Topic Space Research Articles

HierarchicalTopics: Visually Exploring Large Text Collections Using Topic Hierarchies

Analyzing large textual collections has become increasingly challenging given the size of the data available and the rate that more data is being generated. Topic-based text summarization methods coupled with interactive visualizations have presented promising approaches to address the challenge of analyzing large text corpora. As the text corpora and vocabulary grow larger, more topics need to be generated in order to capture the meaningful latent themes and nuances in the corpora. However, it is difficult for most of current topic-based visualizations to represent large number of topics without being cluttered or illegible. To facilitate the representation and navigation of a large number of topics, we propose a visual analytics system--HierarchicalTopic (HT). HT integrates a computational algorithm, Topic Rose Tree, with an interactive visual interface. The Topic Rose Tree constructs a topic hierarchy based on a list of topics. The interactive visual interface is designed to present the topic content as well as temporal evolution of topics in a hierarchical fashion. User interactions are provided for users to make changes to the topic hierarchy based on their mental model of the topic space. To qualitatively evaluate HT, we present a case study that showcases how HierarchicalTopics aid expert users in making sense of a large number of topics and discovering interesting patterns of topic groups. We have also conducted a user study to quantitatively evaluate the effect of hierarchical topic structure. The study results reveal that the HT leads to faster identification of large number of relevant topics. We have also solicited user feedback during the experiments and incorporated some suggestions into the current version of HierarchicalTopics.

Autonomous adaptive exploration using realtime online spatiotemporal topic modeling

The exploration of dangerous environments such as underwater coral reefs and shipwrecks is a difficult and potentially life-threatening task for humans, which naturally makes the use of an autonomous robotic system very appealing. This paper presents such an autonomous system, which is capable of autonomous exploration, and shows its use in a series of experiments to collect image data in challenging underwater marine environments. We present novel contributions on three fronts. First, we present an online topic-modeling-based technique to describe what is being observed using a low-dimensional semantic descriptor. This descriptor attempts to be invariant to observations of different corals belonging to the same species, or observations of similar types of rocks observed from different viewpoints. Second, we use the topic descriptor to compute the surprise score of the current observation. This is done by maintaining an online summary of observations thus far, and then computing the surprise score as the distance of the current observation from the summary in the topic space. Finally, we present a novel control strategy for an underwater robot that allows for intelligent traversal, hovering over surprising observations, and swimming quickly over previously seen corals and rocks.

Mining user-contributed photos for personalized product recommendation

With the advent and popularity of social media, users are willing to share their experiences by photos, reviews, blogs, and so on. The social media contents shared by these users reveal potential shopping needs. Product recommender is not limited to just e-commerce sites, it can also be expanded to social media sites. In this paper, we propose a novel hierarchical user interest mining (Huim) approach for personalized products recommendation. The input of our approach consists of user-contributed photos and user generated content (UGC), which include user-annotated photo tags and the comments from others in a social site. The proposed approach consists of four steps. First, we make full use of the visual information and UGC of its photos to mine user's interest. Second, we represent user interest by a topic distribution vector, and apply our proposed Huim to enhance interest-related topics. Third, we also represent each product by a topic distribution vector. Then, we measure the relevance of user and product in the topic space and determine the rank of each product for the user. We conduct a series of experiments on Flickr users and the products from Bing Shopping. Experimental results show the effectiveness of the proposed approach.

Cross domain recommendation based on multi-type media fusion

Due to the scarcity of user interest information in the target domain, recommender systems generally suffer from the sparsity problem. To alleviate this limitation, one natural way is to transfer user interests in other domains to the target domain. However, objects in different domains may be in different media types, which make it very difficult to find the correlations between them. In this paper, we propose a Bayesian hierarchical approach based on Latent Dirichlet Allocation (LDA) to transfer user interests cross domains or media. We model documents (corresponding to media objects) from different domains and user interests in a common topic space, and learn topic distributions for documents and user interests together. Specifically, to learn the model, we combine multi-type media information: media descriptions, user-generated text data and ratings. With this model, recommendation can be done in multiple ways, via predicting ratings, comparing topic distributions of documents and user interests directly and so on. Experiments on two real world datasets demonstrate that our proposed method is effective in addressing the sparsity problem by transferring user interests cross domains.

A novel machine learning approach to rank web forum posts

Since the user generated contents in Web forums are rich but vary in quality, ranging from excellent detailed opinions to simple repetition of the content of previous, or even spams, it is difficult to find high quality information in the process of post browsing, retrieval and other Web forum applications. In this paper, we propose a novel machine learning approach named LGPRank to evaluate the web forum posts, where a genetic programming architecture is used to rank Web forum posts according to the qualities of their contents. In order to address the shortcomings of current studies, we take both the semantic-free and semantic-specific information of a post into account. We propose a set of new features named Latent Dirichlet Allocation (LDA) semantic features which are computed in LDA topic space. The proposed features as well as content surface features and forum specific features are used in the learning process. Experiments are conducted on three web forum datasets in comparison with methods used in prior ranking research. LGPRank outperforms all the other methods in terms of P@N, NDCG@N and MAP measures. Furthermore, the experimental results also indicate that the proposed LDA semantic features have a positive effect in improving the ranking performance.

