Search

LDA is based on the bag-of-words, which simplifies the complexity of modeling, but makes the semantic coherence of topics poor, and text representation ability is not strong. To solve this problem, this paper proposed the semantic distribution similarity based topic model. This model used GPU (generalized Pólya urn) model to add word-word and document-topic semantic distribution similarity to guide topic modeling under the framework of EM (expectation maximization) algorithm, which weakened the effect of bag-of-words hypothesis on topics from the semantic association level. Experiments on four public datasets show that the semantic distribution similarity based topic model is superior to the currently popular topic modeling algorithms in terms of topic semantic coherence and text classification accuracy, and the model improves the convergence speed and topic accuracy. [ABSTRACT FROM AUTHOR]

The aim of topic detection is to automatically identify the events and hot topics in social networks and continuously track known topics. Applying the traditional methods such as Latent Dirichlet Allocation and Probabilistic Latent Semantic Analysis is difficult given the high dimensionality of massive event texts and the short-text sparsity problems of social networks. The problem also exists of unclear topics caused by the sparse distribution of topics. To solve the above challenge, we propose a novel word embedding topic model by combining the topic model and the continuous bag-of-words mode (Cbow) method in word embedding method, named Cbow Topic Model (CTM), for topic detection and summary in social networks. We conduct similar word clustering of the target social network text dataset by introducing the classic Cbow word vectorization method, which can effectively learn the internal relationship between words and reduce the dimensionality of the input texts. We employ the topic model-to-model short text for effectively weakening the sparsity problem of social network texts. To detect and summarize the topic, we propose a topic detection method by leveraging similarity computing for social networks. We collected a Sina microblog dataset to conduct various experiments. The experimental results demonstrate that the CTM method is superior to the existing topic model method. [ABSTRACT FROM AUTHOR]

In order to solve the problem that the recommendation list :is biased towards popular projects and with poor diversity, this paper proposed the ARIFDP ( aggregation recommendation algorithm for embedding item fatigue and diversity preference) algorithm. First, it analyzed the user's diversity preferences by using the historical feedback data of the user to derive the user, got the degree of diversification. And then it constructed a project fatigue function negatively related to the number of evaluations; eventually aggregated the matrix decomposition and the project fatigue function, and the inclusion of various propensity degrees adjusted the weight of the fatigue function of the project, and increased the probability of the proposed project. Experimental results show that the ARIFDP algorithm can effectively improve the diversity of recommendation results on the premise of ensuring accuracy. [ABSTRACT FROM AUTHOR]

Aiming at the increasingly rich multi language information resources and multi-label data in news reports and scientific literatures, in order to mining the relevance between languages and the correlation between data, this paper proposed labeled bilingual topic model, which was applied on cross-lingual text classification and label recommendation. First of all, it could assume that the keywords in the scientific literature are relevant to the abstract in same article. And then it extracted the keywords and regarded it as labels, and aligned the labels with topics in topic model, instantiated the " latent" topic. Secondly, this paper trained the abstracts in article through the proposed topic model. Finally, it classified the new documents by cross-lingual text classifier, and also recommended the labels. The experiment result show that micro-F1 measure reaches 94. 81 % in cross-lingual text classification task, and the recommended labels also reflects the sematic relevance with documents. [ABSTRACT FROM AUTHOR]

Understandings of racism are produced and circulated in contemporary networked media forms, contributing new opportunities to both extend and challenge racializing discourses, images and frameworks. This study investigated how the concept of racism is used in Finnish public debate by employing a computational text analysis technique to derive topics related to racism from a large corpus of news media content and online discussion forum comments. Our findings show that discourses about racism are different in legacy media and online platform regarding both their prominence and framings. While social media produce various discourses of 'reverse racism', news media connects racism to historical and international contexts. We conclude that what racism is understood to be is not only an intensely political process but also one shaped by the type of media platform, specificities of Finnish language and national ideological battles. [ABSTRACT FROM AUTHOR]

This paper proposed a collaborative topic regression model based on the preference for frequent topic sets to address the item-cold-start problem in academic paper recommendation. The algorithm took into account the user' s preference for research hotspots when selected academic papers, and used frequent topic sets to represent research hotspots. So, it expressed user' s preference for research hotspots as the user's preference for frequent topic sets. Firstly, it obtained the papers-topic probability distribution matrix through LDA algorithm and filter out the topics with higher probability. Then, the algorithm mined the frequently-occurring topic sets and gets the relationships between papers and frequent topic sets. Finally, it used the user's preference for frequent topic sets for the prediction of unknown scores. Experiments on CiteULike datasets show that the algorithm improves the recall, accuracy and RMSE over the matrix factorization model and the collaborative topic regression model. [ABSTRACT FROM AUTHOR]

The traditional collaborative filtering model suffers from high-dimensional sparse user rating information and ignores user preference information contained in user reviews. To address the problem, this paper proposes a new collaborative filtering model UL SAM (UBCF LDA SIMILAR ADD MEAN) which integrates topic model with user-based collaborative filtering model. UL SAM extracts user preference information from user reviews through topic model and then fuses user preference information with user rating information by similarity fusion method to create fusion information. UL SAM creates collaborative filtering recommendations according to fusion information. It is the advantage of UL SAM on improving recommendation effectiveness that UL SAM enriches information for collaborative recommendation by integrating user preference with user rating information. Experimental results of two public datasets demonstrate significant improvement on recommendation effectiveness in our model. [ABSTRACT FROM AUTHOR]

