Learning to rank[1] or machine-learned ranking (MLR) is the application of machine learning, typically supervised, semi-supervised or reinforcement learning, in the construction of ranking models for information retrieval systems.[2]Training data may, for example, consist of lists of items with some partial order specified between items in each list. This order is typically induced by giving a numerical or ordinal score or a binary judgment (e.g. "relevant" or "not relevant") for each item. The goal of constructing the ranking model is to rank new, unseen lists in a similar way to rankings in the training data.
Applications
In information retrieval
A possible architecture of a machine-learned search engine.
A possible architecture of a machine-learned search engine is shown in the accompanying figure.
Training data consists of queries and documents matching them together with the relevance degree of each match. It may be prepared manually by human assessors (or raters, as Google calls them), who check results for some queries and determine relevance of each result. It is not feasible to check the relevance of all documents, and so typically a technique called pooling is used — only the top few documents, retrieved by some existing ranking models are checked. This technique may introduce selection bias. Alternatively, training data may be derived automatically by analyzing clickthrough logs (i.e. search results which got clicks from users),[3]query chains,[4] or such search engines' features as Google's (since-replaced) SearchWiki. Clickthrough logs can be biased by the tendency of users to click on the top search results on the assumption that they are already well-ranked.
Training data is used by a learning algorithm to produce a ranking model which computes the relevance of documents for actual queries.
Typically, users expect a search query to complete in a short time (such as a few hundred milliseconds for web search), which makes it impossible to evaluate a complex ranking model on each document in the corpus, and so a two-phase scheme is used.[5] First, a small number of potentially relevant documents are identified using simpler retrieval models which permit fast query evaluation, such as the vector space model, Boolean model, weighted AND,[6] or BM25. This phase is called top- document retrieval and many heuristics were proposed in the literature to accelerate it, such as using a document's static quality score and tiered indexes.[7] In the second phase, a more accurate but computationally expensive machine-learned model is used to re-rank these documents.
In other areas
Learning to rank algorithms have been applied in areas other than information retrieval:
In computational biology for ranking candidate 3-D structures in protein structure prediction problems.[8]
In recommender systems for identifying a ranked list of related news articles to recommend to a user after he or she has read a current news article.[9]
Feature vectors
For the convenience of MLR algorithms, query-document pairs are usually represented by numerical vectors, which are called feature vectors. Such an approach is sometimes called bag of features and is analogous to the bag of words model and vector space model used in information retrieval for representation of documents.
Components of such vectors are called features, factors or ranking signals. They may be divided into three groups (features from document retrieval are shown as examples):
Query-independent or static features — those features, which depend only on the document, but not on the query. For example, PageRank or document's length. Such features can be precomputed in off-line mode during indexing. They may be used to compute document's static quality score (or static rank), which is often used to speed up search query evaluation.[7][10]
Query-dependent or dynamic features — those features, which depend both on the contents of the document and the query, such as TF-IDF score or other non-machine-learned ranking functions.
Query-level features or query features, which depend only on the query. For example, the number of words in a query.
Some examples of features, which were used in the well-known LETOR dataset:
There are several measures (metrics) which are commonly used to judge how well an algorithm is doing on training data and to compare the performance of different MLR algorithms. Often a learning-to-rank problem is reformulated as an optimization problem with respect to one of these metrics.
DCG and its normalized variant NDCG are usually preferred in academic research when multiple levels of relevance are used.[11] Other metrics such as MAP, MRR and precision, are defined only for binary judgments.
Recently, there have been proposed several new evaluation metrics which claim to model user's satisfaction with search results better than the DCG metric:
Both of these metrics are based on the assumption that the user is more likely to stop looking at search results after examining a more relevant document, than after a less relevant document.
Approaches
This section needs expansion. You can help by adding to it. (December 2009)
Tie-Yan Liu of Microsoft Research Asia has analyzed existing algorithms for learning to rank problems in his book Learning to Rank for Information Retrieval.[1] He categorized them into three groups by their input spaces, output spaces, hypothesis spaces (the core function of the model) and loss functions: the pointwise, pairwise, and listwise approach. In practice, listwise approaches often outperform pairwise approaches and pointwise approaches. This statement was further supported by a large scale experiment on the performance of different learning-to-rank methods on a large collection of benchmark data sets.[14]
In this section, without further notice, denotes an object to be evaluated, for example, a document or an image, denotes a single-value hypothesis, denotes a bi-variate or multi-variate function and denotes the loss function.
