple aspects or subtopics. This paper proposes a 3D water filling model to describe this search process, and derives a new evaluation metric, the Cube Test, to encompass the complex nature of professional search. The new metric is compared against state-of-the-art patent search evaluation metrics as well as Web search evaluation metrics over two distinct patent datasets. The experimental results show that the Cube Test metric effectively captures the characteristics and requirements of professional search.

Search result clustering (SRC) is a post-retrieval process that hierarchically organizes search results. The hierarchical structure offers overview for the search results and displays an "information lay-of-land" that intents to guide the users throughout a search session. However, SRC hierarchies are sensitive to query changes, which are common among queries in the same session. This instability may leave users seemly random overviews throughout the session. We present a new tool called InfoLand that integrates external knowledge from Wikipedia when building SRC hierarchies and increase their stability. Evaluation on TREC 2010-2011 Session Tracks shows that InfoLand produces more stable results organization than a commercial search engine.