The Gender Congruency Effect and the Selection of Freestanding and Bound Morphemes: Evidence From Croatian.

The authors report 3 picture-word interference experiments in which they explore some properties of the agreement process in speech production. In Experiment 1, Croatian speakers were asked to produce utterances in which the noun's gender value had an impact on the selection of gender-marked freestanding morphemes (pronouns) while ignoring the presentation of same- or different-gender distractor words. In Experiments 2 and 3, Croatian speakers were asked to name the same pictures using noun phrases in which the noun's gender value surfaced as an inflectional suffix. Different-gender distractors interfered more than same-gender distractors (the gender congruency effect) in Experiment 1, but not in Experiments 2 and 3. These contrasting results show that the cause of the gender congruency effect is not at the level where lexical-grammatical information is selected but at the level of selection of freestanding morphemes. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

英语单词形素探析

英语学习的困难之一就是单词的识记，而单词的构成有它内在的规律，只要抓住了这些规律，英语词汇的学习就变得比较容易。本文论述英语单词是由形体要素构成的，这些形素构成了英语字母有意义或无意义的单元。在学习英语的过程中，有意识地注意和分析这些形体单元，可以高效地识记大量的英语词汇。     

基于多粒度特征的文本生成评价方法

近年来,基于预训练语言模型的文本生成评价方法得到了广泛关注,其通过计算两个句子间子词粒度的相似度来评价生成文本的质量.但是对于越南语、泰语等存在大量黏着语素的语言,单个音节或子词不能独立成词表达语义,仅基于子词粒度匹配的方法并不能够完整表征两个句子间的语义相似关系.基于此,该文提出一种基于子词、音节、词组等多粒度特征的...     

汉语的语素概念提取与语义构词分析

作为基础的表义单位,语素及此上的构词分析,既是汉语作为意合语言进行语义分析的起点,也是认知、理解词义的关键。该文提出了一种探寻汉语语义基元和分析词义的新的视角和方法: 基于语素义相似度计算形成“同义语素集”,用来表征“语素概念”,并借鉴生成词库理论形成“语素概念体系”;建立在此基础上的汉语语义构词分析,在全局性语义分析、数据挖掘等方面也有新的进展。这些思路、做法及语言资源建设,有望推动人文领域和计算应用等相关工作的开展。     

Different selection principles of freestanding and bound morphemes in language production.

Freestanding and bound morphemes differ in many (psycho)linguistic aspects. Some theorists have claimed that the representation and retrieval of freestanding and bound morphemes in the course of language production are governed by similar processing mechanisms. Alternatively, it has been proposed that both types of morphemes may be selected for production in different ways. In this article, the authors first review the available experimental evidence related to this topic and then present new experimental data pointing to the notion that freestanding and bound morphemes are retrieved following distinct processing principles: freestanding morphemes are subject to competition, bound morphemes not. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The processing of free and bound gender-marked morphemes in speech production: Evidence from Dutch.

In many languages, the production of noun phrases requires the selection of gender-marked elements like determiners or inflectional suffixes. There is a recent debate as to whether the selection of freestanding gender-marked elements, such as determiners, follows the same processing mechanisms as the selection of bound gender-marked morphemes, such as adjective suffixes. Most of the evidence on which this debate is based relates to the gender-congruency effect in picture-word interference experiments. In the present article, the authors address this issue with a pure picture-naming task, extending previous work in German (H. Schriefers, J. D. Jescheniak, & A. Hantsch, 2005). The results of the present study on noun phrase production in Dutch show that both types of gender-marked morphemes are selected via the same basic processing mechanisms. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Prosody-assisted head-driven access to spoken German compounds.

Auditory processing of German 2-noun compound words was investigated with 328 participants in 4 experiments by monitoring semantic priming effects of the left constituents of the compound words. The authors demonstrated that there is no primacy of the left constituents in accessing auditorily presented German compound words in the mental lexicon. A clear priming effect of left constituents occurred only for compound words with a transparent right constituent that is the head of compound words in Germanic languages. The data suggest that the access to German compounds in the auditory domain involves 2 temporally overlapping routes: direct and decompositional. The prosodic structure (i.e., the duration) of the first morphemes of compound words appears to be a determining factor for activation of the decompositional route. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Selection of Gender-Marked Morphemes in Speech Production.

N.O. Schiller and A. Caramazza (2003) and A. Costa, D. Kovacic, E. Fedorenko, and A. Caramazza (2003) have argued that the processing of freestanding gender-marked morphemes (e.g., determiners) and bound gender-marked morphemes (e.g., adjective suffixes) during syntactic encoding in speech production follows distinct principles, with only freestanding morphemes being subject to a competitive selection process. In 3 experiments, the authors tested this hypothesis in German, extending a previous study by H. Schriefers, J.D. Jescheniak, and A. Hantsch (2002). The results suggest that freestanding and bound morphemes are basically processed in the same way, although competition appears to be attenuated for bound morphemes relative to free morphemes. The authors discuss theoretical and methodological implications of this pattern. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

基于词素特征的轻量级域名检测算法

对网络中DNS交互报文进行检测以发现恶意服务,是网络安全监测的一个重要手段,这种检测往往要求系统能够实时或准实时地发现监测域名中的可疑对象.面对庞大的域名集合,若对所有域名使用同样强度的监测通常开销过大.通过挖掘域名字面蕴含的词素（词根、词缀、拼音及缩写）特征,提出一种轻量级检测算法,能够快速锁定可疑域名,以便后续有针对性地进行DPI检测.实验结果表明：基于词素特征的检测算法比统计n元组频率分布的方法虽然略微增加了58.3%的内存开销,但却具备抗逃避能力以及更高的准确率（相对提高35.2%）;与基于单词特征的方法相比,极大地降低了计算复杂度（相对降低64.8%）,并减少了2.6%的内存开销,而准确率仅下降2.5%.     

基于词信息嵌入的汉语构词结构识别研究

作为一种意合型语言,汉语中的构词结构刻画了构词成分之间的组合关系,是认知、理解词义的关键。在中文信息处理领域,此前的构词结构识别工作大多沿用句法层面的粗粒度标签,且主要基于上下文等词间信息建模,忽略了语素义、词义等词内信息对构词结构识别的作用。该文采用语言学视域下的构词结构标签体系,构建汉语构词结构及相关信息数据集,提出了一种基于Bi-LSTM和self-attention的模型,以此来探究词内、词间等多方面信息对构词结构识别的潜在影响和能达到的性能。实验取得了良好的预测效果,准确率达77.87%,F₁值为78.36%;同时,对比测试揭示,词内的语素义信息对构词结构识别具有显著的贡献,而词间的上下文信息贡献较弱且带有较强的不稳定性。