Pointwise approach
In this case, it is assumed that each query-document pair in the training data has a numerical or ordinal score. Then the learning-to-rank problem can be approximated by a regression problem — given a single query-document pair, predict its score. Formally speaking, the pointwise approach aims at learning a function predicting the real-value or ordinal score of a document using the loss function .
A number of existing supervised machine learning algorithms can be readily used for this purpose. Ordinal regression and classification algorithms can also be used in pointwise approach when they are used to predict the score of a single query-document pair, and it takes a small, finite number of values.
Pairwise approach
In this case, the learning-to-rank problem is approximated by a classification problem — learning a binary classifier that can tell which document is better in a given pair of documents. The classifier shall take two documents as its input and the goal is to minimize a loss function . The loss function typically reflects the number and magnitude of inversions in the induced ranking.
In many cases, the binary classifier is implemented with a scoring function . As an example, RankNet [15] adapts a probability model and defines as the estimated probability of the document has higher quality than :
These algorithms try to directly optimize the value of one of the above evaluation measures, averaged over all queries in the training data. This is often difficult in practice because most evaluation measures are not continuous functions with respect to ranking model's parameters, and so continuous approximations or bounds on evaluation measures have to be used. For example the SoftRank algorithm.[16] LambdaMART is a pairwise algorithm which has been empirically shown to approximate listwise objective functions.[17]
List of methods
A partial list of published learning-to-rank algorithms is shown below with years of first publication of each method:
Winning entry in the Yahoo Learning to Rank competition in 2010, using an ensemble of LambdaMART models. Based on MART (1999)[24] “LambdaSMART”, for Lambda-submodel-MART, or LambdaMART for the case with no submodel.
Magnitude-preserving variant of RankBoost. The idea is that the more unequal are labels of a pair of documents, the harder should the algorithm try to rank them.
Combined Regression and Ranking. Uses stochastic gradient descent to optimize a linear combination of a pointwise quadratic loss and a pairwise hinge loss from Ranking SVM.
A multi-variate ranking function that encodes multiple items from an initial ranked list (local context) with a recurrent neural network and create result ranking accordingly.
Differentiable surrogates for ranking able to exactly recover the desired metrics and scales favourably to large list sizes, significantly improving internet-scale benchmarks.
Combining Gradient Ascent with Variable Neighbourhood Search for Learning to Rank
Note: as most supervised learning to Rank algorithms can be applied to pointwise, pairwise and listwise case, only those methods which are specifically designed with ranking in mind are shown above.
History
Norbert Fuhr introduced the general idea of MLR in 1992, describing learning approaches in information retrieval as a generalization of parameter estimation;[39] a specific variant of this approach (using polynomial regression) had been published by him three years earlier.[18] Bill Cooper proposed logistic regression for the same purpose in 1992 [19] and used it with his Berkeley research group to train a successful ranking function for TREC. Manning et al.[40] suggest that these early works achieved limited results in their time due to little available training data and poor machine learning techniques.
Several conferences, such as NeurIPS, SIGIR and ICML had workshops devoted to the learning-to-rank problem since mid-2000s (decade).