Microblogs have a large amount of data but a few characters in the text, and their features are sparse. In order to improve the retrieval precision, we propose a microblog retrieval model combining BTM and graph theory. The lexical semantic correlation is used to calculate the correlation between features with labels in microblog text. Then we construct a bi-term topic model，use JSD distance to calculate the correlation of pair words in the short text that mapped to the model. Thirdly，we extract the entity and graph structure in CN-DBpedia，and then use the SimRank algorithm to calculate inter-entity correlation between graph structures. The above three correlations are the final correlation of the model. Finally，the Sina Weibo data set is used for the retrieval experiments. Experimental results show that compared with the retrieval model based on the combination of the implicit Dirichlet distribution algorithm and graph theory and the system model based on open data correlation and graph theory, the performance of the new model is significantly improved in MAP, accuracy and recall rate, indicating that the model has better retrieval performance. [ABSTRACT FROM AUTHOR]

In order to obtain the topic distribution of bilingual documents effectively, we propose a phrase-based Khmer-Chinese bilingual LDA topic model* We modify the bag-of-word model in the traditional LDA topic model and incorporate the concept of phrase (N-gram). The method considers the word order and context of the article in the topic prediction process and applies it to the bilingual environment of comparable corpus. It is based on a three-layer Bayesian network model. Under this framework，we firstly collect comparable Chinese and Khmer corpus，and each pair of bilingual comparable corpus shares a common topic distribution. And then we introduce the topic model of discovery topic and topic phrase： the topic of each word is firstly sampled； then its status is sampled as a phrase； and finally words from a particular topic phrase distribution are sampled. Experimental results show that the phrase-based bilingual LDA topic model is more capable of grasping the topic of the article than general bilingual LDA models and has better topic prediction ability. [ABSTRACT FROM AUTHOR]

Topic models like Latent Dirichlet Allocation (LDA) and its variants is a type of statistical model for discovering latent topics. However, as revealed by the previous research, some topics generated by LDA may be uninterpretable and semantically incoherent due to the occurrence of irrelevant words in these topics. To improve the semantic qualities of automatically discovered topics, we explore the distributional characteristics of words across topics to identify topic-indiscriminate words which are blamed for the low-quality topics. The main contribution of our research reported in this paper is that we develop a novel framework named Iterative Term Weighting Framework (ITWF) which can effectively identify and filter out topic-indiscriminate words from uncovered topics. In particular, the proposed framework first applies an entropy-based term weighting schemes and adopts a novel iterative method to identify topic-indiscriminate words. To the best of our knowledge, our research is among the very few successful work that aims to enhance both the semantic coherence and the interpretability of LDA-based topic modeling methods. The experimental results show that the proposed framework improves the effectiveness of LDA as well as its variants. [ABSTRACT FROM AUTHOR]

For the acquisition of user behavior preference in social network, usually a data mining will be conducted on the nearest neighborhood users or latestprojects based on the user's historical behavior data, so as to find similar behaviorrelationship for quantitative analysis; it can also focus on the awareness on the user-related context information based on cognitive psychology, so as to find its internal links forthe potential mining. However, these methods ignore the intrinsic link between the browsing behavior and the preferred topics in the user link connection, resulting in the limited precision and accuracy of the preference acquisition. Inspired by the theory of complex network link prediction and the topic model, anacquisition method for users' browsing behavior preference was proposed in this paper. In the multi-dimensional network link environment, by measuring the importance of the node via network centrality and search the social network link via setting the similarity threshold, the real-time multi-link information and the big data about users' browsing under each link were acquired, then the data were filteredand cleaned by using adjustable parameters. On this basis, according to the least squares criterion the data underwent a fusion and were used to construct a data node distribution model for user browsing behavior, then the frequent feature items of user browsing behavior preference were extracted. Based on the extracted feature terms, the variational Bayes approximation reasoning method was used to construct the preference topic model. Finally, the hierarchical VSM model representation method was used to establish the preference acquisition model of user browsing behavior, and the model was updated in real time by user feedback processing mechanism. The experimental results on the real data set showed that the link search method and the preference topic model provided by this paper are accurate and efficient. Compared with the classical cooperative filtering method and the context-awareness method, the precision, accuracy and effectiveness of the preference acquisition model provided this paper are significantly improved, and its adaptability has been significantly strengthened. [ABSTRACT FROM AUTHOR]

In networks, multiple contagions, such as information and purchasing behaviors, may interact with each other as they spread simultaneously. However, most of the existing information diffusion models are built on the assumption that each individual contagion spreads independently, regardless of their interactions. Gaining insights into such interaction is crucial to understand the contagion adoption behaviors, and thus can make better predictions. In this paper, we study the contagion adoption behavior under a set of interactions, specifically, the interactions among users, contagions' contents, and sentiments, which are learned from social network structures and texts. We develop an effective and efficient interaction-aware diffusion (IAD) framework, incorporating these interactions into a unified model. We also present a generative process to distinguish user roles, a co-training method to determine contagions' categories and a new topic model to obtain topic-specific sentiments. Evaluation on the large-scale Weibo dataset demonstrates that our proposal can learn how different users, contagion categories, and sentiments interact with each other efficiently. With these interactions, we can make a more accurate prediction than the state-of-art baselines. Moreover, we can better understand how the interactions influence the propagation process and thus can suggest useful directions for information promotion or suppression in viral marketing. [ABSTRACT FROM AUTHOR]

