Integrative Semantic Dependency Parsing via Efficient Large-scale Feature Selection

Semantic parsing, i.e., the automatic derivation of meaning representation such as an instantiated predicate-argument structure for a sentence, plays a critical role in deep processing of natural language. Unlike all other top systems of semantic dependency parsing that have to rely on a pipeline framework to chain up a series of submodels each specialized for a specific subtask, the one presented in this article integrates everything into one model, in hopes of achieving desirable integrity and practicality for real applications while maintaining a competitive performance. This integrative approach tackles semantic parsing as a word pair classification problem using a maximum entropy classifier. We leverage adaptive pruning of argument candidates and large-scale feature selection engineering to allow the largest feature space ever in use so far in this field, it achieves a state-of-the-art performance on the evaluation data set for CoNLL-2008 shared task, on top of all but one top pipeline system, confirming its feasibility and effectiveness.

Download Full-text

Self-Adaptive Particle Swarm Optimization for Large-Scale Feature Selection in Classification

ACM Transactions on Knowledge Discovery from Data ◽

10.1145/3340848 ◽

2019 ◽

Vol 13 (5) ◽

pp. 1-27 ◽

Cited By ~ 31

Author(s):

Yu Xue ◽

Bing Xue ◽

Mengjie Zhang

Keyword(s):

Feature Selection ◽

Particle Swarm Optimization ◽

Large Scale ◽

Particle Swarm ◽

Swarm Optimization ◽

Adaptive Particle Swarm Optimization ◽

Scale Feature ◽

Self Adaptive

Download Full-text

Modelling above-ground carbon dynamics using multi-temporal airborne lidar: insights from a Mediterranean woodland

Biogeosciences ◽

10.5194/bg-13-961-2016 ◽

2016 ◽

Vol 13 (4) ◽

pp. 961-973 ◽

Cited By ~ 17

Author(s):

W. Simonson ◽

P. Ruiz-Benito ◽

F. Valladares ◽

D. Coomes

Keyword(s):

Climate Change ◽

Large Scale ◽

Critical Role ◽

Carbon Fluxes ◽

Carbon Dynamics ◽

Airborne Lidar ◽

Lidar Data ◽

Low Carbon ◽

Data Set ◽

Multi Temporal

Abstract. Woodlands represent highly significant carbon sinks globally, though could lose this function under future climatic change. Effective large-scale monitoring of these woodlands has a critical role to play in mitigating for, and adapting to, climate change. Mediterranean woodlands have low carbon densities, but represent important global carbon stocks due to their extensiveness and are particularly vulnerable because the region is predicted to become much hotter and drier over the coming century. Airborne lidar is already recognized as an excellent approach for high-fidelity carbon mapping, but few studies have used multi-temporal lidar surveys to measure carbon fluxes in forests and none have worked with Mediterranean woodlands. We use a multi-temporal (5-year interval) airborne lidar data set for a region of central Spain to estimate above-ground biomass (AGB) and carbon dynamics in typical mixed broadleaved and/or coniferous Mediterranean woodlands. Field calibration of the lidar data enabled the generation of grid-based maps of AGB for 2006 and 2011, and the resulting AGB change was estimated. There was a close agreement between the lidar-based AGB growth estimate (1.22 Mg ha−1 yr−1) and those derived from two independent sources: the Spanish National Forest Inventory, and a tree-ring based analysis (1.19 and 1.13 Mg ha−1 yr−1, respectively). We parameterised a simple simulator of forest dynamics using the lidar carbon flux measurements, and used it to explore four scenarios of fire occurrence. Under undisturbed conditions (no fire) an accelerating accumulation of biomass and carbon is evident over the next 100 years with an average carbon sequestration rate of 1.95 Mg C ha−1 yr−1. This rate reduces by almost a third when fire probability is increased to 0.01 (fire return rate of 100 years), as has been predicted under climate change. Our work shows the power of multi-temporal lidar surveying to map woodland carbon fluxes and provide parameters for carbon dynamics models. Space deployment of lidar instruments in the near future could open the way for rolling out wide-scale forest carbon stock monitoring to inform management and governance responses to future environmental change.

Download Full-text