AI帮你理解科学

AI 生成解读视频

AI抽取解析论文重点内容自动生成视频


pub
生成解读视频

AI 溯源

AI解析本论文相关学术脉络


Master Reading Tree
生成 溯源树

AI 精读

AI抽取本论文的概要总结


微博一下
We argue that it is better to take this approach than the traditional pairwise approach in learning to rank

Learning to rank: from pairwise approach to listwise approach

ICML, pp.129-136, (2007)

引用2032|浏览663
WOS SCOPUS EI
下载 PDF 全文
引用
微博一下

摘要

The paper is concerned with learning to rank, which is to construct a model or a function for ranking objects. Learning to rank is useful for document retrieval, collaborative filtering, and many other applications. Several methods for learning to rank have been proposed, which take object pairs as 'instances' in learning. We refer to the...更多

代码

数据

简介
  • The central issues of many applications are ranking.
  • These include document retrieval, collaborative filtering, expert finding, anti web spam, sentiment analysis, and product rating.
  • The authors address learning to rank and without loss of generality the authors take document retrieval as example.
  • Learning to rank, when applied to document retrieval, is a task as follows.
  • The ranking order represents relative relevance of documents with respect to the query.
  • A number of queries are provided; each query is associated with a perfect ranking list of documents; a ranking function is created using the training data, such that the model can precisely predict the ranking lists in the training data
重点内容
  • The central issues of many applications are ranking
  • We propose employing what we call the listwise approach, in which document lists instead of document pairs are used as instances in learning
  • We propose a probabilistic method to calculate the listwise loss function
  • We have proposed a new approach to learning to rank, referred to as the listwise approach
  • We argue that it is better to take this approach than the traditional pairwise approach in learning to rank
  • We have proposed employing a probabilistic method to solve it
方法
  • The authors employ a new learning method for optimizing the listwise loss function based on top one probability, with Neural Network as model and Gradient Descent as optimization algorithm.
  • The authors refer to the method as ListNet. Again, let them take document retrieval as example.
  • The authors denote the ranking function based on the Neural Network model ω as fω.
  • Given a feature vector x, fω(x) assigns a score to it.
  • The authors rewrite the top one probability in Theorem 6 as exp(s j)
结论
  • The authors have proposed a new approach to learning to rank, referred to as the listwise approach.
  • Instead of using object pairs as instances, the authors use list of objects as instances in learning.
  • The key issue for the listwise approach is to define a listwise loss function.
  • The authors make use of probability models: permutation probability and top one probability to transform ranking scores into probability distributions.
  • The authors can view any metric between probability distributions (e.g., Cross Entropy) as the listwise loss function
表格
  • Table1: Ranking accuracies in terms of MAP
  • Table2: Document-pair number distribution
Download tables as Excel
相关工作
  • 2.1. Learning to Rank

    Learning to rank is a new and popular topic in machine learning. There is one major approach to learning to rank, referred to as the pairwise approach in this paper. For other approaches, see (Shashua & Levin, 2002; Crammer & Singer, 2001; Lebanon & Lafferty, 2002), for example.

