AToT 模型可视化工具开发

随着科研人员需要处理的文献集规模的日益庞大,以LDA 为代表的主题模型能够从语义层面挖掘大规模文献集中隐含的主题,因此,LDA 主题模型的应用越来越广泛。LDA 模型仅仅关注文献集的内容,而忽略了文献其他重要的外部信息,AToT 模型在LDA 主题模型的基础上引入了文献作者和文献发表时间两个属性,使AToT 模型不仅可以挖掘文献中隐含的信息,还可以分析文献作者的研究兴趣及文献主题随时间的变化。AToT 模型对文献集建模的结果是以概率矩阵的形式呈现,不能直观、全面、清晰的呈现挖掘出来的信息,特别是对数据挖掘不熟悉的科研人员,因此,本文开发了一个基于AToT 模型的可视化系统,该可视化系统清晰、美观地展现了AToT 模型中文献、主题、作者、时间、词项间的关系。如文档中的主题分布、主题的词项分布、作者的研究兴趣分布、主题的相似主题和主题的演化趋势等。     

基于主题发现的专利发明人推荐方法

LDA 主题模型可用于识别大规模文档集中潜藏的主题信息,本文提出了一种基于LDA 建立发明人兴趣主题模型的方法,合并每位发明人的专利数据,专利信息基于发明人进行划分,将标准的文档- 主题-词的三层LDA 模型变为专利数据中的发明人- 主题- 词的发明人兴趣模型,实现发明人的主题发现,并利用该模型中主题分布之间的相似性进行发明人的个性化推荐。在采集真实专利数据集上的实验结果表明该方法相比传统的向量空间模型方法和隐马尔科夫模型方法具有更高的准确率,推荐效果更优。     

基于LDA模型的医学领域主题分裂融合探测

[目的/意义]随着信息资源在数量和种类上的急剧增长,学科间的交叉融合不断涌现,快速主动地从海量信息资源中识别和判断研究主题的发展演化是实现科技创新的基础。[方法/过程]在相关理论调研的基础上,结合医学领域的资源特点,提出一种基于LDA模型的主题演化探测模型和相应的流程步骤。主要步骤包括医学主题词抽取、主题识别、主题关联、关键主题识别、关键主题的演化主路径识别、演化主路径上主题分裂、融合事件识别,实现深度、细致的主题演化分析。[结果/结论]选用乳腺癌治疗研究文献为实验案例,对判断模型进行试验并对结果进行分析验证,证实提出的技术方法具有一定的可靠性。     

专业社交媒体中的主题图谱构建方法研究——以汽车论坛为例

专业社交媒体中主题图谱的内容包括论坛中的主题及主题之间的关系,其具有挖掘专业产品创新方向、构建专业知识索引等重要应用价值。本文基于深度学习技术与文本挖掘技术,提出了专业社交媒体中的主题图谱构建方法。首先,使用专业社交媒体中的文本训练Skip-Gram模型,利用该模型的隐藏层权重与模型输出的预测结果,分别获取词语间的语义相似度与上下文关联度。其次,基于该语义相似度与上下文关联度,对已有领域种子本体词汇进行扩充,将语义相似或上下文相邻近的词汇纳入本体词汇,为主题抽取提供高质量的领域词汇。然后,基于扩充的专业本体词汇,使用结合本体词汇的LDA主题模型从专业社交媒体文本中抽取主题与主题词。最后,利用语义相似度与上下文关联度,定义关联度权重,通过图模型与谱聚类,获取主题间与主题词的关联关系与层次结构。本文使用汽车论坛语料进行主题图谱生成实验。实验结果表明,本文方法获取的主题词纯净度相比单独使用LDA模型提升了20.2%,且能够清晰合理地展现主题之间的关系。     

基于动态LDA主题模型的内容主题挖掘与演化

指出文本内容主题的挖掘和演化研究对于文本建模和分类及推荐效果提升具有重要作用。从分析基于LDA主题模型的文本内容主题挖掘原理入手,针对当前网络环境下的文本内容特点,构建适用于动态文内容本主题挖掘的LDA模型,并通过改进的Gibbs抽样估计提高主题挖掘的准确性,进而从主题相似度和强度两个方面研究内容主题随时间的演化问题。实验表明,所提方法可行且有效,对后续有关文本语义建模和分类研究等具有重要的实践意义。     