Accurate makespan estimation is imperative during production scheduling to increase the flexibility and efficiency of work plans. However, given the complexities of production systems and product customizations, it is challenging to estimate makespans with high accuracy. In this paper, we propose a topic model-based neural network (TM-NN) method to increase the accuracy of makespan estimation for assembly processes. First, unlike traditional methods that use influential factors as inputs, we extract assembly features using a latent Dirichlet allocation model that mines latent topic information from an assembly instruction corpus. Then, the assembly process is represented as a sequence model with both assembly topics and features of the product physical characteristics, the assembly process, the equipment, the personnel, and uncertainty. Finally, we use a structured numerical vector as the input to machine learning-based predictive models, including a neural network, a random forest, and a support vector machine, and estimate makespans. The results show that the proposed TM-NN method effectively extracts latent topics in assembly documents and significantly increases the accuracy of makespan estimation. • We propose a topic model-based neural network (TM-NN) method to increase the accuracy of makespan estimation for assembly processes. • TM-NN can extract latent topic information from an assembly instruction corpus. • We compare the average errors between the predicted and real makespans. • TM-NN exhibits superior effectiveness and accuracy. [ABSTRACT FROM AUTHOR]

Artificial Intelligence (AI) has affected all aspects of social life in recent years. This study reviews 177,204 documents published in 25 journals and 16 conferences in the AI research from 1990 to 2021, and applies the Latent Dirichlet allocation (LDA) model to extract the 40 topics from the abstracts. According to the topics obtained, 7 subfields of the AI field can be discovered: Approximate Reasoning , Computational Theory , Intelligent Automation , Artificial Neural Network , Machine Learning , Natural Language Processing , and Computer Vision. This study aggregates the results of the LDA model from the perspectives of year, publication source, and country/region. The aggregated result is the topic distribution from different perspectives. Analysis of the aggregated results reveals the research characteristics of different publication sources (countries/regions) in the AI research, and which publication sources (countries/regions) have similar research content. These results provide help for scholars and research institutions to choose research directions, and new entrants to understand the dynamics of the field. [ABSTRACT FROM AUTHOR]

With the help of topic models, social media data have become a valuable information source for manufacturers to identify product defects. However, when absorbing information on product defects, special defect-unrelated texts that mention product failures induced by customers will be mistaken as defect-related by existing methods, leading to biased results. Furthermore, extant topic models suffer from the "topic-indiscriminative" problem, which means topics derived by topic models are similar. In order to address these problems, we first create a lexicon to differentiate whether the texts mentioning product failures are defect-related. Then we propose a topic model named Improved Topic Model for Defect Information Derivation (ITMDID) to derive product defect information from defect-related texts. To enhance the discrimination of extracted topics, we consider the word coherence when inferring variables of ITMDID via Gibbs sampling. Finally, we applied the developed approaches to analyze two case studies of the automobile and laptop industries. Experimental results prove that our methods can derive product defects from social media data more accurately and comprehensively than state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Topic modeling has become an extraordinary tool for analyzing large, unclassified documents to detect patterns that use similar words. It also permits the discovery of hidden themes that pervade the collection, allowing annotation according to those themes, and then using these to summarize and search the text. This paper focuses on Latent Dirichlet Allocation, which is one of the most widely used methods for topic modeling. Trend detection and evolution of topics relating to research topics was attempted from the Journal of the Modern Language Association of America. The study also identified those documents that explained topics according to time series and citation. The experimental data consisted of 5,605 articles from the Journal between the years 1889-2007 held in the Journal Storage digital library. The results show that the Latent Dirichlet Allocation model can effectively detect distinct topics and those documents that explained these topics over time. [ABSTRACT FROM AUTHOR]

The cooperative hierarchical structure is a common and significant data structure observed in, or adopted by, many research areas, such as: text mining (author–paper–word) and multi-label classification (label–instance–feature). Renowned Bayesian approaches for cooperative hierarchical structure modeling are mostly based on hierarchical Bayesian models. However, these approaches suffer from a serious issue in that the number of hidden topics/factors needs to be fixed in advance and an inappropriate number may lead to overfitting or underfitting. One elegant way to resolve this issue is Bayesian nonparametric learning, but existing work in this area still cannot be applied to cooperative hierarchical structure modeling. In this paper, we propose a cooperative hierarchical Dirichlet process (CHDP) to fill this gap. Each node in a cooperative hierarchical structure is assigned a Dirichlet process to model its weights on the infinite hidden factors/topics. Together with measure inheritance from hierarchical Dirichlet process, two kinds of measure cooperation, i.e., superposition and maximization, are defined to capture the many-to-many relationships in the cooperative hierarchical structure. Furthermore, two constructive representations for CHDP, i.e., stick-breaking and international restaurant process, are designed to facilitate the model inference. Experiments on synthetic and real-world data with cooperative hierarchical structures demonstrate the properties and the ability of CHDP for cooperative hierarchical structure modeling and its potential for practical application scenarios. [ABSTRACT FROM AUTHOR]

Abstract Semantic embedding has demonstrated its value in latent representation learning of data, and can be effectively adopted for many applications. However, it is difficult to propose a joint learning framework for semantic embedding in Community Question Answer (CQA), because CQA data have multi-view and sparse properties. In this paper, we propose a generic Multi-modal Multi-view Semantic Embedding (MMSE) framework via a Bayesian model for question answering. Compared with existing semantic learning methods, the proposed model mainly has two advantages: (1) To deal with the multi-view property, we utilize the Gaussian topic model to learn semantic embedding from both local view and global view. (2) To deal with the sparse property of question answer pairs in CQA, social structure information is incorporated to enhance the quality of general text content semantic embedding from other answers by using the shared topic distribution to model the relationship between these two modalities (user relationship and text content). We evaluate our model for question answering and expert finding task, and the experimental results on two real-world datasets show the effectiveness of our MMSE model for semantic embedding learning. [ABSTRACT FROM AUTHOR]

