The dorsal area of the anterior cingulate cortex (ACC) constructs the salience network associated with the anterior insular cortex. Conventional brain imaging studies, such as functional magnetic resonance imaging (fM...The dorsal area of the anterior cingulate cortex (ACC) constructs the salience network associated with the anterior insular cortex. Conventional brain imaging studies, such as functional magnetic resonance imaging (fMRI), have demonstrated that relational memory formation occurs in the ACC. However, how such memory is encoded and retrieved remains unknown due to limited time resolution of conventional fMRI. This study aimed to investigate temporal dynamics of the dorsal ACC (dACC) during word-pair tasks based on a newly developed event-related deep brain activity (ER-DBA) method using occipital electroencephalogram (EEG) signal powers. The method assesses dACC activity at a temporal resolution of approximately 0.3 s beyond the conventional resolution limit. We found that transient deactivation of dACC during the presentation of the second word of each pair was essential for encoding success regardless of whether the words were related or unrelated. We also found that memory accuracy was not affected by the intervention of inter-trials until the recall trial. Taken together, these findings suggest that dACC deactivation for encoding success is accompanied with short-term potentiation essential for durability of memory. We further found that false memory formation associated with the presentation of word pairs was occasionally committed. In such cases, dACC exhibited a similar transient deactivation although false memory commission was independent of related or unrelated conditions. Our findings suggest that encoding and retrieval of associates are paralleled and that simultaneous production of associates seems to be an essential strategy for successful relational memory formation. The study was limited to the assessment of dACC activity and did not account for other regional brain activities or receptor regulation related to short-term potentiation. We detected fast behavior of dACC during relational memory formation using the novel ER-DBA method. Such temporal dynamics will be important for eliciting underlying m展开更多
Sentence alignment provides multi-lingual or cross-lingual natural language processing(NLP)applications with high-quality parallel sentence pairs.Normally,an aligned sentence pair contains multiple aligned words,which...Sentence alignment provides multi-lingual or cross-lingual natural language processing(NLP)applications with high-quality parallel sentence pairs.Normally,an aligned sentence pair contains multiple aligned words,which intuitively play different roles during sentence alignment.Inspired by this intuition,we propose to deal with the problem of sentence alignment by exploring the semantic interactionship among fine-grained word pairs within the framework of neural network.In particular,we first employ various relevance measures to capture various kinds of semantic interactions among word pairs by using a word-pair relevance network,and then model their importance by using a multi-view attention network.Experimental results on both monotonic and non-monotonic bitexts show that our proposed approach significantly improves the performance of sentence alignment.展开更多
文摘The dorsal area of the anterior cingulate cortex (ACC) constructs the salience network associated with the anterior insular cortex. Conventional brain imaging studies, such as functional magnetic resonance imaging (fMRI), have demonstrated that relational memory formation occurs in the ACC. However, how such memory is encoded and retrieved remains unknown due to limited time resolution of conventional fMRI. This study aimed to investigate temporal dynamics of the dorsal ACC (dACC) during word-pair tasks based on a newly developed event-related deep brain activity (ER-DBA) method using occipital electroencephalogram (EEG) signal powers. The method assesses dACC activity at a temporal resolution of approximately 0.3 s beyond the conventional resolution limit. We found that transient deactivation of dACC during the presentation of the second word of each pair was essential for encoding success regardless of whether the words were related or unrelated. We also found that memory accuracy was not affected by the intervention of inter-trials until the recall trial. Taken together, these findings suggest that dACC deactivation for encoding success is accompanied with short-term potentiation essential for durability of memory. We further found that false memory formation associated with the presentation of word pairs was occasionally committed. In such cases, dACC exhibited a similar transient deactivation although false memory commission was independent of related or unrelated conditions. Our findings suggest that encoding and retrieval of associates are paralleled and that simultaneous production of associates seems to be an essential strategy for successful relational memory formation. The study was limited to the assessment of dACC activity and did not account for other regional brain activities or receptor regulation related to short-term potentiation. We detected fast behavior of dACC during relational memory formation using the novel ER-DBA method. Such temporal dynamics will be important for eliciting underlying m
基金The work was supported by the National Natural Science Foundation of China under Grant Nos.61876120,61751206,and 61673290.
文摘Sentence alignment provides multi-lingual or cross-lingual natural language processing(NLP)applications with high-quality parallel sentence pairs.Normally,an aligned sentence pair contains multiple aligned words,which intuitively play different roles during sentence alignment.Inspired by this intuition,we propose to deal with the problem of sentence alignment by exploring the semantic interactionship among fine-grained word pairs within the framework of neural network.In particular,we first employ various relevance measures to capture various kinds of semantic interactions among word pairs by using a word-pair relevance network,and then model their importance by using a multi-view attention network.Experimental results on both monotonic and non-monotonic bitexts show that our proposed approach significantly improves the performance of sentence alignment.
