Optimizing search engines using clickthrough data

By Thorsten Joachims. 2002

4,680 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Information Systems, Information Retrieval & Search

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Information Systems

In Information Retrieval & Search

Citations in the corpus

Word-Entity Duet Representations for Document Ranking (2017, methodology): “Baselines included standard word-based baselines: Indri’s language model ( Lm ), sequential dependency model ( SDM ), and two state-of-the-art learning to rank methods: RankSVM [10] and co-ordinate ascent ( Coor-Ascent ) [15].”
Learning to summarize from human feedback (2020, methodology): “Our reward modeling approach was developed in prior work on learning to rank [40], which has been applied to ranking search results using either explicit feedback [2, 18] or implicit feedback in the form of click-through data [29, 30].”
Pretrained Transformers for Text Ranking: BERT and Beyond (2020, methodology): “important role in enabling this growth was a series of techniques for interpreting noisy user click data and converting them into training examples that could be fed into machine-learning algorithms [Joachims, 2002, Radlinski and Joachims, 2005]. As we lack the space for a detailed treatment of lear”
Word Representations via Gaussian Embedding (2014, methodology): “We chose a max-margin ranking objective, similar to that used in Rank-SVM (Joachims, 2002) or Wsabie (Weston et al., 2011), which pushes scores of positive pairs above negatives by a margin: In this terminology, the contribution of our work is a pair of energy functions for training Gaussian…”
Learning to Branch in Mixed Integer Programming (2016, background): “While minimizing the number of violated constraints in (5) is NP-hard, Joachims (2002) proposed a Support Vector Machine (SVM) approach that optimizes an upper bound on that number.”
Learning in a Large Function Space: Privacy-Preserving Mechanisms for SVM Learning (2009, background): “, 2008), recommendation systems (Xu and Araki, 2006), web search (Joachims, 2002), cancer diagnosis (Ramaswamy et al.”
Zero-Shot Learning — The Good, the Bad and the Ugly (2017, methodology): “SJE [4] gives full weight to the top of the ranked list and is inspired from the structured SVM [36]: ESZSL [32] adds the following regularization term to the unregularized risk minimization formulation: where γ, λ, β are parameters of this regularizer.”
Deep Convolutional Ranking for Multilabel Image Annotation (2013, methodology): “The second loss was a simple modiﬁcation of a pairwise-ranking loss [19], which takes multiple labels into account.”
Modeling Local Coherence: An Entity-Based Approach (2005, methodology): “The problem is typically treated as a Support Vector Machine constraint optimization problem, and can be solved using the search technique described in Joachims (2002a).”
Adversarial Ranking for Language Generation (2017, background): “Given a reference, the desired information (such as click-through logs [15]) is incorporated into the ranking function which aims to encourage the relevant documents to be returned as early as possible.”
Sampling Matters in Deep Embedding Learning (2017, background): “The basic idea can be traced back to the insight that in ordinal regression only the relative order of scores matters [17].”
Neural Ranking Models with Weak Supervision (2017, methodology): “itialization. Aredeepneuralnetworksagoodchoiceforlearningtorank with weak supervision? To see if there is a real bene•t from using a non-linear neural network in di‡erent se−ings, we examined RankSVM [20] as a strong-performing pair-wise learning to rank method with linear kernel that is fed with di”
Semi-supervised Question Retrieval with Gated Convolutions (2015, methodology): “• SVM, we trained a re-ranker using SVM-Light (Joachims, 2002) with a linear kernel incorporating several similarity measures from the DKPro similarity package (Bär et al.”
Detecting adversarial advertisements in the wild (2011, methodology): “Linear SVM’s have been found to be highly effective at high dimensional classification tasks similar to those encountered here, such as text classification [17].”
Substructure counting graph kernels for machine learning from RDF data (2015, methodology): “(18) The advantage of the feature vector representation is that it leads to efficient computations of the kernels [33] and for larger datasets the feature vectors can be used directly as input to linear learning methods, such as linear Support Vector Machines [34].”
Deep Relevance Ranking Using Enhanced Document-Query Interactions (2018, neutral): “Kleinberg, 1999) and/or user feedback (Joachims, 2002).”