Massey Research Online - Browsing by Author "Hou F"

Browsing by Author "Hou F"

Now showing 1 - 9 of 9

Anisotropic span embeddings and the negative impact of higher-order inference for coreference resolution: An empirical analysis
(Cambridge University Press, 2024-01-25) Hou F; Wang R; Ng S-K; Zhu F; Witbrock M; Cahan SF; Chen L; Jia X
Coreference resolution is the task of identifying and clustering mentions that refer to the same entity in a document. Based on state-of-the-art deep learning approaches, end-to-end coreference resolution considers all spans as candidate mentions and tackles mention detection and coreference resolution simultaneously. Recently, researchers have attempted to incorporate document-level context using higher-order inference (HOI) to improve end-to-end coreference resolution. However, HOI methods have been shown to have marginal or even negative impact on coreference resolution. In this paper, we reveal the reasons for the negative impact of HOI coreference resolution. Contextualized representations (e.g., those produced by BERT) for building span embeddings have been shown to be highly anisotropic. We show that HOI actually increases and thus worsens the anisotropy of span embeddings and makes it difficult to distinguish between related but distinct entities (e.g., pilots and flight attendants). Instead of using HOI, we propose two methods, Less-Anisotropic Internal Representations (LAIR) and Data Augmentation with Document Synthesis and Mention Swap (DSMS), to learn less-anisotropic span embeddings for coreference resolution. LAIR uses a linear aggregation of the first layer and the topmost layer of contextualized embeddings. DSMS generates more diversified examples of related but distinct entities by synthesizing documents and by mention swapping. Our experiments show that less-anisotropic span embeddings improve the performance significantly (+2.8 F1 gain on the OntoNotes benchmark) reaching new state-of-the-art performance on the GAP dataset.
C3 and C4 grass species: who can reduce soil nitrous oxide emissions in a continental arid region?
(MDPI AG, 8/09/2020) Ning J; He XZ; Hou F
In order to relieve grazing pressure, drought-tolerant grass species are widely cultivated in arid regions. However, soil N emission is largely neglected while pursuing forage yield. We carried out a randomized block study to investigate whether and how the C3 and C4 grass species differ in soil N emission in a typical salinized field with temperate continental arid climate in the northwest inland regions, China. We quantified soil N2O flux from two C3 (barley and rye) and two C4 grass species [corngrass and sorghum hybrid sudangrass (SHS)] in fields during the growing season (from May to September) by using the static box method, and then determined the relationships between soil N2O fluxes and forage yield and soil properties. Results show that soil available nitrogen, soil temperature, pH, soil organic carbon, and total nitrogen were correlated, but soil water was anti-correlated with soil N2O fluxes. In addition, N2O flux increased significantly faster with soil temperature in C4 than in C3 grass fields. Although the lower total N2O emission fluxes were detected for C3 species, the lower yield-scaled N2O was detected for C4 species. Our study provided insights into the determination of grass species and the understanding of mechanisms regulating N2O fluxes in C3 and C4 species in the continental arid regions.
Different effects of grazing and nitrogen addition on ecosystem multifunctionality are driven by changes in plant resource stoichiometry in a typical steppe
(5/08/2022) Li L; He XZ; Zhang X; Hu J; Wang M; Wang Z; Hou F
Purpose: Herbivore grazing and nitrogen (N) input may alter the multiple ecosystem functions (i.e., multifunctionality, hereafter) associated with carbon (C), N, and phosphorus (P) cycling. Most studies on variations in plant diversity, soil biotic or abiotic factors, and linkages to ecosystem functions have focused on grazing or N enrichment alone. Few studies have combined these two factors to explore the role of plant resource stoichiometry (C:N:P ratios) in ecosystem multifunctionality (EMF) control. Here, we evaluated the direct and indirect effects of stocking rate (0, 2.7, 5.3, and 8.7 sheep ha− 1) and N addition rate (0, 5, 10, and 20 g N m− 2 yr− 1) on a range of ecosystem functions and EMF via changing plant diversity, soil pH and plant resource stoichiometry in a typical steppe on the Loess Plateau. Results: We found that increasing stocking rate and interaction between grazing and N addition significantly decreased EMF, while increasing N addition rate significantly promoted EMF. Grazing decreased soil NH4+-N, soil NO3−-N, aboveground biomass, and plant C, N, and P pools, but increased soil total N and P at 8.7 and 5.3 sheep ha− 1, respectively. N addition increased soil NH4+-N, NO3−-N, and total P. Plant aboveground biomass, and plant C, N, and P pools increased at the lower N addition rate (≤ 5 g N m− 2 yr− 1) under grazing. The structural equation models indicated that (1) EMF was driven by the direct effects of grazing and the indirect effects of grazing on plant resource stoichiometry and soil pH; (2) EMF increased with increasing N addition rates, but such positive response of EMF to increasing N addition rates was alleviated at high levels of plant resource stoichiometry and diversity; and (3) the indirect effects of plant diversity induced by grazing and N addition had moderate effects on EMF via the variations of plant resource stoichiometry. Conclusions: This study demonstrated grazing and N addition had contrasting effects on ecosystem multifunctionality in a typical steppe, and highlighted the capacity of plant diversity in balancing plant elements that serve as a key mechanism in the maintenance of EMF in response to intensive grazing and N enrichment.
