ABSTRACT
This paper studies quality of human labels used to train search engines' rankers. Our specific focus is performance improvements obtained by using overlapping relevance labels, which is by collecting multiple human judgments for each training sample. The paper explores whether, when, and for which samples one should obtain overlapping training labels, as well as how many labels per sample are needed. The proposed selective labeling scheme collects additional labels only for a subset of training samples, specifically for those that are labeled relevant by a judge. Our experiments show that this labeling scheme improves the NDCG of two Web search rankers on several real-world test sets, with a low labeling overhead of around 1.4 labels per sample. This labeling scheme also outperforms several methods of using overlapping labels, such as simple k-overlap, majority vote, the highest labels, etc. Finally, the paper presents a study of how many overlapping labels are needed to get the best improvement in retrieval accuracy.
- A. Bernstein and J. Li. From active towards interactive learning: Using consideration information to improve labeling correctness. University of Zurich, Dynamic and distributed information systems group working paper. www.ifi.uzh.ch/ddis/nc/publications.Google Scholar
- C. J. C. Burges, R. Ragno, and Q. V. Le. Learning to rank with nonsmooth cost functions. NIPS'06.Google Scholar
- Kalervo Järvelin and Jaana Kekäläinen. Ir evaluation methods for retrieving highly relevant documents. SIGIR'00. Google ScholarDigital Library
- NAACL-HLT 2010 Workshop on Amazon Mechanical Turk. http://sites.google.com/site/amtworkshop2010/.Google Scholar
- V. S. Sheng, F. Provost, and P. G. Lpeirotis. Get another label? Improving data quality and data mining using multiple noisy labelers. KDD'08. Google ScholarDigital Library
- P. Smyth, U. M. Fayyad, M. C. Burl, P. Perona, and P. Baldi. Inferring ground truth from subjective labeling of Venus images. NIPS'94.Google Scholar
- P. Smyth, M. C. Burl, U. M. Fayyad, P. Perona. Knowledge discovery in large image databases: Dealing with uncertainties in ground truth. AAAI Knowledge Discovery in Databases Workshop of KDD'94.Google Scholar
- P. Smyth. Bounds on the mean classification error rate of multiple experts. Pattern Recognition Letters 17, 12. 1996. Google ScholarDigital Library
- R. Snow, B. O'Connor, D. Jurafsky and A. Y. Ng. 2008. Cheap and Fast - But is it Good? Evaluating Non-Expert Annotations for Natural Language Tasks. EMNLP'08. Google ScholarDigital Library
- E. M. Voorhees. Evaluating by highly relevant documents. SIGIR'01. Google ScholarDigital Library
- Q. Wu, C. J. C. Burges, K. M. Svore and J. Gao. Ranking, Boosting, and Model Adaptation. Microsoft Research Technical Report MSR-TR-2008-109.Google Scholar
Index Terms
- Collecting high quality overlapping labels at low cost
Recommendations
Training query filtering for semi-supervised learning to rank with pseudo labels
Semi-supervised learning is a machine learning paradigm that can be applied to create pseudo labels from unlabeled data for learning a ranking model, when there is only limited or no training examples available. However, the effectiveness of semi-...
Clustering-based transduction for learning a ranking model with limited human labels
CIKM '13: Proceedings of the 22nd ACM international conference on Information & Knowledge ManagementTransductive learning is a semi-supervised learning paradigm that can leverage unlabeled data by creating pseudo labels for learning a ranking model, when there is only limited or no training examples available. However, the effectiveness of ...
Multi-Label Low-dimensional Embedding with Missing Labels
A novel multi-label method is developed to handle label recovery to enhance classifying performance.The proposed method exploits the low rank property to complete a label imputation process.An inductive classifier is trained using the underling ...
Comments