基于关键词关联度指标(KRI)进行LDA噪声主题过滤的方法研究

[目的/意义] 针对LDA模型主题识别结果通常包含噪声主题的问题,建立科学有效的主题过滤方法,排除噪声主题,确保主题识别及后续演化分析的准确性。[方法/过程] 基于关键词之间的共现关系,构建关键词关联度指标（KRI）,借助定量手段进行主题筛选和过滤。以单细胞研究领域为例,计算各主题-关键词分布的KRI值,与人工判读结果进行对比分析。[结果/结论] 实验结果表明,该方法能够有效排除LDA模型识别结果中的噪声主题,提高主题识别的准确性,也在一定程度上降低了主题识别过程对人工判读的依赖性。     

���ڹؼ��ʹ�����ָ��(KRI)����LDA����������˵ķ����о�

��[Purpose/significance] The identification results of the LDA model is sometimes unsatisfactory due to some meaningless topics mixed together. Therefore, it's quite necessary to establish an effective topic filtering method to eliminate these noise topics and to ensure the accuracy of subsequent evolution analysis.[Method/process] Based on the co-occurrence relationship between keywords, keywords relevance index (KRI) was constructed. Taking the field of single cell research as an example, KRI values of the distribution of theme-keywords were calculated and compared with the results of manual interpretation.[Result/conclusion] Experimental results show that this method can effectively eliminate meaningless noise topics in the LDA model recognition results, which can improve the accuracy of topic recognition and the subsequent topic evolution analysis. It also helps to reduce the dependence on manual interpretation in the process of topic identification through the topic model method.     

多源数据环境下科研热点识别方法研究

[目的/意义] 在科学研究中,从不同来源的科技文献中识别挖掘科研热点对于开展科研工作具有指导意义。旨在通过本研究提出的模型方法,快速准确地识别蕴含在多源文本中的热点主题,为科研创新提供支撑服务。[方法/过程] 提出一种基于LDA2vec模型的多源文本下科研热点识别的方法并针对科研热点识别构建模型,该方法融合LDA主题模型对隐含语义挖掘的优势和Word2Vec词向量模型对于上下文关系把握的优势。以机器学习领域的科技文献为例,利用模型困惑度和主题一致性两个指标对LDA2vec的在本领域应用的可行性和有效性进行验证,并与LDA的主题提取效果进行对比。[结果/结论] 实验结果表明,提出的方法在面对多源数据情况下,进行科研热点识别挖掘是可行的,且在一定程度上有效果的提升,对利用单一数据源进行主题分析的不足进行补充,对多数据源融合的实践应用进行丰富。     

��Դ���ݻ����¿����ȵ�ʶ�𷽷��о�

��[Purpose/significance] In scientific research, identifying mining scientific research hotspots from different sources of scientific literature is of guiding significance for carrying out the next scientific research work. It aims to quickly and accurately identify hot topics contained in multi-source texts through the model method proposed in this study, and provide support services for scientific research innovation.[Method/process] This paper proposed a method based on LDA2vec model for multi-source text research hotspot identification and built a model for scientific research hotspot identification. This method combined the advantages of LDA topic model on implicit semantic mining and the context of Word2Vec word vector model. Taking the scientific literature in the field of machine learning as an example, the model extraction degree (perplexity) and topic coherence (topic coherence) were used to compare the topic extraction effects of LDA2vec and LDA in the context of multi-source text.[Result/conclusion] After experiments, the results show that the method proposed in this paper is feasible and can be improved to some extent in the face of multi-source data. The method can relatively quickly and accurately identify the hot content in the multi-data source text, make up for the shortcoming of the single analysis data source for subject detection, and enrich the practical application of the multi-data source fusion theory system.     

A non-parametric topical relevance model

An information retrieval (IR) system can often fail to retrieve relevant documents due to the incomplete specification of information need in the user’s query. Pseudo-relevance feedback (PRF) aims to improve IR effectiveness by exploiting potentially relevant aspects of the information need present in the documents retrieved in an initial search. Standard PRF approaches utilize the information contained in these top ranked documents from the initial search with the assumption that documents as a whole are relevant to the information need. However, in practice, documents are often multi-topical where only a portion of the documents may be relevant to the query. In this situation, exploitation of the topical composition of the top ranked documents, estimated with statistical topic modeling based approaches, can potentially be a useful cue to improve PRF effectiveness. The key idea behind our PRF method is to use the term-topic and the document-topic distributions obtained from topic modeling over the set of top ranked documents to re-rank the initially retrieved documents. The objective is to improve the ranks of documents that are primarily composed of the relevant topics expressed in the information need of the query. Our RF model can further be improved by making use of non-parametric topic modeling, where the number of topics can grow according to the document contents, thus giving the RF model the capability to adjust the number of topics based on the content of the top ranked documents. We empirically validate our topic model based RF approach on two document collections of diverse length and topical composition characteristics: (1) ad-hoc retrieval using the TREC 6-8 and the TREC Robust ’04 dataset, and (2) tweet retrieval using the TREC Microblog ’11 dataset. Results indicate that our proposed approach increases MAP by up to 9% in comparison to the results obtained with an LDA based language model (for initial retrieval) coupled with the relevance model (for feedback). Moreover, the non-parametric version of our proposed approach is shown to be more effective than its parametric counterpart due to its advantage of adapting the number of topics, improving results by up to 5.6% of MAP compared to the parametric version.     

