Scholarlink

<p>Emerging research topics represent the forefront of scientific discoveries and technological breakthroughs, holding significant potential to drive innovation across industries. Identifying and recognizing these topics at an early stage facilitates proactive planning and better preparedness for future technological transformations. Traditional topic modeling methods often preprocess textual data in a simplistic manner, which inevitably incorporates irrelevant textual information, such as research background, significance, and limitations, that may deviate from the core content. This inclusion of “noisy” text results in suboptimal topic quality. Moreover, existing studies frequently evaluate topics based on factors such as the number of related publications, publication time, and citation counts, leading to a biased perception that topics with a larger volume of associated literature are automatically considered emerging. To address these limitations, this study proposes a semantic filtering approach to enhance the quality of topic identification. Furthermore, it introduces a set of indicators grounded in the theory of knowledge element metrics: knowledge novelty, knowledge growth, and knowledge impact. By employing a three-dimensional strategic coordinate system, this method enables the precise identification of emerging research topics. An empirical study in the field of brain-like intelligence shows that: (1) Semantic filtering significantly improves topic modeling performance. For instance, the topic coherence of text filtered using the DeepSeek method reached 0.42422, with a diversity score of 0.73064, and the number of noise documents was reduced from 5469 to 5066. DeepSeek outperformed other semantic filtering methods in maintaining topic coherence and reducing textual noise. (2) The knowledge element-based evaluation system allows for finer-grained assessments of emerging research topics. It effectively identifies promising research directions, providing a more nuanced understanding of emerging trends. These results underscore the potential of the proposed method in accurately identifying and evaluating emerging research topics, offering valuable insights for strategic decision-making in research and development.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Semantic filtering meets knowledge elements: a novel approach for emerging research topics discovery

  • 期刊论文

  • 发表日期：

    2025年12月18日

查看全文
  • Biao Zhang,
  • Yunwei Chen

摘要

Emerging research topics represent the forefront of scientific discoveries and technological breakthroughs, holding significant potential to drive innovation across industries. Identifying and recognizing these topics at an early stage facilitates proactive planning and better preparedness for future technological transformations. Traditional topic modeling methods often preprocess textual data in a simplistic manner, which inevitably incorporates irrelevant textual information, such as research background, significance, and limitations, that may deviate from the core content. This inclusion of “noisy” text results in suboptimal topic quality. Moreover, existing studies frequently evaluate topics based on factors such as the number of related publications, publication time, and citation counts, leading to a biased perception that topics with a larger volume of associated literature are automatically considered emerging. To address these limitations, this study proposes a semantic filtering approach to enhance the quality of topic identification. Furthermore, it introduces a set of indicators grounded in the theory of knowledge element metrics: knowledge novelty, knowledge growth, and knowledge impact. By employing a three-dimensional strategic coordinate system, this method enables the precise identification of emerging research topics. An empirical study in the field of brain-like intelligence shows that: (1) Semantic filtering significantly improves topic modeling performance. For instance, the topic coherence of text filtered using the DeepSeek method reached 0.42422, with a diversity score of 0.73064, and the number of noise documents was reduced from 5469 to 5066. DeepSeek outperformed other semantic filtering methods in maintaining topic coherence and reducing textual noise. (2) The knowledge element-based evaluation system allows for finer-grained assessments of emerging research topics. It effectively identifies promising research directions, providing a more nuanced understanding of emerging trends. These results underscore the potential of the proposed method in accurately identifying and evaluating emerging research topics, offering valuable insights for strategic decision-making in research and development.