基于社交圈层和注意力机制的信息热度预测

社交网络现已成为现实世界中信息传播与扩散的主要媒介,对其中的热点信息进行建模和预测有着广泛的应用场景和商业价值,比如进行信息传播挖掘、广告推荐和用户行为分析等.目前的相关研究主要利用特征和时间序列进行建模,但是并没有考虑到社交网络中用户的社交圈层对于信息传播的作用.本文提出了一种基于社交圈层和注意力机制的热度预测模型SCAP(Social Circle and Attention based Popularity Prediction),首先对社交圈层进行定义,通过自动编码器提取用户历史文本序列的特征,对不同用户的社交圈层进行聚类划分,得到社交圈层特征.进而对于一条新发布的文本信息,通过长短期记忆网络与嵌入层提取其文本特征、用户特征和时序特征,并基于注意力机制,捕获到不同社交圈层对于该文本信息的影响程度,得到社交圈层注意力特征.最后将文本特征、用户特征、时序特征和社交圈层注意力特征进行特征融合,并通过两个全连接层进行建模学习,对社交信息的热度进行预测.在推特、微博和豆瓣等四个数据集上的实验结果表明,SCAP模型的预测表现相比于多个对比模型总体呈优,在不同数据集上均方误差(MSE)分别降低了0.017,0.022,0.021和0.031,F1分数分别提升0.034,0.021,0.034和0.025,能够较为准确地预测社交信息的热度.本文同时探究了不同实验参数对于模型的影响效果,如用户历史文本序列的数量、社交圈层的数量和时间序列的长度,最后验证了模型输入的各个特征和注意力机制的引入对于模型预测性能提升的有效性,在推特数据集中,引入社交圈层和注意力机制,模型的MSE指标分别降低了0.065和0.019.     

融合传播影响力的热点事件时序摘要研究

技术获得了一定程度的发展,但仍具有较大局限性,主要表现为缺乏对事件发展演化阶段的追踪挖掘,以及对社交媒体文本特性的挖掘利用不够充分,忽视了社交文本所蕴含的时序信息及传播影响力与文本概述信息能力之间的联系等问题.     

基于社交媒体文本信息的金融时序预测

针对传统股票趋势预测模型中忽略社交媒体文本信息对股价变化的影响和时间序列的平稳性处理、长期依赖等问题,提出一种融合社交媒体文本信息和LSTM的股票趋势预测模型(BiTCN-LSTM).该模型分为情感分析和金融时序预测两部分.情感分析层将社交媒体文本信息输入到双向时间卷积网络进行特征提取和情感分析,得到积极或者消极的情感分类表示;金融时序预测层使用LSTM神经网络,将差分运算后的股票历史数据和文本情感特征向量加权融合作为网络输入,完成金融时序预测任务.通过上海证券综合指数数据集的实验验证,与传统金融时序预测模型相比,该模型的RMSE指标降低3.44-43.62.     

时序感知的异质图神经谣言检测

近年来,在线社交媒体的发展大大加速了谣言的滋生和传播,谣言的危害性使得谣言的自动检测技术受到研究学者的广泛关注.本文同时考虑事件与事件之间的全局结构关系以及事件内部消息传播的时序关系,以异质图为载体共同显式建模两种关系,提出一种新的时序感知的异质图神经谣言检测模型.该模型利用时序感知的自注意力机制捕获事件内部转发(或评论)贴之间的时序关系,并将具有时序信息的转发(或评论)贴与源贴融合,得到事件的局部时序表征;接着利用元素级注意力机制捕捉事件与事件之间的全局结构关系,学习事件的全局结构表征;最后将二者融合用于检测谣言.实验结果表明,该模型优于大多数现有模型,可以提高谣言检测性能,并且同样具有优秀的早期检测性能.     

融合上下文信息的篇章级事件时序关系抽取方法

事件时序关系抽取是一项重要的自然语言理解任务,可以广泛应用于诸如知识图谱构建、问答系统等任务.已有事件时序关系抽取方法往往将该任务视为句子级事件对的分类问题,而基于有限的局部句子信息导致其抽取的事件时序关系的精度较低,且无法保证整体时序关系的全局一致性.针对此问题,提出一种融合上下文信息的篇章级事件时序关系抽取方法,使用基于双向长短期记忆(bidirectional long short-term memory,Bi-LSTM)的神经网络模型学习文章中事件对的时序关系表示,再利用自注意力机制融入上下文中其他事件对信息,从而得到更丰富的事件对时序关系表示用于时序关系分类通过 TB-Dense(timebank dense)和 M ATRES(multi-axis temporal relations for start-points)数据集的实验表明:此方法能够取得比当前主流的句子级方法更佳的抽取效果.     

基于级联时空特征的信息传播预测方法

现有信息传播预测方法对级联序列和拓扑结构独立建模,难以学习级联时序特征和结构特征在嵌入空间的交互表达,造成对信息传播动态演化的刻画不足.因此,文中提出基于级联时空特征的信息传播预测方法.基于社交关系网络和传播路径构建异质图,使用图神经网络学习异质图和社交关系网络节点的结构上下文,引入门控循环单元提取级联时序特征,融合结构上下文和时序特征,构建级联时空特征,进行信息传播的微观预测.在Twitter、Memes数据集上的实验表明,文中方法性能得到一定提升.     

