动态文本会话抽取技术研究

对海量的短文数据进行自动分析和挖掘,从中获取有价值的隐含知识已经成为一项迫切的需求。动态文本会话抽取是针对MSN,QQ等动态数据,将属于同一会话但相互交错的不同消息文本组织在一起,属于在线话题发现追踪的新兴领域,在信息检索,文本挖掘和话题检测追踪等方面有着重要应用。首先介绍了文本会话抽取的必要性和重要性,介绍其主要研究内容和结果评测方法;然后对其中多个研究内容提出一个统一研究框架,并对该框架中的关键技术进行了详细分析;最后指出该领域中的关键问题及难点,并对未来研究做出展望。     

Identification of key cyberbullies: A text mining and social network analysis approach

Cyberbullying is a major problem in society, and the damage it causes is becoming increasingly significant. Previous studies on cyberbullying focused on detecting and classifying malicious comments. However, our study focuses on a substantive alternative to block malicious comments via identifying key offenders through the application of methods of text mining and social network analysis (SNA). Thus, we propose a practical method of identifying social network users who make high rates of insulting comments and analyzing their resultant influence on the community. We select the Korean online community of Daum Agora to validate our proposed method. We collect over 650,000 posts and comments via web crawling. By applying a text mining method, we calculate the Losada ratio, a ratio of positive-to-negative comments. We then propose a cyberbullying index and calculate it based on text mining. By applying the SNA method, we analyze relationships among users so as to ascertain the influence that the core users have on the community. We validate the proposed method of identifying key cyberbullies through a real-world application and evaluations. The proposed method has implications for managing online communities and reducing cyberbullying.     

Correlation between university students’ online trolling behavior and online trolling victimization forms,current conditions,and personality traits

Few studies have specifically examined online trolling behavior and the forms and current conditions of online trolling victimization that may develop among university students, as well as the correlation of these to personality traits. The valid sample included 285 university students, with 80.6% and 19.3% being male and female, respectively. The research results are: 1. Online trolling behavior is more common in those who more frequently post text information on Facebook than those who do not; 203 (71.2%) and 211 (74.0%) experienced at least 1 instance of online trolling behavior or being an online trolling victim, respectively, in the previous week; 2. University students’ online trolling behavior types are ranked by quantity as evocative trolling, malicious trolling, obstruction trolling, and pathological trolling; 3. University students’ online trolling victimization types are ranked by quantity as identity victimization, dissemination victimization, malicious victimization, and obstruction victimization; 4. Sense of inferiority is a significant predictive variable for online trolling behavior and online trolling victimization. At the same time, social extraversion and depression significantly and positively predict online trolling behavior. Based on the foregoing results, the study proposed discussion and recommendations for university students and future research.     

Truncation noise in fixed-point SFGs [digital filters]

A new model for predicting truncation error variance in fixed-point filter implementations is introduced. The proposed model is shown to be more accurate than existing models, particularly for some direct hardware implementations. In addition, some comments are made on the applicability of existing error models     

IMPACT OF BUSY LINES AND MOBILITY ON CALL BLOCKING IN A PCS NETWORK

Several analytical models have been proposed to study the blocking probability for personal communications service networks or mobile phone networks. These models cannot accurately predict the blocking probability because they do not capture two important features. First, they do not capture the busy-line effect. Even if a cell has free channels, incoming and outgoing calls must be dropped when the destination portable is already in a conversation. Second, they do not capture the mobility of individual portables. In these models, mobility is addressed by net hand-off traffic to a cell, which results in traffic with a smaller variance to a cell compared with the true situation. We propose a new analytic model which addresses both the busy-line effect and individual portable mobility. Furthermore, our model can be used to derive the portable population distribution in a cell. The model is validated against the simulation experiments. We indicate that the previously proposed models approximate a special case of our model where the number of portables in a cell is 40 times larger than the number of channels.     

Sports match prediction model for training and exercise using attention-based LSTM network

Sports matches are very popular all over the world. The prediction of a sports match is helpful to grasp the team's state in time and adjust the strategy in the process of the match. It's a challenging effort to predict a sports match. Therefore, a method is proposed to predict the result of the next match by using teams' historical match data. We combined the Long Short-Term Memory (LSTM) model with the attention mechanism and put forward an AS-LSTM model for predicting match results. Furthermore, to ensure the timeliness of the prediction, we add the time sliding window to make the prediction have better timeliness. Taking the football match as an example, we carried out a case study and proposed the feasibility of this method.     

A new point process transition density model for space-time event prediction

A new point process transition density model is proposed based on the theory of point patterns for predicting the likelihood of occurrence of spatial-temporal random events. The model provides a framework for discovering and incorporating event initiation preferences in terms of clusters of feature values. Components of the proposed model are specified taking into account additional behavioral assumptions such as the "journey to event" and "lingering period to resume act." Various feature selection techniques are presented in conjunction with the proposed model. Extending knowledge discovery into feature space allows for extrapolation beyond spatial or temporal continuity and is shown to be a major advantage of our model over traditional approaches. We examine the proposed model primarily in the context of predicting criminal events in space and time.     