Abstract In topic modeling framework, many Dirichlet-based models performances have been hindered by the limitations of the conjugate prior. It led to models with more flexible priors, such as the generalized Dirichlet distribution, that tend to capture semantic relationships between topics (topic correlation). Now these extensions also suffer from incomplete generative processes that complicate performances in traditional inferences such as VB (Variational Bayes) and CGS (Collaspsed Gibbs Sampling). As a result, the new approach, the CVB-LGDA (Collapsed Variational Bayesian inference for the Latent Generalized Dirichlet Allocation) presents a scheme that integrates a complete generative process to a robust inference technique for topic correlation and codebook analysis. Its performance in image classification, facial expression recognition, 3D objects categorization, and action recognition in videos shows its merits. [ABSTRACT FROM AUTHOR]

In multi-label classification settings, one of the most common problems is the massive label output space. To alleviate this, some methods opt to exploit label correlations to reduce the output space during prediction. However, these methods sacrifice efficiency or ignore global label correlations. In addition, label imbalances are another problem that is prevalent in multi-label classification. Current methods of correcting for imbalance oftentimes use single-label methods, which fail to consider label correlations. In this paper, we introduce general frameworks that incorporate topic modeling to seamlessly address both problems. We show that these frameworks can allow even the most naïve methods, such as Binary Relevance, to perform similarly to state-of-the-art methods. Furthermore, we show that our frameworks can also adapt state-of-the-art methods to perform better than the methods by themselves. [ABSTRACT FROM AUTHOR]

The volume of short text data increases rapidly these years. Data examples include tweets and online Q&A pairs. It is essential to organize and summarize these data automatically. Topic model is one of the effective approaches, whose application domains include text mining, personalized recommendation and so on. Conventional models like pLSA and LDA are designed for long text data. However, these models may suffer from the sparsity problem brought by lacking words in short text scenarios. Recent studies such as BTM show that using word co-occurrent pairs is effective to relieve the sparsity problem. However, both BTM and extended models ignore the quantifiable relationship between words. From our perspectives, two more related words should occur in the same topic. Based on this idea, we introduce a model named RIBS, which makes use of RNN to learn relationship. By using the learned relationship, we introduce a model named RIBS-Bigrams, which can display topics with bigrams. Through experiments on two open-source and real-world datasets, RIBS achieves better coherence in topic discovery, and RIBS-Bigrams achieves better readability in topic display. In the document characterization task, the document representation of RIBS can lead better purity and entropy in clustering, higher accuracy in classification. [ABSTRACT FROM AUTHOR]

Policy text contains large amount of diversified data and strictly conforms to standards and specifications, but the traditional text clustering method cannot solve the problems of high dimensionality, sparse features, and similar meanings, so this paper proposes a weighted algorithm based on the LDA-Gibbs model to improve the accuracy of policy text clustering. Firstly, it provides realistic basis for the assumptions of the LDA-Gibbs topic model and the weighted algorithm; secondly, it pre-processes the existing policy text simulated data, establishes the LDA-Gibbs model, forms a weighted algorithm, and generates training data to determine the number of optimal topics in the LDAGibbs model and completes the final clustering of the policy text; finally, by summarizing, classifying and deducing the conclusions of the experimental data, this paper proves the objective validity and effects of this method. Hopefully the overall design of this method can be applied in the prospective study on the formulation of new policies in the future, the retrospective evaluation and testing of the existing policies and the formation of a two-way interactive mechanism. [ABSTRACT FROM AUTHOR]

Abstract The development of mobile positioning technologies makes massive individual trajectory data easily accessible, which facilitates the revisit of spatial interaction issue in recent years. Researchers have proposed many methods to investigate the spatial interactions derived from human movements, such as the gravity model and radiation model. However, these studies have mainly focused on the interactions among areal units at an aggregated level, neglecting that in most cases, human movements are carried by vehicles and constrained by the underlying road network, which causes the interactions among roads. To fill this gap, we propose a novel approach to identify spatial interaction patterns of vehicle movements on urban road network. As the topic model originating from the domain of natural language processing has powerful advantages in extracting semantic relations of words from corpus, we utilize it to extract interaction relations of urban roads from massive vehicle trajectories. First, "strokes" (i.e., natural streets) are chosen as geographical units to represent the vehicle moving paths. Then, an analogy between geographical elements (i.e., stroke, moving path) and textual elements (i.e., word, document) is established, and a topic model is applied to the moving paths to identify the spatial interaction patterns on road network. From a mass of trajectory data collected by GNSS-equipped taxis in Beijing, the "topic patterns", which can be viewed as clusters of spatially interacted strokes, are identified, visualized and verified. It is argued that our proposed approach is effective in identifying spatial interaction patterns, which provides a new perspective for spatial interaction modelling and enriches the current spatial interaction studies. Highlights • We originally proposed to investigate the spatial interactions among linear streets instead of areal regions. • We proposed a method to represent vehicle moving paths, which can well reflect human driving behavior on road network. • We proposed an innovative approach to identify spatial interaction patterns on road network by topic modelling. • Our approach can better reveal the spatial interaction patterns compared to the community detection-based method. [ABSTRACT FROM AUTHOR]

