Regularized Least-Squares for Parse Ranking

Tsivtsivadze, Evgeni; Pahikkala, Tapio; Pyysalo, Sampo; Boberg, Jorma; Mylläri, Aleksandr; Salakoski, Tapio

doi:10.1007/11552253_42

Cited by 10 publications

(28 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For a more detailed description, we refer to Tsivtsivadze et al (2005). Next, we describe the task of parse ranking.…”

Section: Ranking Of Dependency Parsesmentioning

confidence: 99%

“…However, the ranking performance of the heuristics has been found to be poor when applied to biomedical text (Pyysalo et al 2006), and hence subsequent ranking or selection methods are needed. In our previous studies, we used regularized least-squares regression for the reranking task that notably outperformed the LG heuristics (Tsivtsivadze et al 2005).…”

Section: Ranking Of Dependency Parsesmentioning

confidence: 99%

“…We obtain a scoring for an input by comparing its parse to the hand annotated correct parse of its sentence. Tsivtsivadze et al (2005) describes in detail how the scores are calculated. The relevant input pairs are those of which both inputs are associated to the same sentence and have different scores.…”

Section: Ranking Of Dependency Parsesmentioning

confidence: 99%

See 2 more Smart Citations

An efficient algorithm for learning to rank from preference graphs

et al. 2009

Self Cite

View full text Add to dashboard Cite

In this paper, we introduce a framework for regularized least-squares (RLS) type of ranking cost functions and we propose three such cost functions. Further, we propose a kernel-based preference learning algorithm, which we call RankRLS, for minimizing these functions. It is shown that RankRLS has many computational advantages compared to the ranking algorithms that are based on minimizing other types of costs, such as the hinge cost. In particular, we present efficient algorithms for training, parameter selection, multiple output learning, cross-validation, and large-scale learning. Circumstances under which these computational benefits make RankRLS preferable to RankSVM are considered. We evaluate RankRLS on four different types of ranking tasks using RankSVM and the standard RLS regression as the baselines. RankRLS outperforms the standard RLS regression and its performance is very similar to that of RankSVM, while RankRLS has several computational benefits over RankSVM.

show abstract

“…For a more detailed description, we refer to Tsivtsivadze et al (2005). Next, we describe the task of parse ranking.…”

Section: Ranking Of Dependency Parsesmentioning

confidence: 99%

Section: Ranking Of Dependency Parsesmentioning

confidence: 99%

See 1 more Smart Citation

An efficient algorithm for learning to rank from preference graphs

et al. 2009

Self Cite

View full text Add to dashboard Cite

show abstract

“…Recently, we proposed a method for dependency parse ranking [8] that uses Regularized Least-Squares (RLS) algorithm [9] and grammatically motivated features. The method, called RLS ranker, worked notably better giving 0.42 correlation compared to 0.16 of the LG heuristics.…”

Section: Introductionmentioning

confidence: 99%

“…In this paper, we propose a Locality-Convolution (LC) kernel that provides a correlation of 0.46 when used in RLS algorithm. In all experiments, we applied the F-score based parse goodness function [8], and evaluated the ranking performance with Kendall's correlation coefficient τ b described in [10].…”

Section: Introductionmentioning

confidence: 99%

Locality-Convolution Kernel and Its Application to Dependency Parse Ranking

Tsivtsivadze

Pahikkala

Boberg

et al. 2006

Advances in Applied Artificial Intelligence

Self Cite

View full text Add to dashboard Cite

Abstract. We propose a Locality-Convolution (LC) kernel in application to dependency parse ranking. The LC kernel measures parse similarities locally, within a small window constructed around each matching feature. Inside the window it makes use of a position sensitive function to take into account the order of the feature appearance. The similarity between two windows is calculated by computing the product of their common attributes and the kernel value is the sum of the window similarities. We applied the introduced kernel together with Regularized Least-Squares (RLS) algorithm to a dataset containing dependency parses obtained from a manually annotated biomedical corpus of 1100 sentences. Our experiments show that RLS with LC kernel performs better than the baseline method. The results outline the importance of local correlations and the order of feature appearance within the parse. Final validation demonstrates statistically significant increase in parse ranking performance.

show abstract