    In the pairwise approach, the learning task is formalized as classification of object pairs into two categories (correctly ranked and incorrectly ranked). Herbrich et al (1999) proposed employing the approach and using the SVM techniques to build the classification model. The method is referred to as Ranking SVM. Freund et al (1998) proposed performing the task in the same way but by means of Boosting. Burges et al (2005) also adopted the approach and developed a method called RankNet. They employed Cross Entropy as loss function and Gradient Descent as algorithm to train a Neural Network model.
引用论文
  • Baeza-Yates, R., & Ribeiro-Neto, B. (1999). Modern information retrieval. Addison Wesley.
    Google ScholarFindings
  • Burges, C., Shaked, T., Renshaw, E., Lazier, A., Deeds, M., Hamilton, N., & Hullender, G. (2005). Learning to rank using gradient descent. Proceedings of ICML 2005 (pp. 89–96).
    Google ScholarLocate open access versionFindings
  • Cao, Y. B., Xu, J., Liu, T. Y., Li, H., Huang, Y. L., & Hon, H. W. (2006). Adapting ranking SVM to document retrieval. Proceedings of SIGIR 2006 (pp. 186–193).
    Google ScholarLocate open access versionFindings
  • Cohen, W. W., Schapire, R. E., & Singer, Y. (1998). Learning to order things. Advances in Neural Information Processing Systems. The MIT Press.
    Google ScholarLocate open access versionFindings
  • Crammer, K., & Singer, Y. (2001). Pranking with ranking. Proceedings of NIPS 2001.
    Google ScholarLocate open access versionFindings
  • Craswell, N., Hawking, D., Wilkinson, R., & Wu, M. (2003). Overview of the TREC 2003 web track. Proceedings of TREC 2003 (pp. 78–92).
    Google ScholarLocate open access versionFindings
  • Freund, Y., Iyer, R., Schapire, R. E., & Singer, Y. (1998). An efficient boosting algorithm for combining preferences. Proceedings of ICML 1998 (pp. 170–178).
    Google ScholarLocate open access versionFindings
  • Herbrich, R., Graepel, T., & Obermayer, K. (1999). Support vector learning for ordinal regression. Proceedings of ICANN 1999 (pp. 97–102).
    Google ScholarLocate open access versionFindings
  • Hersh, W. R., Buckley, C., Leone, T. J., & Hickam, D. H. (1994). OHSUMED: An interactive retrieval evaluation and new large test collection for research. Proceedings of SIGIR 1994 (pp. 192–201).
    Google ScholarLocate open access versionFindings
  • Jarvelin, K., & Kekanainen, J. (2000). IR evaluation methods for retrieving highly relevant documents. Proceedings of SIGIR 2000 (pp. 41–48).
    Google ScholarLocate open access versionFindings
  • Joachims, T. (1999). Making large-scale support vector machine learning practical. Advances in kernel methods: support vector learning, 169–184.
    Google ScholarFindings
  • Joachims, T. (2002). Optimizing search engines using clickthrough data. Proceedings of KDD 2002 (pp. 133– 142).
    Google ScholarLocate open access versionFindings
  • Lebanon, G., & Lafferty, J. (2002). Cranking: Combining rankings using conditional probability models on permutations. Proceedings of ICML 2002 (pp. 363–370).
    Google ScholarLocate open access versionFindings
  • Luce, R. D. (1959). Individual choice behavior. Wiley.
    Google ScholarFindings
  • Matveeva, I., Burges, C., Burkard, T., Laucius, A., & Wong, L. (2006). High accuracy retrieval with multiple nested ranker. Proceeings of SIGIR 2006 (pp. 437–444).
    Google ScholarLocate open access versionFindings
  • Nallapati, R. (2004). Discriminative models for information retrieval. Proceedings of SIGIR 2004 (pp. 64–71).
    Google ScholarLocate open access versionFindings
  • Plackett, R. L. (1975). The analysis of permutations. Applied Statistics, 24(2), 193–202.
    Google ScholarLocate open access versionFindings
  • Shashua, A., & Levin, A. (2002). Taxonomy of large margin principle algorithms for ordinal regression problems. Proceedings of NIPS 2002.
    Google ScholarLocate open access versionFindings
  • Yu, H. (2005). SVM selective sampling for ranking with application to data retrieval. Proceedings of KDD 2005 (pp. 354–363).
    Google ScholarLocate open access versionFindings
0
您的评分 :

暂无评分

标签
评论
数据免责声明
页面数据均来自互联网公开来源、合作出版商和通过AI技术自动分析结果,我们不对页面数据的有效性、准确性、正确性、可靠性、完整性和及时性做出任何承诺和保证。若有疑问,可以通过电子邮件方式联系我们:report@aminer.cn