Abstract Hashtags provide a simple and natural way of organizing content in microblog services. Along with the fast growing of microblog services, the task of recommending hashtags for microblogs has been given increasing attention in recent years. However, much of the research depends on hand-crafted features. Motivated by the successful use of neural models for many natural language processing tasks, in this paper, we adopt an attention based neural network to learn the representation of a microblog post. Unlike previous works, which only focus on content attention of microblogs, we propose a novel Topical Co-Attention Network (TCAN) that jointly models content attention and topic attention simultaneously, in the sense that the content representation(s) are used to guide the topic attention and the topic representation is used to guide content attention. We conduct experiments and test with different settings of TCAN on a large real-world dataset. Experimental results show that our model significantly outperforms various competitive baseline methods. Furthermore, the incorporation of topical co-attention mechanism gives more than 13.6% improvement in F1 score compared with the standard LSTM based methods. [ABSTRACT FROM AUTHOR]

The online reviews are one type of social media which are opinions generated by the users to comment on some special items. Since the sentiments are dependent on topics, probabilistic topic models have been widely used for sentiment analysis. However, most of existing methods only model the text, but rarely consider the users, who express the opinions, and the items, which the opinions are expressed on. Different users are usually concerned with different topics and use different sentiment expressions, a lenient user might tend to give positive review than a critical user. High-quality items tend to receive positive reviews than low-quality items. To better model the topics and sentiments, we argue that it is essential to explore reviews as well as users and items. To this end, we propose a novel model called User Item Sentiment Topic (UIST) which incorporates users and items for topic modeling and produces topic-word, user-topic, and item-topic distributions simultaneously. Extensive experiments on several datasets demonstrate the advantages and effectiveness of our method. The extracted topics with our method are more coherent and informative; consequently, the performance of sentiment classification is also improved. Furthermore, the user preference obtained with our method could be utilized for many personalized applications. [ABSTRACT FROM AUTHOR]

Abstract With the fast development of online social platforms, social emotion detection, focusing on predicting readers' emotions evoked by news articles, has been intensively investigated. Considering emotions as latent variables, various probabilistic graphical models have been proposed for emotion detection. However, the bag-of-words assumption prohibits those models from capturing the inter-relations between sentences in a document. Moreover, existing models can only detect emotions at either the document-level or the sentence-level. In this paper, we propose an effective Bayesian model, called hidden Topic–Emotion Transition model, by assuming that words in the same sentence share the same emotion and topic and modeling the emotions and topics in successive sentences as a Markov chain. By doing so, not only the document-level emotion but also the sentence-level emotion can be detected simultaneously. Experimental results on the two public corpora show that the proposed model outperforms state-of-the-art approaches on both document-level and sentence-level emotion detection. [ABSTRACT FROM AUTHOR]

One important necessity is to filter out the irrelevant information and organize the relevant information into meaningful categories. However, developing text classifiers often requires a large number of labeled documents as training examples.Manually labeling documents is costly and time-consuming.More importantly, it becomes unrealistic to knowall the categories covered by the documents beforehand. Recently, a fewmethods have been proposed to label documents by using a small set of relevant keywords for each category, known as dataless text classification. In this article, we propose a seed-guided topic model for the dataless text filtering and classification (named DFC). Given a collection of unlabeled documents, and for each specified category a small set of seed words that are relevant to the semantic meaning of the category, DFC filters out the irrelevant documents and classifies the relevant documents into the corresponding categories through topic influence. DFC models two kinds of topics: category-topics and general-topics. Also, there are two kinds of category-topics: relevant-topics and irrelevant-topics. Each relevant-topic is associated with one specific category, representing its semantic meaning. The irrelevant-topics represent the semantics of the unknown categories covered by the document collection. And the general-topics capture the global semantic information. DFC assumes that each document is associated with a single category-topic and a mixture of general-topics. A novelty of the model is that DFC learns the topics by exploiting the explicit word co-occurrence patterns between the seed words and regular words (i.e., non-seed words) in the document collection. A document is then filtered, or classified, based on its posterior category-topic assignment. Experiments on two widely used datasets show that DFC consistently outperforms the state-of-the-art dataless text classifiers for both classi- fication with filtering and classification without filtering. In many tasks, DFC can also achieve comparable or even better classification accuracy than the state-of-the-art supervised learning solutions. Our experimental results further show that DFC is insensitive to the tuning parameters.Moreover, we conduct a thorough study about the impact of seed words for existing dataless text classification techniques. The results reveal that it is not using more seed words but the document coverage of the seed words for the corresponding category that affects the dataless classification performance. [ABSTRACT FROM AUTHOR]

Abstract Emotional conversation generation has elicited a wide interest in both academia and industry. However, existing emotional neural conversation systems tend to ignore the necessity to combine topic and emotion in generating responses, possibly leading to a decline in the quality of responses. This paper proposes a topic-enhanced emotional conversation generation model that incorporates emotional factors and topic information into the conversation system, by using two mechanisms. First, we use a Twitter latent Dirichlet allocation (LDA) model to obtain topic words of the input sequences as extra prior information, ensuring the consistency of content between posts and responses for emotional conversation generation. Second, the system uses a dynamic emotional attention mechanism to adaptively acquire content-related and affective information of the input texts and extra topics. The advantage of this study lies in the fact that the presented model can generate abundant emotional responses, with the contents being related and diverse. To demonstrate the effectiveness of our method, we conduct extensive experiments on large-scale Weibo post–response pairs. Experimental results show that our method achieves good performance, even outperforming some existing models. Highlights • We present a topic-enhanced neural emotion conversation generation model (TE-ECG) with attention mechanism. • The topic words are obtained from a pre-trained Twitter LDA model to ensure the generated response is related to the post. • A novel dynamic emotional attention mechanism is proposed to capture the emotional context and topic information. • The TE-ECG model can generate responses at both the emotion- and content-related levels. [ABSTRACT FROM AUTHOR]