Modeling and Verification of a Sentiment Analysis System Using Aspect-Oriented Petri Nets

An increasing number of social media and networking platforms have been widely used. People usually post the online comments to share their own opinions on the networking platforms with social media. Business companies are increasingly seeking effective ways to mine what people think and feel regarding their products and services. How to correctly understand the online customers’ reviews becomes an important issue. This study aims to propose a method with the aspect-oriented Petri nets (AOPN) to improve the examination correctness without changing any process and program. We collect those comments from the online reviews with Scrapy tools, perform sentiment analysis using SnowNLP, and examine the analysis results to improve the correctness. In this paper, we apply our method for a case of the online movie comments. The experimental results have shown that AOPN is helpful for the sentiment analysis and verifying its correctness.     

Calculating the feature method of short text based on analytic hierarchy process

In order to model the accurate interest preference of microblog users and discover user groups with similar in-terest, a new method was proposed which considered the total amount of retweets, comments and attitudes of each mi-croblog for text feature calculation with utilizing classic analytical hierarchy process method. The proposed method used three indicators to evaluate the importance of the text feature representation and made an improvement on traditional tf-idf feature calculation method to fit for short text. Furthermore, this method was also implemented in the traditional clustering algorithm. Experimental results show that, compared with the traditional tf-idf method, the improved approach has a better clustering effect on the average scattering for clusters and the total separation between clusters.     

Factors influencing users’ post replying behavior in a senior online community: An empirical investigation

Senior online communities (SOCs) have become an important venue for older people to seek support and exchange information. While online community engagement has been well studied in the existing literature, few studies have explored how older adults behave in online communities. Therefore, drawing upon signaling theory, this study aims to investigate how different content-related and social-related signals influence users’ post replying behavior (i.e., reply to another user’s post) in SOCs. We collected 7486 health-related posts and 71,859 comments from one of the most popular Chinese SOCs, Keai (https://www.keai99.com). Information signals in the posts were operationalized using different techniques such as text mining and social network analysis. Results from negative binomial regression indicated that content-related signals (posts’ topic and length) and social-related signals (authors’ position and centrality) were related to replying behavior. In addition, we revealed some differences between the effects of these signals on informational replies and emotional replies. More specifically, compared to posts mentioning traditional Chinese medicine, posts mentioning western medicine received more informational replies, but less emotional replies. Original posts triggered more informational replies, whereas shared posts attracted more emotional replies. Average reply length was positively related to informational replies, but negatively related to emotional replies. Considering the important role of SOCs in satisfying older adults’ social and informational needs, future research is needed to promote user social engagement in SOCs, thereby maintaining their sustainability.     

Multi-layered attentional peephole convolutional LSTM for abstractive text summarization

Abstractive text summarization is a process of making a summary of a given text by paraphrasing the facts of the text while keeping the meaning intact. The manmade summary generation process is laborious and time-consuming. We present here a summary generation model that is based on multilayered attentional peephole convolutional long short-term memory (MAPCoL; LSTM) in order to extract abstractive summaries of large text in an automated manner. We added the concept of attention in a peephole convolutional LSTM to improve the overall quality of a summary by giving weights to important parts of the source text during training. We evaluated the performance with regard to semantic coherence of our MAPCoL model over a popular dataset named CNN/Daily Mail, and found that MAPCoL outperformed other traditional LSTM-based models. We found improvements in the performance of MAPCoL in different internal settings when compared to state-of-the-art models of abstractive text summarization.     

基于电信大数据的5G网络海量用户复访行为预测模型

5G网络中的用户会产生大量的访问数据，导致用户复访行为难以精准预测，因此提出基于电信大数据的5G网络海量用户复访行为预测模型。从电信大数据中提取用户上网历史行为特征数据，构建数据集。引入多阶加权马尔可夫链模型，通过计算各阶自相关系数，得到模型权重值，计算模型的统计量。经过分析后得到各阶步长的马尔可夫氏链一步转移概率矩阵，从而实现对5G网络海量用户复访行为的精准预测。实验结果表明，该模型拥有最低的均值误差和标准差，以及最高的精度、查全率、查准率、F1指标，可证明该方法在预测用户复访行为方面有着非常明显的优势。     

一种热真空环境下的GaAs太阳电池温度预测方法

GaAs太阳电池目前已广泛地应用于空间领域。太阳电池本质上为p-n结结构,性能受到温度的显著影响,故准确地预测空间的实际应用中太阳电池的温度有助于对其进行性能评估。空间用太阳电池的温度在热真空环境与太阳光谱辐射环境的双重作用下达到稳态。以少数载流子连续方程为基础,通过对电池热能与光谱辐射能的传输过程的研究,提出了一种预测GaAs电池温度的迭代求解方法。     

Online blackmail of Czech children focused on so-called “sextortion” (analysis of culprit and victim behaviors)