Grazing activity increases decomposition of yak dung and litter in an alpine meadow on the Qinghai-Tibet plateau
(Springer Nature Switzerland AG on behalf of the Royal Netherlands Society of Agricultural Science, 2019-11) Yang C; Zhang Y; Hou F; Millner JP; Wang Z; Chang S; Shang Z
Aims: This study investigated the influences of herbivore grazing intensity and grazing season on decomposition and nutrient release of dung and litter, which aimed to improve our understandings of grazing affecting nutrient cycling in alpine meadows on the Qinghai-Tibetan Platean. Methods: A factorial design experiment comprising 3 grazing intensities (non-grazing, moderate grazing, and heavy grazing) and 2 grazing seasons (summer and winter), was applied to quantify the decomposition and chemistry of dung and litter in an alpine pasture using the litterbag technique. Litterbags were retrieved for analysis of mass loss and nutrient release with 180, 360, 540, and 720 days after placement. Results: Grazing activity accelerated the decomposition of dung and litter and increased nutrient release from dung and litter by increasing soil temperature compared with non-grazing pastures, whereas grazing season had no effect on decomposition. The decomposition time was shorter for dung than that for litter. Conclusions: Herbivores grazing benefited dung and litter decomposition and nutrient cycling directly by increasing soil temperature, which is likely to promote soil microbial activity due to low temperatures in alpine meadows, and indirectly through herbage ingestion and dung deposition which increase the organic debris concentration used for microorganisms growth and reproduction. This study provides insights into the mechanisms of grazing regulating nutrient cycling in alpine ecosystems.
Learning and integration of adaptive hybrid graph structures for multivariate time series forecasting
(Elsevier Inc., 2023-11-01) Guo T; Hou F; Pang Y; Jia X; Wang Z; Wang R
Recent status-of-the-art methods for multivariate time series forecasting can be categorized into graph-based approach and global-local approach. The former approach uses graphs to represent the dependencies among variables and apply graph neural networks to the forecasting problem. The latter approach decomposes the matrix of multivariate time series into global components and local components to capture the shared information across variables. However, both approaches cannot capture the propagation delay of the dependencies among individual variables of a multivariate time series, for example, the congestion at intersection A has delayed effects on the neighboring intersection B. In addition, graph-based forecasting methods cannot capture the shared global tendency across the variables of a multivariate time series; and global-local forecasting methods cannot reflect the nonlinear inter-dependencies among variables of a multivariate time series. In this paper, we propose to combine the advantages of both approaches by integrating Adaptive Global-Local Graph Structure Learning with Gated Recurrent Units (AGLG-GRU). We learn a global graph to represent the shared information across variables. And we learn dynamic local graphs to capture the local randomness and nonlinear dependencies among variables. We apply diffusion convolution and graph convolution operations to global and dynamic local graphs to integrate the information of graphs and update gated recurrent unit for multivariate time series forecasting. The experimental results on seven representative real-world datasets demonstrate that our approach outperforms various existing methods.
Pasture Performance: Perspectives on Plant Persistence and Renewal in New Zealand Dairy Systems
(MDPI (Basel, Switzerland), 2024-08) Cartmill AD; Donaghy DJ; Hou F
Pasture systems dominate the landscape of Aotearoa, New Zealand, and are an integral component of sustainable and resilient livestock production. Predicting the response, performance, and dynamics of pasture species and adapting management practices is key to the long-term economic and environmental sustainability and resilience of the agricultural sector. However, there is limited information on the long-term productivity, performance, and persistence of forage cultivars and species for pasture production systems, particularly when linked to grazing and animal performance. Here, we sought to reduce scientific uncertainty, inform modelling efforts, and contribute to a predictive framework for understanding pasture performance, persistence, and renewal. Inter-annual pasture renewal (direct drilling and cultivation) rates vary by region and year, reflecting both opportunity and problem-based drivers, with the highest pasture renewal rates in Waikato and Canterbury on the North and South Island, respectively.
