Current Location:home > Browse
Your conditions: 江西师范大学(3)

1. chinaXiv:202201.00009 [pdf]

面向低资源命名实体识别BiLSTM-BCRF模型

钟茂生; 吴佳华
Subjects: Computer Science >> Other Disciplines of Computer Science

[目的]当标注数据较少时,现有模型受训练数据量少的限制,参数没有拟合到预期效果,导致在低资源命名实体识别任务中模型识别性能不佳。[方法]本文提出一种融入伯努利分布(Bernoulli distribution)的新型损失函数,让模型较好拟合数据。此外,本文在BiLSTM-CRF模型基础上融合多层字符特征信息,结合基于伯努利分布的新型损失函数,构建了BiLSTM-BCRF模型。[结果]本文提出的BiLSTM-BCRF模型在20%的CoNLL2003和20%的BC5CDR的数据集上,F1值在BiLSTM-CRF模型基础上分别提升了6.16%、3.35%。[结论]该模型能较好地适应低资源命名实体识别任务。[局限]该模型识别专有名词的性能还有待提升。

submitted time 2022-01-02 Hits2524Downloads107 Comment 1

2. chinaXiv:202107.00074 [pdf]

一种高效的CD-CAT在线标定新方法:基于熵的信息增益与EM视角

谭青蓉; 汪大勋; 罗芬; 蔡艳; 涂冬波
Subjects: Psychology >> Psychological Measurement

项目增补(Item Replenishing)对认知诊断计算机自适应测验(CD-CAT)题库的维护有着至关重要的作用,而在线标定是一种重要的项目增补方式。基于数据挖掘中特征选择(Feature Selection)的思路,提出一种高效的基于熵的信息增益的在线标定方法(记为IGEOCM),该方法利用被试在新旧题上的作答联合估计新题的Q矩阵和项目参数。研究采用Monte Carlo模拟实验验证所开发新方法的效果,并同时与已有的在线标定方法SIE (Chen et al., 2015)、SIE-R-BIC和RMSEA-N (谭青蓉,2019)进行比较。结果表明:新开发的IGEOCM在各实验条件下均具有较好的项目标定精度和项目估计效率,且整体上优于已有的SIE等方法;同时,IGEOCM标定新题所需的时间低于SIE等方法。总之,研究为CD-CAT题库中项目的增补提供了一种更为高效、准确的方法。

submitted time 2021-07-30 Hits659Downloads350 Comment 0

3. chinaXiv:202004.00019 [pdf]

The phylogenetic position of Cladiucha within Tenthredinidae based on comprehensive mitochondrial phylogenomics and the evidence from comparative analyses of rRNA secondary structure

Niu, Gengyun; Jiang Sijia; Do?an, ?zgül ; Korkmaz, Ertan Mahir; Budak, Mahir ; Wu, Duo; Wei, Meicai
Subjects: Biology >> Entomology

Two mitogenomes of Cladiucha were newly reported and shown typical pattern of gene arrangement. The phylogenetic position of Cladiucha was obtained from tree reconstruction using various data treatment methods and substitution models. Significant heterogeneity in the nucleotide composition and mutational biases was found in the mitochondrial protein-coding genes, and the third codon position exhibited high levels of saturation. Therefore, 14 datasets were conducted under both site-homogeneous and site-heterogeneous models. The following conclusions were drawn from the phylogenetic analyses: (i) the monophyly of Tenthredinidae was confirmed, (ii) the monophyly of Allantinae + Tenthredininae + Megabelesinae was approved, and (iii) within the family, ((((Tenthredininae + Fenusinae) + Allantinae) + Megabelesinae) + Nematinae) is probably the most acceptable cladogram for the phylogeny of Tenthredinidae, which is also supported by the morphological analysis and a comparative study on the rRNA secondary structure. Divergence time estimation analyses indicated that diversification of the major superfamilies of the suborder Symphyta occurred around 232.9 Ma, and the splits of Tenthredinidae were dated to 146 Ma, which corresponding to the origin of the earliest lineages of flowering plants and major diversifications of core angiosperms, respectively. The Cladiucha arose in the Mid-Miocene; at that time, magnoliids are rapidly undergoing genus-species differentiation. "

submitted time 2020-04-19 Hits12572Downloads1646 Comment 0

  [1 Pages/ 3 Totals]