Highlights • The effective and targeted interaction between driving environment and drivers. • Using topic models, the underlying structure of driving styles is established. • Two topic models (mLDA and mHLDA) are presented for driving style analysis. • The driving styles can be recognized properly using the proposed models. Abstract With the explosion of information in our current era, senders of information increasingly need to target their messages to recipients. However, messages within transportation systems, including traffic information and commercial advertisements, tend to be transmitted to all drivers indiscriminately. This is because the information providers (such as other vehicles, roads, facilities, buildings etc.), can hardly recognize the variations within drivers, who should be treated differently as information recipients. As a result of the rapid development of data collection technologies and machine learning techniques in recent years, extraction and recognition of drivers' unique driving style from actual driving behaviour data become possible. In this paper, two kinds of topic models are investigated: mLDA and mHLDA, to discover distinguishable driving style information with hidden structure from the real-world driving behaviour data. The results show that the proposed models can successfully recognize the differences between driving styles. The study is of great value for providing deep insight into the underlying structure of driving styles and can effectively support the recognition of drivers with different driving styles. [ABSTRACT FROM AUTHOR]

A phrase is a natural, meaningful, and essential semantic unit. In topic modeling, visualizing phrases for individual topics is an effective way to explore and understand unstructured text corpora. However, from phrase quality and topical cohesion perspectives, the outcomes of existing approaches remain to be improved. Usually, the process of topical phrase mining is twofold: phrase mining and topic modeling. For phrase mining, existing approaches often suffer from order sensitive and inappropriate segmentation problems, which make them often extract inferior quality phrases. For topic modeling, traditional topic models do not fully consider the constraints induced by phrases, which may weaken the cohesion. Moreover, existing approaches often suffer from losing domain terminologies since they neglect the impact of domain-level topical distribution. In this paper, we propose an efficient method for high quality and cohesive topical phrase mining. A high quality phrase should satisfy frequency, phraseness, completeness, and appropriateness criteria. In our framework, we integrate quality guaranteed phrase mining method, a novel topic model incorporating the constraint of phrases, and a novel document clustering method into an iterative framework to improve both phrase quality and topical cohesion. We also describe efficient algorithmic designs to execute these methods efficiently. The empirical verification demonstrates that our method outperforms the state-of-the-art methods from the aspects of both interpretability and efficiency. [ABSTRACT FROM AUTHOR]

This paper proposes a new approach to multi-object tracking by semantic topic discovery. We dynamically cluster frame-by-frame detections and treat objects as topics, allowing the application of the Dirichlet process mixture model. The tracking problem is cast as a topic-discovery task, where the video sequence is treated analogously to a document. It addresses tracking issues such as object exclusivity constraints as well as tracking management without the need for heuristic thresholds. Variation of object appearance is modeled as the dynamics of word co-occurrence and handled by updating the cluster parameters across the sequence in the dynamical clustering procedure. We develop two kinds of visual representation based on super-pixel and deformable part model and integrate them into the model of automatic topic discovery for tracking rigid and non-rigid objects, respectively. In experiments on public data sets, we demonstrate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Highlights • A novel universal affective model for classifying social emotions is proposed. • ATF-IDF is developed to enhance the semantic relationships between biterms. • A word-level emotional lexicon is established for background words by using SWAT. • UAM is very effective in detecting emotions in both short texts and long texts. Graphical abstract Abstract As the rapid development of Web 2.0 communities, social media service providers offer users a convenient way to share and create their own contents such as online comments, blogs, microblogs/tweets, etc. Understanding the latent emotions of such short texts from social media via the computational model is an important issue as such a model will help us to identify the social events and make better decisions (e.g., investment in stocking market). However, it is always very challenge to detect emotions from above user-generated contents due to the sparsity problem (e.g., a tweet is a short message). In this article, we propose an universal affective model (UAM) to classify readers’ emotions over unlabeled short texts. Different from conventional text classification model, the UAM structurally consists of topic-level and term-level sub-models, and detects social emotions from the perspective of readers in social media. Through the evaluation on real-world data sets, the experimental results validate the effectiveness of the proposed model in terms of the effectiveness and accuracy. [ABSTRACT FROM AUTHOR]

Background: Proteins are a kind of macromolecules and the main component of a cell, and thus it is the most essential and versatile material of life. The research of protein functions is of great significance in decoding the secret of life. In recent years, researchers have introduced multi-label supervised topic model such as Labeled Latent Dirichlet Allocation (Labeled-LDA) into protein function prediction, which can obtain more accurate and explanatory prediction. However, the topic-label corresponding way of Labeled-LDA is associating each label (GO term) with a corresponding topic directly, which makes the latent topics to be completely degenerated, and ignores the differences between labels and latent topics. Result: To achieve more accurate probabilistic modeling of function label, we propose a Partially Function-to-Topic Prediction (PFTP) model for introducing the local topics subset corresponding to each function label. Meanwhile, PFTP not only supports latent topics subset within a given function label but also a background topic corresponding to a 'fake' function label, which represents common semantic of protein function. Related definitions and the topic modeling process of PFTP are described in this paper. In a 5-fold cross validation experiment on yeast and human datasets, PFTP significantly outperforms five widely adopted methods for protein function prediction. Meanwhile, the impact of model parameters on prediction performance and the latent topics discovered by PFTP are also discussed in this paper. Conclusion: All of the experimental results provide evidence that PFTP is effective and have potential value for predicting protein function. Based on its ability of discovering more-refined latent sub-structure of function label, we can anticipate that PFTP is a potential method to reveal a deeper biological explanation for protein functions. [ABSTRACT FROM AUTHOR]