Real and synthetic Punjabi speech datasets for automatic speech recognition
(Elsevier Inc, 2024-02) Singh S; Hou F; Wang R
Automatic speech recognition (ASR) has been an active area of research. Training with large annotated datasets is the key to the development of robust ASR systems. However, most available datasets are focused on high-resource languages like English, leaving a significant gap for low-resource languages. Among these languages is Punjabi, despite its large number of speakers, Punjabi lacks high-quality annotated datasets for accurate speech recognition. To address this gap, we introduce three labeled Punjabi speech datasets: Punjabi Speech (real speech dataset) and Google-synth/CMU-synth (synthesized speech datasets). The Punjabi Speech dataset consists of read speech recordings captured in various environments, including both studio and open settings. In addition, the Google-synth dataset is synthesized using Google's Punjabi text-to-speech cloud services. Furthermore, the CMU-synth dataset is created using the Clustergen model available in the Festival speech synthesis system developed by CMU. These datasets aim to facilitate the development of accurate Punjabi speech recognition systems, bridging the resource gap for this important language.
Seasonal variation in soil and herbage CO2 efflux for a sheep-grazed alpine meadow on the north-east Qinghai-Tibetan Plateau and estimated net annual CO2 exchange
(2/06/2022) Yuan H; Matthew C; He XZ; Sun Y; Liu Y; Zhang T; Gao X; Yan C; Chang S; Hou F
The Qinghai-Tibetan Plateau is a vast geographic area currently subject to climate warming. Improved knowledge of the CO2 respiration dynamics of the Plateau alpine meadows and of the impact of grazing on CO2 fluxes is highly desirable. Such information will assist land use planning. We measured soil and vegetation CO2 efflux of alpine meadows using a closed chamber technique over diurnal cycles in winter, spring and summer. The annual, combined soil and plant respiration on ungrazed plots was 28.0 t CO2 ha-1 a-1, of which 3.7 t ha-1 a-1occurred in winter, when plant respiration was undetectable. This suggests winter respiration was driven mainly by microbial oxidation of soil organic matter. The winter respiration observed in this study was sufficient to offset the growing season CO2 sink reported for similar alpine meadows in other studies. Grazing increased herbage respiration in summer, presumably through stimulation of gross photosynthesis. From limited herbage production data, we estimate the sustainable yield of these meadows for grazing purposes to be about 500 kg herbage dry matter ha-1 a-1. Addition of photosynthesis data and understanding of factors affecting soil carbon sequestration to more precisely determine the CO2 balance of these grasslands is recommended.
Soil C, N, and P stocks evaluation under major land uses on China’s Loess Plateau
(Society for Range Management, 1/03/2017) Chen X; Hou F; Matthew C; He XZ
Loess Plateau covers 640 000 km2 in the central northern China. Despite a semiarid environment, harsh winters, and hot summers, agriculture has been practiced in this region for > 5 000 yr, and the food production systems are among China's oldest. The environment is fragile because the loessial soils are prone to erosion. Sound scientific information is therefore required to underpin future land use planning in the region. To this end, total soil organic carbon (SOC), N, and P stocks were measured in Huanxian County of the wider Loess Plateau, representing five major land use categories. Sites were sampled three times over 3 yr. In all, almost 2 800 soil analyses were performed. A feature of these soils is low SOC content in the A horizon but comparatively small decline with soil depth. For example, SOC levels for the 0-20 cm and 70-100 cmsoil depths averaged 6.1 and 4.1Mg ha-1, respectively. Alfalfa and rangeland sites had 5.1 Mg ha-1 (10%) more total than cropland and 7.5 t ha-1 (16%) more total SOC to 100-cm soil depth than the two silvopastoral sites. For total soil N (0- to 100-cm soil depth) the averages of alfalfa and RL siteswere 20% and 28%, respectively, higher than the cropland and silvopastoral site group means, although soil C, N, and P levels are very low, relative to those of typical soils elsewhere. When these observations are scaled up to a regional level, it can be calculated that a 5% shift in land use from cropping or silvopastoral systems to alfalfa-based systems could increase soil C sequestration by as many as 20 million t CO2 per yr, although some caution is needed in making extrapolations, as the present data are from a single locality on the Loess Plateau.

Browsing by Author "Hou F"

Results Per Page

Sort Options