摘要 :
Recent work suggests knowledge sources can be added into the topic modeling process to label topics and improve topic discovery. The knowledge sources typically consist of a collection of human-constructed articles, each describin...
展开
Recent work suggests knowledge sources can be added into the topic modeling process to label topics and improve topic discovery. The knowledge sources typically consist of a collection of human-constructed articles, each describing a topic (article-topic) for an entire domain. However, these semisupervised topic models assume a corpus to contain topics on only a subset of a domain. Therefore, during inference, the model must consider which article-topics were theoretically used to generate the corpus. Since the knowledge sources tend to be quite large, the many article-topics considered slow down the inference process. The increase in execution time is significant, with knowledge source input greater than 103 becoming unfeasible for use in topic modeling. To increase the applicability of semisupervised topic models, approaches are needed to speed up the overall execution time. This paper presents a way of ranking knowledge source topics to satisfy the above goal. Our approach utilizes a knowledge source ranking, based on the PageRank algorithm, to determine the importance of an article-topic. By applying our ranking technique we can eliminate low scoring article-topics before inference, speeding up the overall process. Remarkably, this ranking technique can also improve perplexity and interpretability. Results show our approach to outperform baseline methods and significantly aid semisupervised topic models. In our evaluation, knowledge source rankings yield a 44% increase in topic retrieval f-score, a 42.6% increase in inter-inference topic elimination, a 64% increase in perplexity, a 30% increase in token assignment accuracy, a 20% increase in topic composition interpretability, and a 5% increase in document assignment interpretability over baseline methods.
收起
摘要 :
Abstract Fundamental theories of human cognition have long posited that the short‐term maintenance of actions is supported by one of the “core knowledge” systems of human visual cognition, yet its neural substrates are still no...
展开
Abstract Fundamental theories of human cognition have long posited that the short‐term maintenance of actions is supported by one of the “core knowledge” systems of human visual cognition, yet its neural substrates are still not well understood. In particular, it is unclear whether the visual short‐term memory (VSTM) of actions has distinct neural substrates or, as proposed by the spatio‐object architecture of VSTM, shares them with VSTM of objects and spatial locations. In two experiments, we tested these two competing hypotheses by directly contrasting the neural substrates for VSTM of actions with those for objects and locations. Our results showed that the bilateral middle temporal cortex (MT) was specifically involved in VSTM of actions because its activation and its functional connectivity with the frontal–parietal network (FPN) were only modulated by the memory load of actions, but not by that of objects/agents or locations. Moreover, the brain regions involved in the maintenance of spatial location information (i.e., superior parietal lobule, SPL) was also recruited during the maintenance of actions, consistent with the temporal–spatial nature of actions. Meanwhile, the frontoparietal network (FPN) was commonly involved in all types of VSTM and showed flexible functional connectivity with the domain‐specific regions, depending on the current working memory tasks. Together, our results provide clear evidence for a distinct neural system for maintaining actions in VSTM, which supports the core knowledge system theory and the domain‐specific and domain‐general architectures of VSTM.
收起
原文获取