The quality of any text mining technique is highly dependent on the features that are used to represent the document collection. A classical form of document representation is the vector space model (VSM), according to which the documents are represented as vectors of weights that correspond to the features of the documents. The bag-of-words model is the most popular VSM approach due to its simplicity and general applicability, but this model does not include term dependency and has a high dimensionality. In the literature, several models for document representation have been proposed in order to capture the dependency of terms. Among them, the topic model representation is one of the most interesting approaches - since it describes the collection of documents in a way that reveals their internal structure and the interrelationships therein, and also provides a dimensionality reduction. However, even for topic models, the efficient extraction of information concerning the relations among terms for document representation is still a major research challenge. In order to address this issue, we proposed thelatent association rule cluster based model (LARCM). The LARCM is a non-probabilistic topic model that makes use of association rule clustering to build a document representation with low dimensionality in such a way that each feature (i.e., topic) is comprised of information concerning relations among the terms. We evaluated the interpretability of the topics obtained by using our proposed model against the ones provided by the traditional latent dirichlet allocation (LDA) model and the LDA model using a document representation that includes correlated terms (i.e., bag-of-related-words). The experimental results indicated that the LARCM provides topics with better interpretability than the LDA models. Additionally, we used the topics obtained by the LARCM in two different applications: text classification and page recommendation. With respect to text classification, the topics were used to improve document collection representation. Concerning page recommendation, topics were used as contextual information in context-aware recommender systems. Results have shown that the topics provided by the LARCM can be used to improve both applications. [ABSTRACT FROM AUTHOR]

Discussion forums is a major component of MOOC platform. The teaching interaction in MOOC learning process mainly occurs in discussion forums by topic. Aiming at the problem of students' ability difference in the course of MOOC teaching, in this paper a topic interaction model based on local community detection is proposed. Through topic modeling for students' interaction ability in discussion forums, local community detection algorithm is used to classify students' various abilities reasonably. Through tracking and analyzing student behavior information on a highly interactive MOOC platform, the accuracy rate of the proposed model is obviously higher than that of the traditional assessment methods. The teaching practice using the model shows that students' abilities in all aspects are improved by means of pertinence classroom communication and training. [ABSTRACT FROM AUTHOR]

As an efficient sampling algorithm for latent dirichlet allocation SparseLDA uses cache strategy to improve the time and space efficiency of its standard gibbs sampling algorithm (StdGibbs) by recycling previous computation. However, SparseLDA cannot further improve the time-efficiency of StdGibbs, since the amount of recycled computation is limited. This is because the word types of two adjacent tokens are usually different and the previous computation cannot be further recycled easily. To solve this problem, in this paper we propose a new algorithm named Efficient SparseLDA (ESparseLDA) based on SparseLDA. The main idea of ESparseLDA is to first rearrange the tokens within one text according to the word types so that the tokens of the same word type are aggregated together and then recycle more computation while making no approximation and ensuring the exactness. In this paper, we make detailed theoretical explanations and comparative experimental analyses on the correctness, exactness and time-efficiency of ESparseLDA. In detail, the statistical significance tests on perplexities strictly show that ESparseLDA is correct and exact. In addition, the running time results show that the time-efficiency of ESparseLDA is the higher than SparseLDA in varying degrees from 5.06% to 31.85% on the different datasets used in experiments. [ABSTRACT FROM AUTHOR]

Short text categorization is a crucial issue to many applications, e.g., Information Retrieval, Question-Answering System, MRI Database Construction and so forth. Many researches focus on data sparsity and ambiguity issues in short text categorization. To tackle these issues, we propose a novel short text categorization strategy based on abundant representation, which utilizes Bi-directional Recurrent Neural Network(Bi-RNN) with Long Short-Term Memory(LSTM) and topic model to catch more contextual and semantic information. Bi-RNN enriches contextual information, and topic model discovers more latent semantic information for abundant text representation of short text. Experimental results demonstrate that the proposed model is comparable to state-of-the-art neural network models and method proposed is effective. [ABSTRACT FROM AUTHOR]

Latent Dirichlet allocation is the prevalent topic model and performs well for image classification. However, it ignores visual word spatial information, which affects topic assignment accuracy. This paper proposes an effective topic model framework based on spatial pyramids including visual word regional information: spatial topic pyramid model (STPM). STPM divides the images into different scale regions and uses the regional topic distributions to represent the images. The regional topic distributions effectively represent image characteristics, because they include global information (regarding the image as a single region) and the regional relationships of visual words in different scale regions. Since the pyramid layers are independent, different topic models and parameters can be used for different scale layers. It makes STPM flexible and easily extensible. [ABSTRACT FROM AUTHOR]

Several types of online chat system have been developed; however, there exist no recommendation systems for the recommendation of topics suitable for discussion with a given individual at a particular time. This paper proposes a hot-topic recommendation system based on analysis of the tweets posted by the user, his/her chat partners, and similar users of his/her chat partners, as well as hashtags trending in Twitter. In experiments, the proposed system, which is based on the well-known Latent Dirichlet Allocation (LDA) algorithm, was shown to outperform existing recommendation systems with regard to computational efficiency as well as prediction accuracy. [ABSTRACT FROM AUTHOR]