Research frontier detection and analysis based on research grants information: A case study on health informatics in the US

Identifying research fronts is an essential aspect of promoting scientific development. Many researchers choose their research directions and topics by analyzing their field's current research fronts. Many previous researchers have used academic papers or patents to identify research fronts; however, this is potentially outdated and reduces the prospective value of the research front detection. Considering this, this work proposes adapted indicators to conduct research front topic detection based on research grant data, which aims to identify research front topics and forecast trends using path analysis. First, research topics were identified using topic modeling, and then the mapping relations from topics to both fund projects and cross-domain categories were built. Then, research front topics were detected by multi-dimensional measurements, and the evolution of research topics was analyzed using topic evolution visualization to predict development trends. Finally, the Brillouin index was used to measure the cross-domain degree. Our method was evaluated using a dataset from the field of health informatics and was shown to be effective in research front identification. We found that the proposed adapted indicators were informative in identifying the evolutional trends in the health informatics field. In addition, research grants with higher cross-domain degrees are more likely to receive a high amount of funding.     

基于Chunk-LDAvis的核心技术主题识别方法研究

[目的/意义]基于大量专利文献数据的核心技术主题识别有助于识别某技术领域的关键技术、分析关键技术的发展方向,是进行技术创新的基础情报工作,对于研究人员、企业乃至国家层面都具有一定的意义。[方法/过程]提出基于Chunk-LDAvis的核心技术主题识别方法,首先基于经典LDA模型进行主题识别,然后利用名词组块对初始LDA主题识别结果进行标注,构建Chunk-LDA主题识别结果,提高其可解读性;然后基于社会网络分析方法构建主题网络,识别核心技术主题;基于R语言的LDAvis工具包绘制可交互的Chunk-LDAvis核心技术主题关联分析图谱,发现核心技术主题的隐含联系,辅助进行核心技术主题识别。[结果/结论]通过对纳米农业领域进行实证研究,验证了本文提出方法的准确性和可行性。     

Topics in dynamic research communities: An exploratory study for the field of information retrieval

Research topics and research communities are not disconnected from each other: communities and topics are interwoven and co-evolving. Yet, scientometric evaluations of topics and communities have been conducted independently and synchronically, with researchers often relying on homogeneous unit of analysis, such as authors, journals, institutions, or topics. Therefore, new methods are warranted that examine the dynamic relationship between topics and communities. This paper examines how research topics are mixed and matched in evolving research communities by using a hybrid approach which integrates both topic identification and community detection techniques. Using a data set on information retrieval (IR) publications, two layers of enriched information are constructed and contrasted: one is the communities detected through the topology of coauthorship network and the other is the topics of the communities detected through the topic model. We find evidence to support the assumption that IR communities and topics are interwoven and co-evolving, and topics can be used to understand the dynamics of community structures. We recommend the use of the hybrid approach to study the dynamic interactions of topics and communities.     

�ڿ����ĺ����о�����ʶ�����ݻ�·�����ӻ������о�——���ҹ�ҽ�ƽ�����Ϣ�����ڿ�����Ϊ��

��[Purpose/significance] This paper proposes the identification of the core research topics and their evolution path visualization methods, in order to provide reference for the field subject evolution analysis research, which has certain significance for revealing the evolution characteristics and development laws of the core topics.[Method/process] Using the LDA model for topic recognition and combining multi-dimensional scaling analysis and visualization techniques to map LDA topic recognition results to two-dimensional space. The topic similarity algorithm was used to detect the association between adjacent time topics, a new visual display method was proposed. We constructed cross-evolution paths of different types of research topics to reveal the dynamic changes of core topics and secondary topics in the evolution process.[Result/conclusion] Taking the medical health information field in China as an example, the research results show that the core research topics in the field of medical and health information in China mainly include electronic health records and Internet medical treatment. Among them, core themes such as health management and smart medical treatment show a good development trend.     