面向俄乌冲突的时序知识图谱推理系统设计与实现

提出了一种面向俄乌冲突的时序知识图谱推理方法,该方法基于进化图演变的思路,将历史事件和相关知识构建成一个时序知识图谱,并通过对知识图谱中节点和边的演化过程进行建模,实现了对未来事件的预测和分析。具体而言,首先利用自然语言处理技术对公开的新闻报道、社交媒体信息等进行爬取和处理,提取关键词和时间信息,并将其转换为结构化数据。然后,利用这些数据构建起一个时序知识图谱,其中每个节点表示一个事件或概念,每条边表示它们之间的关系。其次,通过基于进化图演变的算法对知识图谱建模,从而能够模拟节点和边在不同时间段内的演化特征。最后,通过在俄乌冲突数据集上构建推理系统,实验结果表明该模型与系统设计具有良好的效果和应用前景。     

基于时序感知DAG的多模态对话情绪识别模型

针对现有对话情绪识别方法中对时序信息、话语者信息、多模态信息利用不充分的问题,提出了一个时序信息感知的多模态有向无环图模型（MTDAG）。其中所设计的时序感知单元能按照时间顺序优化话语权重设置,并收集历史情绪线索,实现基于近因效应下对时序信息和历史信息更有效的利用;设计的上下文和话语者信息融合模块,通过提取上下文语境和话语者自语境的深度联合信息实现对话语者信息的充分利用;通过设置DAG（directed acyclic graph）子图捕获多模态信息并约束交互方向的方式,在减少噪声引入的基础上充分利用多模态信息。在两个基准数据集IEMOCAP和MELD的大量实验表明该模型具有较好的情绪识别效果。     

基于用户活动轨迹和个性化区域划分的兴趣点推荐

随着基于位置的社交网络(Location-Based Social Networks,LBSNs)与兴趣点(Point of Interest,POI)推荐的有效组合,近年来已涌现出大量的相关研究,这些方法主要可分为将地理、社会、类别、文本以及时间等上下文信息进行建模并融合,进而克服数据稀疏问题并提升兴趣点推荐的性能.但已有的兴趣点推荐方法认为不同上下文间相互独立,在对不同上下文建模并融合的过程中忽略了其内在联系,导致上下文信息未得到充分利用.另外,在将上下文模型融合到用户自身偏好模型时,未考虑上下文信息对用户历史签到记录的不同影响.为应对上述挑战,本文合理地重构了上下文信息模型并有效地融合到用户偏好模型中,且提出了一种基于用户活动轨迹和个性化区域划分的兴趣点推荐方法.该方法根据用户的活动轨迹刻画出其日常活动区域,并探索了不同用户间的地理距离分布以及活动轨迹的相似度以建模社会关系对用户签到的影响.进一步地,结合用户活动轨迹区域内的POI的地理信息,使用带有自适应带宽的核密度估计方法评估POI间的地理相关性,以建模POI地理信息对用户签到的影响.最后,将用户社会关系模型和POI地理信息模...     

融合依存关系和篇章修辞关系的事件时序关系识别

已有事件间时序关系识别只考虑两个事件所在上下文的局部信息,忽略事件间篇章视角的关联关系.针对这一问题,文中给出融合句子级依存关系和篇章层修辞关系的事件时序关系识别方法.将事件间关联关系分两部分进行表征:事件所在句子的依存路径信息和事件所在基本篇章单元间的修辞关系信息.基于这一表征体系构建可以捕获更多有效信息的神经网络模型,提高事件时序关系识别的性能.在TimeBank-Dense语料上的一系列实验验证文中方法的优越性.     

Twitter summarization with social-temporal context

Twitter is one of the most popular social media platforms for online users to create and share information. Tweets are short, informal, and large-scale, which makes it difficult for online users to find reliable and useful information, arising the problem of Twitter summarization. On the one hand, tweets are short and highly unstructured, which makes traditional document summarization methods difficult to handle Twitter data. On the other hand, Twitter provides rich social-temporal context beyond texts, bringing about new opportunities. In this paper, we investigate how to exploit social-temporal context for Twitter summarization. In particular, we provide a methodology to model temporal context globally and locally, and propose a novel unsupervised summarization framework with social-temporal context for Twitter data. To assess the proposed framework, we manually label a real-world Twitter dataset. Experimental results from the dataset demonstrate the importance of social-temporal context in Twitter summarization.     