In recent years, the explosive growth of the number of App brings difficulties for user to find a suitable App. There are many limitations of the traditional recommendation approaches when they are applied to App, such as the cold start problem and different choice bias on different types of applications. This paper proposed a personalized recommendation algorithm TGRW, which was based on topic grouping and random walk. Because users have different biases and underling choice factors on the App of different categories, TGRW firstly divided App into categories based on their description information. Then it constructed a triple tuple graphic model consisting of user, App group and App. By applying random walk to calculate the preferential probability of user to each App, it obtained the recommendation list. Extensive experiments on the real data set shows TGRW gains significant improvements on recommendation performances than other methods. [ABSTRACT FROM AUTHOR]

Abstract: The spatio-temporal urban movement patterns can be extracted from the massive trajectory data recorded by GPS devices. Effectively analyzing the massive and complex traffic data and then finding useful information hidden in such data constitute challenging yet meaningful research. By providing the interactive visual analysis of the underlying traffic patterns of the city, the results can guide the users in choosing ideal locations for setting up shops for business operations. We construct the topic model to analyze the GPS taxi trajectory data. The topic information is combined with the traffic volume information to choose the representative candidate areas. Then, traffic flow graphs are generated between candidate areas to show the distribution of such areas and the taxi running rules. We study the distribution and semantics of the topics from three aspects: time, space, and POIs (points of interest). Thus, we can enhance the user’s understanding of area characters by semantics. In addition, inspired by the wheels of vehicles, we design a metaphor-based glyph to summarize the multi-dimensional attributes of each candidate area. Users can explore the prospective areas’ multiple attributes over time through varied interactions to learn the details of the area from multiple perspectives. Finally, we design and implement a visual analysis prototype system of traffic trajectory data as well as verify the feasibility and validity of the system in the case study.Graphical Abstract: [ABSTRACT FROM AUTHOR]

Topic and sentiment joint modelling has been successfully used in sentiment analysis for product reviews. However, the problem of text sparse is universal with the widespread smart devices and the shorter product reviews. In this paper, we propose a joint sentiment-topic model WSTM (Word-pair Sentiment-Topic Model) for the short text reviews, detecting sentiments and topics simultaneously from the text, especially considering the text sparse problem. Unlike other topic models modelling the generative process of each document, our directly models the generation of the word-pair set from the whole global corpus. In the generative process of WSTM, all of the words in a sentence have the same sentiment polarity, and two words in a word-pair have the same topic. We apply WSTM to two real-life Chinese product review datasets to verify its performance. In three experiments, compared with the existing approaches, the results demonstrate WSTM is quantitatively effective on both topic discovery and document level sentiment. [ABSTRACT FROM AUTHOR]

In order to solve the problems of data sparseness and only using the user's rating behavior to recommend in the traditional collaborative filtering algorithm, a kind of matrix factorization recommendation algorithm combining users' reviews was proposed. First, the algorithms utilized topic model to generate review topics distribution, and extracted accurate user interests and items characteristics by integrating the score. Then, combined with users' interests and item characteristics, the nearest neighbors of users and items were calculated according to the cosine similarity. Last, the nearest regularization terms were introduced into the matrix decomposition model.The USRMF was compared with the traditional collaborative filtering algorithm and the regularization matrix decomposition algo-rithm.The experimental result shows that the USRMF can improve the accuracy of recommendation in sparse datascts. [ABSTRACT FROM AUTHOR]

Since the traditional methods of micro-blog sentiment analysis ignore the influence of images, the special symbol information and the context semantic information, which leads to the accuracy rate of the emotion analysis method is not high. Therefore , this paper presented a micro-blog sentiment method, which was illustrated and based on transferred variable. Firstly, it paper constructed an unsupervised sentiment analysis model USAMTV, which was based on transferred variable, to analysis the distribution of emotion in the text. That was to say, the emotion and topic were categorized by introducing the sentimental transferred variable of conjunction and thematic transferred variable of reposting symbols. Then it introduced the image factors as emotional weight to decide the emotional distribution of the text. Finally, it obtained the overall sentimental orientation of the micro-blog. Compared with the JST model and the ASUM model, this proposed method had higher accuracy in the real-world test dataset. The experimental results show' that the proposed method can predict the emotional tendency of microblog more accurately. [ABSTRACT FROM AUTHOR]

Many of today’s online news websites and aggregator apps have enabled users to publish their opinions without respect to time and place. Existing works on topic-based sentiment analysis of product reviews cannot be applied to online news directly because of the following two reasons: (1) The dynamic nature of news streams require the topic and sentiment analysis model also to be dynamically updated. (2) The user interactions among news comments can easily lead to inaccurate topic extraction and sentiment classification. In this paper, we propose a novel probabilistic generative model (DTSA) to extract topics and the specified sentiments from news streams and analyze their evolution over time simultaneously. In DTSA, three different timescale models are studied to account for the historical dependencies of sentiment-topic word distributions at current epoch, continuous, skip and multiple timescale models. Additionally, we further consider the links among news comments to avoid the error caused by user interactions. In order to mine more interpretable topics, a Conditional Random Fields (CRF) model is adopted to label a set of meaningful phrases for augmenting the bag-of-word features. Finally, we derive distributed online inference procedures to update the model with newly arrived data and show the effectiveness of our proposed model on real-world data sets. [ABSTRACT FROM AUTHOR]