Practical usage by search engines
Commercial web search engines began using machine learned ranking systems since the 2000s (decade). One of the first search engines to start using it was AltaVista (later its technology was acquired by Overture, and then Yahoo), which launched a gradient boosting-trained ranking function in April 2003.[41][42]
In November 2009 a Russian search engine Yandex announced[44] that it had significantly increased its search quality due to deployment of a new proprietary MatrixNet algorithm, a variant of gradient boosting method which uses oblivious decision trees.[45] Recently they have also sponsored a machine-learned ranking competition "Internet Mathematics 2009"[46] based on their own search engine's production data. Yahoo has announced a similar competition in 2010.[47]
As of 2008, Google's Peter Norvig denied that their search engine exclusively relies on machine-learned ranking.[48]Cuil's CEO, Tom Costello, suggests that they prefer hand-built models because they can outperform machine-learned models when measured against metrics like click-through rate or time on landing page, which is because machine-learned models "learn what people say they like, not what people actually like".[49]
Similar to recognition applications in computer vision, recent neural network based ranking algorithms are also found to be susceptible to covert adversarial attacks, both on the candidates and the queries.[53] With small perturbations imperceptible to human beings, ranking order could be arbitrarily altered. In addition, model-agnostic transferable adversarial examples are found to be possible, which enables black-box adversarial attacks on deep ranking systems without requiring access to their underlying implementations.[53][54]
Conversely, the robustness of such ranking systems can be improved via adversarial defenses such as the Madry defense.[55]
^ abManning C.; Raghavan P.; Schütze H. (2008), Introduction to Information Retrieval, Cambridge University Press. Section 7.1Archived 2009-07-19 at the Wayback Machine
^Richardson, M.; Prakash, A.; Brill, E. (2006). "Beyond PageRank: Machine Learning for Static Ranking"(PDF). Proceedings of the 15th International World Wide Web Conference. pp. 707–715. Archived(PDF) from the original on 2009-08-15. Retrieved 2009-11-18.
^Burges, Chris J. C.; Shaked, Tal; Renshaw, Erin; Lazier, Ari; Deeds, Matt; Hamilton, Nicole; Hullender, Greg (1 August 2005). "Learning to Rank using Gradient Descent". Archived from the original on 26 February 2021. Retrieved 31 March 2021. {{cite journal}}: Cite journal requires |journal= (help)
^Taylor, M.J., Guiver, J., Robertson, S.E., & Minka, T.P. (2008). SoftRank: optimizing non-smooth rank metrics. Web Search and Data Mining.
^ abFuhr, Norbert (1989), "Optimum polynomial retrieval functions based on the probability ranking principle", ACM Transactions on Information Systems, 7 (3): 183–204, doi:10.1145/65943.65944, S2CID16632383
^ abCooper, William S.; Gey, Frederic C.; Dabney, Daniel P. (1992), "Probabilistic retrieval based on staged logistic regression", Proceedings of the 15th annual international ACM SIGIR conference on Research and development in information retrieval - SIGIR '92, pp. 198–210, doi:10.1145/133160.133199, ISBN978-0897915236, S2CID125993
^Ai, Qingyao; Bi, Keping; Jiafeng, Guo; Croft, W. Bruce (2018), "Learning a Deep Listwise Context Model for Ranking Refinement", The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, pp. 135–144, arXiv:1804.05936, doi:10.1145/3209978.3209985, ISBN9781450356572, S2CID4956076
^Fatih Cakir, Kun He, Xide Xia, Brian Kulis, Stan Sclaroff, Deep Metric Learning to RankArchived 2019-05-14 at the Wayback Machine, In Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
^Ai, Qingyao; Wang, Xuanhui; Bruch, Sebastian; Golbandi, Nadav; Bendersky, Michael; Najork, Marc (2019), "Learning Groupwise Multivariate Scoring Functions Using Deep Neural Networks", Proceedings of the 2019 ACM SIGIR International Conference on Theory of Information Retrieval, pp. 85–92, arXiv:1811.04415, doi:10.1145/3341981.3344218, ISBN9781450368810, S2CID199441954
^Pang, Liang; Xu, Jun; Ai, Qingyao; Lan, Yanyan; Cheng, Xueqi; Wen, Jirong (2020), "SetRank", Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 499–508, doi:10.1145/3397271.3401104, ISBN9781450380164, S2CID241534531
^Swezey, Robin; Grover, Aditya; Charron, Bruno; Ermon, Stefano (2021-11-27). "PiRank: Scalable Learning To Rank via Differentiable Sorting". Advances in Neural Information Processing Systems. NeurIPS '21. 34. Virtual Event, Ireland. arXiv:2012.06731.
^Fuhr, Norbert (1992), "Probabilistic Models in Information Retrieval", Computer Journal, 35 (3): 243–255, doi:10.1093/comjnl/35.3.243
^Manning C.; Raghavan P.; Schütze H. (2008), Introduction to Information Retrieval, Cambridge University Press. Sections 7.4Archived 2009-07-21 at the Wayback Machine and 15.5Archived 2010-05-09 at the Wayback Machine