A novel approach to update summarization using evolutionary manifold-ranking and spectral clustering

Update summarization is a new challenge in automatic text summarization. Different from the traditional static summarization, it deals with the dynamically evolving document collections of a single topic changing over time, which aims to incrementally deliver salient and novel information to a user who has already read the previous documents. How to have a content selection and linguistic quality control in a temporal context are the two new challenges brought by update summarization. In this paper, we address a novel content selection framework based on evolutionary manifold-ranking and normalized spectral clustering. The proposed evolutionary manifold-ranking aims to capture the temporal characteristics and relay propagation of information in dynamic data stream and user need. This approach tries to keep the summary content to be important, novel and relevant to the topic. Incorporation with normalized spectral clustering is to make summary content have a high coverage for each sub-topic. Ordering sub-topics and selecting sentences are dependent on the rank score from evolutionary manifold-ranking and the proposed redundancy removal strategy with exponent decay. The evaluation results on the update summarization task of Text Analysis Conference (TAC) 2008 demonstrate that our proposed approach is competitive. In the 71 run systems, we receive three top 1 under PYRAMID metrics, ranking 13th in ROUGE-2, 15th in ROUGE-SU4 and 21st in BE.     

Time Aware Knowledge Extraction for microblog summarization on Twitter

Microblogging services like Twitter and Facebook collect millions of user generated content every moment about trending news, occurring events, and so on. Nevertheless, it is really a nightmare to find information of interest through the huge amount of available posts that are often noisy and redundant. In the era of Big Data, social media analytics services have caught increasing attention from both research and industry. Specifically, the dynamic context of microblogging requires to manage not only meaning of information but also the evolution of knowledge over the timeline. This work defines Time Aware Knowledge Extraction (briefly TAKE) methodology that relies on temporal extension of Fuzzy Formal Concept Analysis. In particular, a microblog summarization algorithm has been defined filtering the concepts organized by TAKE in a time-dependent hierarchy. The algorithm addresses topic-based summarization on Twitter. Besides considering the timing of the concepts, another distinguishing feature of the proposed microblog summarization framework is the possibility to have more or less detailed summary, according to the user’s needs, with good levels of quality and completeness as highlighted in the experimental results.     

One-class learning and concept summarization for data streams

In this paper, we formulate a new research problem of concept learning and summarization for one-class data streams. The main objectives are to (1) allow users to label instance groups, instead of single instances, as positive samples for learning, and (2) summarize concepts labeled by users over the whole stream. The employment of the batch-labeling raises serious issues for stream-oriented concept learning and summarization, because a labeled instance group may contain non-positive samples and users may change their labeling interests at any time. As a result, so the positive samples labeled by users, over the whole stream, may be inconsistent and contain multiple concepts. To resolve these issues, we propose a one-class learning and summarization (OCLS) framework with two major components. In the first component, we propose a vague one-class learning (VOCL) module for concept learning from data streams using an ensemble of classifiers with instance level and classifier level weighting strategies. In the second component, we propose a one-class concept summarization (OCCS) module that uses clustering techniques and a Markov model to summarize concepts labeled by users, with only one scanning of the stream data. Experimental results on synthetic and real-world data streams demonstrate that the proposed VOCL module outperforms its peers for learning concepts from vaguely labeled stream data. The OCCS module is also able to rebuild a high-level summary for concepts marked by users over the stream.     

基于条件随机场的多时相遥感影像分类*

多时相遥感影像分类方法通常使用人工设置的转移矩阵作为时间上下文信息,这样不仅难以获得准确的转移矩阵,而且没有充分利用时间上下文信息。针对多时相遥感图像中的时间与空间上下文信息难以构建的问题,提出了一种基于条件随机场模型的多时遥感影像分类方法。首先运用最大期望算法生成用于描述时间上下文信息的时间势能,然后结合空间以及时间上下文信息构造了条件随机场模型,最后使用该模型对多时相遥感影像进行分类。一系列的实验结果表明,该方法可以有效提高遥感影像的分类精度。     

Model‐Based Visualization of Temporal Abstractions

We describe a new conceptual methodology and related computational architecture called Knowledge‐based Navigation of Abstractions for Visualization and Explanation (KNAVE). KNAVE is a domain‐independent framework specific to the task of interpretation, summarization, visualization, explanation, and interactive exploration, in a context‐sensitive manner, of time‐oriented raw data and the multiple levels of higher level, interval‐based concepts that can be abstracted from these data. The KNAVE domain‐independent exploration operators are based on the relations defined in the knowledge‐based temporal‐abstraction problem‐solving method, which is used to abstract the data, and thus can directly use the domain‐specific knowledge base on which that method relies. Thus, the domain‐specific semantics are driving the domain‐independent visualization and exploration processes, and the data are viewed through a filter of domain‐specific knowledge. By accessing the domain‐specific temporal‐abstraction knowledge base and the domain‐specific time‐oriented database, the KNAVE modules enable users to query for domain‐specific temporal abstractions and to change the focus of the visualization, thus reusing for a different task (visualization and exploration) the same domain model acquired for abstraction purposes. We focus here on the methodology, but also describe a preliminary evaluation of the KNAVE prototype in a medical domain. Our experiment incorporated seven users, a large medical patient record, and three complex temporal queries, typical of guideline‐based care, that the users were required to answer and/or explore. The results of the preliminary experiment have been encouraging. The new methodology has potentially broad implications for planning, monitoring, explaining, and interactive data mining of time‐oriented data.