Báo cáo khoa học: Learning to Rank

In this tutorial I will introduce ‘learning to rank’, a machine learning technology on constructing a model for ranking objects using training data. I will first explain the problem formulation of learning to rank, and relations between learning to rank and the other learning tasks. I will then describe learning to rank methods developed in recent years, including pointwise, pairwise, and listwise approaches.
Nội dung trích xuất từ tài liệu:
Báo cáo khoa học: "Learning to Rank" Learning to Rank Hang Li Microsoft Research Asia 4F Sigma Building, No 49 Zhichun Road, Haidian, Beijing China hangli@microsoft.com1 Introduction 3. Learning to Rank MethodsIn this tutorial I will introduce ‘learning to rank’, (a) Pointwise Approacha machine learning technology on constructing a i. McRankmodel for ranking objects using training data. I (b) Pairwise Approachwill first explain the problem formulation of learn- i. Ranking SVMing to rank, and relations between learning to ii. RankBoostrank and the other learning tasks. I will then de- iii. RankNetscribe learning to rank methods developed in re-cent years, including pointwise, pairwise, and list- iv. IR SVMwise approaches. I will then give an introduction (c) Listwise Approachto the theoretical work on learning to rank and the i. ListNetapplications of learning to rank. Finally, I will ii. ListMLEshow some future directions of research on learn- iii. AdaRanking to rank. The goal of this tutorial is to give the iv. SVM Mapaudience a comprehensive survey to the technol- v. PermuRankogy and stimulate more research on the technol- vi. SoftRankogy and application of the technology to natural (d) Other Methodslanguage processing. Learning to rank has been successfully applied 4. Learning to Rank Theoryto information retrieval and is potentially useful (a) Pairwise Approachfor natural language processing as well. In factmany NLP tasks can be formalized as ranking i. Generalization Analysisproblems and NLP technologies may be signifi- (b) Listwise Approachcantly improved by using learning to rank tech- i. Generalization Analysisniques. These include question answering, sum- ii. Consistency Analysismarization, and machine translation. For exam-ple, in machine translation, given a sentence in the 5. Learning to Rank Applicationssource language, we are to translate it to a sentence (a) Search Rankingin the target language. Usually there are multi- (b) Collaborative Filteringple possible translations and it would be better to (c) Key Phrase Extractionsort the possible translations in descending orderof their likelihood and output the sorted results. (d) Potential Applications in Natural Lan-Learning to rank can be employed in the task. guage Processing 6. Future Directions for Learning to Rank Re-2 Outline search 1. Introduction 7. Conclusion 2. Learning to Rank Problem (a) Problem Formulation (b) Evaluation 5 Tutorial Abstracts of ACL-IJCNLP 2009, page 5, Suntec, Singapore, 2 August 2009. c 2009 ACL and AFNLP