中外图书馆出版服务研究主题比较分析

[目的/意义]图书馆出版服务受到广泛的关注且不断地发展,国内外学者纷纷开始研究其对旧图书出版服务和单一的图书馆场所服务模式的挑战及影响。通过对国内外学者研究方向和主题进行归纳总结,对比分析中外学者的关注热点,旨在引进思路并为我国图书馆出版服务提供借鉴。[方法/过程]采用LDA文档主题提取方法,并在提取出来的数据中绘制主题分布热点图。[结果/结论]通过对比国内外研究热点,发现由于国外图书馆较早开展出版服务,很多高校会根据自己的情况来制定出版服务的模式,因此国外学者的研究更多集中在分析图书馆的角色定位、线上服务、数字化趋势和战略性创新等方面;而国内学者主要是通过借鉴国外案例,并结合国内的实际情况,如经费问题、体制问题等,指出我国当前图书馆出版服务中所面临的挑战和新机遇,同时提出符合国内发展状况的模式和内外结合的创新性思路。     

Can epidemic models describe the diffusion of topics across disciplines?

This paper introduces a new approach to describe the spread of research topics across disciplines using epidemic models. The approach is based on applying individual-based models from mathematical epidemiology to the diffusion of a research topic over a contact network that represents knowledge flows over the map of science—as obtained from citations between ISI Subject Categories. Using research publications on the protein class kinesin as a case study, we report a better fit between model and empirical data when using the citation-based contact network. Incubation periods on the order of 4–15.5 years support the view that, whilst research topics may grow very quickly, they face difficulties to overcome disciplinary boundaries.     

Communication Strategies for Instant Messaging and Chat Reference Services

Summary

Instant messaging (IM) reference is gaining in popularity but still faces resistance. Some librarians agree with some researchers in the field of computer-mediated communications (CMC) that it can never approach the complexity of face-to-face communication, and is therefore an unsuitable medium for reference. Librarians in face-to-face reference use nonverbal communication skills such as a welcoming expression and an interested tone of voice to encourage patrons to approach the desk and discuss their topic; they also interpret the nonverbal cues of patrons. This analysis of online reference conversations shows how online skills can substitute for many of these nonverbal cues. Some skills are unique to computer-mediated communication while others involve written language skills to encourage exploration of the topic, increase clarity, demonstrate approachability and empathy, and instruct. The study illustrates communication problems and solutions using actual conversations, giving particular attention to the reference interview.     

微博舆情传播周期中不同传播者的主题挖掘与观点识别

[目的/意义] 探索微博舆情传播周期中不同传播者关注的舆情热点和传播内容的主要观点,进而发现舆情传播的特点和规律,为舆情分析与决策提供依据。[方法/过程] 以特定舆情事件的事实文本数据为来源,以生命周期理论和LDA方法为指导,设计研究流程与构建研究模型,对微博舆情事件中不同传播者的话题进行主题研究,其中包括主题抽取和结果语义标注、各阶段的不同传播者主题的语义分析、基于时间维度的舆情主题观点识别与刻画。[结果/结论] 研究发现,论文所提出的研究模型能够挖掘出舆情传播周期中不同传播者的主题结构、观点脉络以及特征,研判出分布在文字当中有关联性的、代表性的、重要的词语。同时,结论中还发现微博中的官媒、大众媒体发布信息中的话题和用户谈论的热点话题具有明显的差异性。     

Exploring scientific trajectories of a large-scale dataset using topic-integrated path extraction

Main path analysis (MPA) is the most widely accepted approach to tracing knowledge transfer in a research field. In this study, we extracted multiple longest paths from the multidisciplinary academic field's citation network and integrating topic modeling to the extracted paths. We consider three main aspects of trajectory analysis when analyzing the represented documents through the extracted paths: emergence, authority, and topic dynamics. For path extraction, we adopt the longest path algorithm that consists of the following three steps: 1) topological sort, 2) edge relaxation, and 3) multiple path extraction. For topic integration into multiple paths, we employ latent Dirichlet allocation (LDA) by utilizing the topic-document matrix that LDA derives to select an article's topic from the citation network, where each article is labeled with the topic that is assigned with the highest topical probability for that article. We conduct a series of experiments to examine the results on a dataset from the field of healthcare informatics that PubMed provides.