Limits...
Protein interaction sentence detection using multiple semantic kernels.

Polajnar T, Damoulas T, Girolami M - J Biomed Semantics (2011)

Bottom Line: We show that combinations of semantic kernels lead to statistically significant improvements in recognition rates and receiver operating characteristic (ROC) scores over the plain Gaussian kernel, when applied to a well-known labelled collection of abstracts.The results from this paper indicate that using semantic information from unlabelled text, and combinations of such information, can be valuable for classification of short texts such as PPI sentences.The method described herein is modular, and can be applied with a variety of feature types, kernels, and semantic models, in order to facilitate full extraction of interacting proteins.

View Article: PubMed Central - HTML - PubMed

Affiliation: School of Computing Science, University of Glasgow, Glasgow, UK. tamara@dcs.gla.ac.uk.

ABSTRACT

Background: Detection of sentences that describe protein-protein interactions (PPIs) in biomedical publications is a challenging and unresolved pattern recognition problem. Many state-of-the-art approaches for this task employ kernel classification methods, in particular support vector machines (SVMs). In this work we propose a novel data integration approach that utilises semantic kernels and a kernel classification method that is a probabilistic analogue to SVMs. Semantic kernels are created from statistical information gathered from large amounts of unlabelled text using lexical semantic models. Several semantic kernels are then fused into an overall composite classification space. In this initial study, we use simple features in order to examine whether the use of combinations of kernels constructed using word-based semantic models can improve PPI sentence detection.

Results: We show that combinations of semantic kernels lead to statistically significant improvements in recognition rates and receiver operating characteristic (ROC) scores over the plain Gaussian kernel, when applied to a well-known labelled collection of abstracts. The proposed kernel composition method also allows us to automatically infer the most discriminative kernels.

Conclusions: The results from this paper indicate that using semantic information from unlabelled text, and combinations of such information, can be valuable for classification of short texts such as PPI sentences. This study, however, is only a first step in evaluation of semantic kernels and probabilistic multiple kernel learning in the context of PPI detection. The method described herein is modular, and can be applied with a variety of feature types, kernels, and semantic models, in order to facilitate full extraction of interacting proteins.

No MeSH data available.


The weights assigned by pMKL to words at different distances from the target word. The bars represent the kernels constructed from HAL matrices with, from left to right, l = 1 to l = 10.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3116455&req=5

Figure 7: The weights assigned by pMKL to words at different distances from the target word. The bars represent the kernels constructed from HAL matrices with, from left to right, l = 1 to l = 10.

Mentions: Figure 7 shows the estimated weights for kernels constructed with XHl for l = 1 ... 10 (C4, in Table 2). The assigned weightings closely mirror the sparsity of the HAL matrices. Matrices 2 and 3 have the lowest sparsity, and while the contribution of l = 1 seems to be underestimated, l = 3 seems to be overestimated. This would indicate that, perhaps, a scheme weighted by the information stored in matrices representing various window lengths would lead to best performance when applying the HAL algorithm to various tasks. The AUC (0.8883 ± 0.0029) is slightly higher than the uniform combination of these kernels while the F-score (0.6633 ± 0.0080) is significantly lower than in uniform combination of these kernels (C1, in Table 2). Due to the computationally intensive nature of this experiment the parameters for each of the kernels was set at 1, which is one of the parameters the cross-validated tuning approach favoured for XHl = 1 (Figure 6).


Protein interaction sentence detection using multiple semantic kernels.

Polajnar T, Damoulas T, Girolami M - J Biomed Semantics (2011)

The weights assigned by pMKL to words at different distances from the target word. The bars represent the kernels constructed from HAL matrices with, from left to right, l = 1 to l = 10.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3116455&req=5

Figure 7: The weights assigned by pMKL to words at different distances from the target word. The bars represent the kernels constructed from HAL matrices with, from left to right, l = 1 to l = 10.
Mentions: Figure 7 shows the estimated weights for kernels constructed with XHl for l = 1 ... 10 (C4, in Table 2). The assigned weightings closely mirror the sparsity of the HAL matrices. Matrices 2 and 3 have the lowest sparsity, and while the contribution of l = 1 seems to be underestimated, l = 3 seems to be overestimated. This would indicate that, perhaps, a scheme weighted by the information stored in matrices representing various window lengths would lead to best performance when applying the HAL algorithm to various tasks. The AUC (0.8883 ± 0.0029) is slightly higher than the uniform combination of these kernels while the F-score (0.6633 ± 0.0080) is significantly lower than in uniform combination of these kernels (C1, in Table 2). Due to the computationally intensive nature of this experiment the parameters for each of the kernels was set at 1, which is one of the parameters the cross-validated tuning approach favoured for XHl = 1 (Figure 6).

Bottom Line: We show that combinations of semantic kernels lead to statistically significant improvements in recognition rates and receiver operating characteristic (ROC) scores over the plain Gaussian kernel, when applied to a well-known labelled collection of abstracts.The results from this paper indicate that using semantic information from unlabelled text, and combinations of such information, can be valuable for classification of short texts such as PPI sentences.The method described herein is modular, and can be applied with a variety of feature types, kernels, and semantic models, in order to facilitate full extraction of interacting proteins.

View Article: PubMed Central - HTML - PubMed

Affiliation: School of Computing Science, University of Glasgow, Glasgow, UK. tamara@dcs.gla.ac.uk.

ABSTRACT

Background: Detection of sentences that describe protein-protein interactions (PPIs) in biomedical publications is a challenging and unresolved pattern recognition problem. Many state-of-the-art approaches for this task employ kernel classification methods, in particular support vector machines (SVMs). In this work we propose a novel data integration approach that utilises semantic kernels and a kernel classification method that is a probabilistic analogue to SVMs. Semantic kernels are created from statistical information gathered from large amounts of unlabelled text using lexical semantic models. Several semantic kernels are then fused into an overall composite classification space. In this initial study, we use simple features in order to examine whether the use of combinations of kernels constructed using word-based semantic models can improve PPI sentence detection.

Results: We show that combinations of semantic kernels lead to statistically significant improvements in recognition rates and receiver operating characteristic (ROC) scores over the plain Gaussian kernel, when applied to a well-known labelled collection of abstracts. The proposed kernel composition method also allows us to automatically infer the most discriminative kernels.

Conclusions: The results from this paper indicate that using semantic information from unlabelled text, and combinations of such information, can be valuable for classification of short texts such as PPI sentences. This study, however, is only a first step in evaluation of semantic kernels and probabilistic multiple kernel learning in the context of PPI detection. The method described herein is modular, and can be applied with a variety of feature types, kernels, and semantic models, in order to facilitate full extraction of interacting proteins.

No MeSH data available.