Search (3 results, page 1 of 1)

0.0108891195 = product of:
  0.032667357 = sum of:
    0.032667357 = weight(_text_:on in 5103) [ClassicSimilarity], result of:
      0.032667357 = score(doc=5103,freq=12.0), product of:
        0.109763056 = queryWeight, product of:
          2.199415 = idf(docFreq=13325, maxDocs=44218)
          0.04990557 = queryNorm
        0.29761705 = fieldWeight in 5103, product of:
          3.4641016 = tf(freq=12.0), with freq of:
            12.0 = termFreq=12.0
          2.199415 = idf(docFreq=13325, maxDocs=44218)
          0.0390625 = fieldNorm(doc=5103)
  0.33333334 = coord(1/3)

Abstract

Numerous studies have been conducted on the information interaction behavior of search engine users. Few studies have considered information interactions in the domain of academic search. We focus on conversion behavior in this domain. Conversions have been widely studied in the e-commerce domain, e.g., for online shopping and hotel booking, but little is known about conversions in academic search. We start with a description of a unique dataset of a particular type of conversion in academic search, viz. users' downloads of scientific papers. Then we move to an observational analysis of users' download actions. We first characterize user actions and show their statistics in sessions. Then we focus on behavioral and topical aspects of downloads, revealing behavioral correlations across download sessions. We discover unique properties that differ from other conversion settings such as online shopping. Using insights gained from these observations, we consider the task of predicting the next download. In particular, we focus on predicting the time until the next download session, and on predicting the number of downloads. We cast these as time series prediction problems and model them using LSTMs. We develop a specialized model built on user segmentations that achieves significant improvements over the state-of-the art.

0.009409276 = product of:
  0.028227827 = sum of:
    0.028227827 = weight(_text_:on in 5033) [ClassicSimilarity], result of:
      0.028227827 = score(doc=5033,freq=14.0), product of:
        0.109763056 = queryWeight, product of:
          2.199415 = idf(docFreq=13325, maxDocs=44218)
          0.04990557 = queryNorm
        0.25717056 = fieldWeight in 5033, product of:
          3.7416575 = tf(freq=14.0), with freq of:
            14.0 = termFreq=14.0
          2.199415 = idf(docFreq=13325, maxDocs=44218)
          0.03125 = fieldNorm(doc=5033)
  0.33333334 = coord(1/3)

Abstract

Academic search concerns the retrieval and profiling of information objects in the domain of academic research. In this paper we reveal important observations of academic search queries, and provide an algorithmic solution to address a type of failure during search sessions: null queries. We start by providing a general characterization of academic search queries, by analyzing a large-scale transaction log of a leading academic search engine. Unlike previous small-scale analyses of academic search queries, we find important differences with query characteristics known from web search. E.g., in academic search there is a substantially bigger proportion of entity queries, and a heavier tail in query length distribution. We then focus on search failures and, in particular, on null queries that lead to an empty search engine result page, on null sessions that contain such null queries, and on users who are prone to issue null queries. In academic search approximately 1 in 10 queries is a null query, and 25% of the sessions contain a null query. They appear in different types of search sessions, and prevent users from achieving their search goal. To address the high rate of null queries in academic search, we consider the task of providing query suggestions. Specifically we focus on a highly frequent query type: non-boolean informational queries. To this end we need to overcome query sparsity and make effective use of session information. We find that using entities helps to surface more relevant query suggestions in the face of query sparsity. We also find that query suggestions should be conditioned on the type of session in which they are offered to be more effective. After casting the session classification problem as a multi-label classification problem, we generate session-conditional query suggestions based on predicted session type. We find that this session-conditional method leads to significant improvements over a generic query suggestion method. Personalization yields very little further improvements over session-conditional query suggestions.

0.008711295 = product of:
  0.026133886 = sum of:
    0.026133886 = weight(_text_:on in 3527) [ClassicSimilarity], result of:
      0.026133886 = score(doc=3527,freq=12.0), product of:
        0.109763056 = queryWeight, product of:
          2.199415 = idf(docFreq=13325, maxDocs=44218)
          0.04990557 = queryNorm
        0.23809364 = fieldWeight in 3527, product of:
          3.4641016 = tf(freq=12.0), with freq of:
            12.0 = termFreq=12.0
          2.199415 = idf(docFreq=13325, maxDocs=44218)
          0.03125 = fieldNorm(doc=3527)
  0.33333334 = coord(1/3)

Abstract

Personalized search approaches tailor search results to users' current interests, so as to help improve the likelihood of a user finding relevant documents for their query. Previous work on personalized search focuses on using the content of the user's query and of the documents clicked to model the user's preference. In this paper we focus on a different type of signal: We investigate the use of behavioral information for the purpose of search personalization. That is, we consider clicks and dwell time for reranking an initially retrieved list of documents. In particular, we (i) investigate the impact of distributions of users and queries on document reranking; (ii) estimate the relevance of a document for a query at 2 levels, at the query-level and at the word-level, to alleviate the problem of sparseness; and (iii) perform an experimental evaluation both for users seen during the training period and for users not seen during training. For the latter, we explore the use of information from similar users who have been seen during the training period. We use the dwell time on clicked documents to estimate a document's relevance to a query, and perform Bayesian probabilistic matrix factorization to generate a relevance distribution of a document over queries. Our experiments show that: (i) for personalized ranking, behavioral information helps to improve retrieval effectiveness; and (ii) given a query, merging information inferred from behavior of a particular user and from behaviors of other users with a user-dependent adaptive weight outperforms any combination with a fixed weight.

Footnote

A preliminary version of this paper was published in the proceedings of SIGIR '14. In this extension, we (i) extend the behavioral personalization search model introduced there to deal with queries issued by new users for whom long-term search logs are unavailable; (ii) examine the impact of sparseness on the performance of our model by considering both word-level and query-level modeling, as we find that the word-document relevance matrix is less sparse than the query-document relevance matrix; (iii) investigate the effectiveness of our behavior-based reranking model with and without assuming a uniform distribution of users as users may behave differently; (iv) include more related work and provide a detailed discussion of the experimental results.