The online blackmail of children within the environment of Internet services (specifically, social networks) has become an extremely dangerous phenomenon that affects 6–8% of Czech children. Within the following text, data obtained from research conducted in 2013–2015 by the Centre for the Prevention of Risky Virtual Communication, Faculty of Education, Palacký University in Olomouc will be discussed. A selection of analyses aimed at gender and age differences connected with the blackmailing of children will be presented. Additionally, whether the children participate in blackmailing on their own and whether they eventually become offenders of other individuals will be examined.The text will also be supplemented with a typical blackmail model using intimate and sexually explicit materials (which are referred to as “sextortion”) that was created based on a detailed analysis of 25 serious cases of online blackmail. The model has been divided into several connected stages through which the online attack is carried out. The specific stages are subsequently elaborated on using documented evidence. Based on the analysis of individual cases, we have compiled a model that describes the different stages of the process of extortion and can predict communication of the attacker and the victim.     

Reliability prediction models to support conceptual design

During the early stages of conceptual design, the ability to predict reliability is very limited. Without a prototype to test in a lab environment or without field data, component failure rates and system reliability performance are usually unknown. A popular method for early reliability prediction is to develop a computer model for the system. However, most of these models are extremely specific to an individual system or industry. This paper presents three general procedures (using both simulation and analytic solution techniques) for predicting system reliability and average mission cost. The procedures consider both known and unknown failure rates and component-level and subsystem-level analyzes. The estimates are based on the number of series subsystems and redundant (active or stand-by) components for each subsystem. The result is a set of approaches that engineers can use to predict system reliability early in the system-design process. Software was developed (and is discussed in this paper) that facilitates the application of the simulation-based techniques. For the specific type of system and mission addressed in this paper, the analytic approach is superior to the simulation-based prediction models. However, all three approaches are presented for two reasons: (1) to convey the development process involved with building these prediction tools; and (2) the simulation-based approaches are of greater value as the research is extended to consider more complex systems and scenarios     

基于神经网络与马尔可夫组合模型的视频流行度预测算法

为了提升用户体验,降低运营商的成本,将播放最多的视频内容提前放入用户侧缓存是业界的通用做法,如何有效预测视频播放热度已经成为业界热点问题。针对传统预测算法非线性映射能力差、预测精度低及自适应性弱等缺点,提出基于神经网络与马尔可夫组合模型的视频流行度预测算法（Mar-BiLSTM）,该算法通过构建双向长短期记忆（bi-directional long short-term memory,BiLSTM）网络模型可以保留时间序列两个方向的信息依赖;同时在避免引入外部变量导致模型复杂度增加的情况下,利用马尔可夫性质进一步提高了模型的预测精度。实验结果表明,与传统的时间序列和经典的神经网络算法相比,所提算法提升了视频流行度预测的准确性、时效性,并降低了计算量。     

面向匿名通信监控的流量捕获技术研究

匿名通信技术隐藏了通信双方的身份或通信关系,在保障了公民言论自由和隐私权的同时,也不可避免地带来了许多安全隐患,甚至可能被不法分子所利用给国家安全带来极大的危害。文章基于协议特征、数据流特征、匿名代理地址信息这三个匿名会话检测的标准提出了一种基于多代理技术进行匿名通信的系统模型,实现了对匿名会话的有效检测。     

TSL:predicting popularity of Facebook content based on tie strength

The rapid development of online social networks leads to an explosion of information,however,there are great differences in the popularity of different messages,and accurate prediction is always a great difficulty is the current study.Popularity prediction of online content aims to predict the popularity in the future based on its early diffusion status.Existing models for popularity prediction were mostly based on discovering network features or fitting the equation into a varying time function that the accuracy of current popularity prediction model was not high enough.Therefore,with the help of the weak ties theory in sociology,the concept of tie strength was introduced and a multilinear regression equation was constructed combined with the early popularity.A TSL model to predict the popularity of Facebook’s well-known pages was proposed.The main contribution of this article was to solve the problem and few or no work based on sociology.A high linear correlation between the proportion of faithful fans was existed in Facebook homepage with frequent shares in the early and the future popularity.Compared with other baseline models,an experimental study of Facebook (including 1.54 million shares) illustrates the effectiveness of the proposed TSL model,and the performance is better than the existing similar methods.     

A Novel Dual Pointer Approach for Entity Mention Extraction

The named entity extraction task aims to extract entity mentions from the unstructured text, including names of people, places, institutions and so on. It plays an important role in many Natural language processing (NLP) tasks, such as knowledge bases construction, automatic question answering system and information extraction. Most of the existing entity extraction studies are based on the long text data, which are easier to annotate due to the sufficient contextual information. Extracting entities from short texts such as search queries, conversations is still a challenging task. This paper proposes a dual pointer approach for entity mention extraction, it extracts one entities by two position pointers of the input sentence. The end-to-end deep neural networks model based on the proposed approach can extract the entities by serially generating the dual pointers. The evaluation results on the Chinese public dataset show that the model achieves the state-of-the-art results over the baseline models.