Social judgments are usually made in the context of complex information including verbal cues.Here we investigated the impact of verbal statements on social judgments by biasing male and female neutral faces with desc...Social judgments are usually made in the context of complex information including verbal cues.Here we investigated the impact of verbal statements on social judgments by biasing male and female neutral faces with descriptives of differentially-valenced behaviour（criticizing or praising） targeting others or objects. Results showed significant main effects of valence and target, such that critical individuals were rated lower in likeability than praising ones and those targeting others relative to objects were valued less. In particular, those who criticized others were the most unlikeable. Among critical individuals, men were less likeable than women. Similarly, men became less valued while targeting others. Overall these findings suggest that the negative impact of critical attributes may trigger avoidance in social interaction while the positive impact of praise may trigger approach.展开更多
针对现有的序列化模型对中文隐式情感分析中特征信息提取不准确以及对篇章级的文本信息提取存在的梯度爆炸或者梯度消失的问题,提出了双向长短时神经网络和上下文感知的树形递归神经网络(context-aware tree recurrent neutral network,...针对现有的序列化模型对中文隐式情感分析中特征信息提取不准确以及对篇章级的文本信息提取存在的梯度爆炸或者梯度消失的问题,提出了双向长短时神经网络和上下文感知的树形递归神经网络(context-aware tree recurrent neutral network,CA-TRNN)的并行混合模型。该模型分别利用双向循环长短时记忆神经网络(BiLSTM)提取文本中的上下文信息,树形递归神经网络(TRNN)提取文本中目标句的语义特征信息,最后,使用特定目标句的注意力机制将两个表示信息进行融合表示后,经过softmax得出文本的情感分类结果。采用SMP2019微博中文隐式情感分析任务中的数据进行验证,实验结果表明,所使用的模型(CA-TRNN)可以有效提高分类结果的准确度,时间代价小,具有更好的应用能力。展开更多
基金supported by the National Natural Science Foundation of China under Grant No.31600880Chinese Fundamental Research Funding for Central Universities under Grant No.ZYGX2015002Interdisciplinary Development under Grant No.Y03111023901014007
文摘Social judgments are usually made in the context of complex information including verbal cues.Here we investigated the impact of verbal statements on social judgments by biasing male and female neutral faces with descriptives of differentially-valenced behaviour（criticizing or praising） targeting others or objects. Results showed significant main effects of valence and target, such that critical individuals were rated lower in likeability than praising ones and those targeting others relative to objects were valued less. In particular, those who criticized others were the most unlikeable. Among critical individuals, men were less likeable than women. Similarly, men became less valued while targeting others. Overall these findings suggest that the negative impact of critical attributes may trigger avoidance in social interaction while the positive impact of praise may trigger approach.
文摘针对现有的序列化模型对中文隐式情感分析中特征信息提取不准确以及对篇章级的文本信息提取存在的梯度爆炸或者梯度消失的问题,提出了双向长短时神经网络和上下文感知的树形递归神经网络(context-aware tree recurrent neutral network,CA-TRNN)的并行混合模型。该模型分别利用双向循环长短时记忆神经网络(BiLSTM)提取文本中的上下文信息,树形递归神经网络(TRNN)提取文本中目标句的语义特征信息,最后,使用特定目标句的注意力机制将两个表示信息进行融合表示后,经过softmax得出文本的情感分类结果。采用SMP2019微博中文隐式情感分析任务中的数据进行验证,实验结果表明,所使用的模型(CA-TRNN)可以有效提高分类结果的准确度,时间代价小,具有更好的应用能力。