The Need for More Workshops in Laparoscopic Surgery and Surgical Anatomy for European Gynaecological Oncology Trainees

The objective of this study was to highlight the relative preference of European gynecologic oncology trainees for workshops that could support and supplement their training needs. A Web-based survey was sent to 900 trainees on the European Network of Young Gynaecological Oncologists database in November 2011. Respondents were asked to rate a 13-item questionnaire (using a 1- to 5-point Likert scale) on workshop topics they felt would most benefit their training requirements. Free text space for additional topics was also provided. Descriptive analysis was used to describe the mean scores reported for different items. A complete linkage hierarchical cluster analysis with Dendron plot was used to assess any clustering of data, and Cronbach α was used to assess the internal reliability of the questionnaire. One hundred ninety trainees from 37 countries responded to the survey, giving a 21% response rate. The 3 most important topics reported were laparoscopic surgery; surgical anatomy, and imaging techniques in gynecologic oncology. The Dendron plot indicated 4 different clusters of workshops (research related skills, supportive ancillary skills, related nonsurgical subspecialties, and core surgical skills) reflecting different competencies trainees need to meet. There was no significant association between individual country of training and workshop preference. The mean duration of the workshop preferred by 71% of respondents was 2 days. Cronbach α of the 13-item questionnaire was 0.78, which suggests good internal consistency/reliability. This report for the first time highlights the relative importance and significance European trainees attach to some of their training needs in gynecologic oncology. Laparoscopic surgery, surgical anatomy, and imaging appear to be the 3 areas of greatest need. The European Society of Gynaecological Oncology, other national specialist societies, and institutions should direct additional training efforts at these areas.

PREMENY SEMINÁRA SV. LADISLAVA V NITRE V ARCHÍVNYCH PRAMEŇOCH 18. STOROČIA

Konštantínove listy/Constantine’s Letters is an international scientific journal which has been published in Constantine the Philosopher University in Nitra since 2008. The main goal of the journal is to present problems and themes intertwined with the cultural and political history of Great Moravia during the 9th century as well as continuity of Cyrillo-Methodian tradition in later times. Next, there is also space for other adjacent topics connected with Slavonic and Byzantine history, philology, philosophy, literature, cultural relations as well as medieval and early modern age ecclesiastical history and iconography.

MENOVANIA ČLENOV SPIŠSKEJ KAPITULY DO VZNIKU SPIŠSKÉHO BISKUPSTVA V ROKU 1776

Konštantínove listy/Constantine’s Letters is an international scientific journal which has been published in Constantine the Philosopher University in Nitra since 2008. The main goal of the journal is to present problems and themes intertwined with the cultural and political history of Great Moravia during the 9th century as well as continuity of Cyrillo-Methodian tradition in later times. Next, there is also space for other adjacent topics connected with Slavonic and Byzantine history, philology, philosophy, literature, cultural relations as well as medieval and early modern age ecclesiastical history and iconography.

KONŠTANTÍN, METOD A ICH UČENÍCI PRED TVÁRAMI BYZANTSKÝCH CISÁROV

Konštantínove listy/Constantine’s Letters is an international scientific journal which has been published in Constantine the Philosopher University in Nitra since 2008. The main goal of the journal is to present problems and themes intertwined with the cultural and political history of Great Moravia during the 9th century as well as continuity of Cyrillo-Methodian tradition in later times. Next, there is also space for other adjacent topics connected with Slavonic and Byzantine history, philology, philosophy, literature, cultural relations as well as medieval and early modern age ecclesiastical history and iconography.

MEDIÁLNA GRAMOTNOSŤ – PREDPOKLAD PRE ŠÍRENIE KULTÚRNYCH TRADÍCIÍ

Konštantínove listy/Constantine’s Letters is an international scientific journal which has been published in Constantine the Philosopher University in Nitra since 2008. The main goal of the journal is to present problems and themes intertwined with the cultural and political history of Great Moravia during the 9th century as well as continuity of Cyrillo-Methodian tradition in later times. Next, there is also space for other adjacent topics connected with Slavonic and Byzantine history, philology, philosophy, literature, cultural relations as well as medieval and early modern age ecclesiastical history and iconography.

A link-bridged topic model for cross-domain document classification

Transfer learning utilizes labeled data available from some related domain (source domain) for achieving effective knowledge transformation to the target domain. However, most state-of-the-art cross-domain classification methods treat documents as plain text and ignore the hyperlink (or citation) relationship existing among the documents. In this paper, we propose a novel cross-domain document classification approach called Link-Bridged Topic model (LBT). LBT consists of two key steps. Firstly, LBT utilizes an auxiliary link network to discover the direct or indirect co-citation relationship among documents by embedding the background knowledge into a graph kernel. The mined co-citation relationship is leveraged to bridge the gap across different domains. Secondly, LBT simultaneously combines the content information and link structures into a unified latent topic model. The model is based on an assumption that the documents of source and target domains share some common topics from the point of view of both content information and link structure. By mapping both domains data into the latent topic spaces, LBT encodes the knowledge about domain commonality and difference as the shared topics with associated differential probabilities. The learned latent topics must be consistent with the source and target data, as well as content and link statistics. Then the shared topics act as the bridge to facilitate knowledge transfer from the source to the target domains. Experiments on different types of datasets show that our algorithm significantly improves the generalization performance of cross-domain document classification.

Mining One Percent of Twitter: Collections, Baselines, Sampling

The objective of the paper is to reflect on the affordances of different techniques for making Twitter collections and to suggest the use of a random sampling technique, made possible by Twitter’s Streaming API (Application Programming Interface), for baselining, scoping, and contextualising practices and issues. It discusses this technique by analysing a one per cent sample of all tweets posted during a 24-hour period and introducing a number of analytical directions considered useful for qualifying some of the core elements of the platform, in particular hashtags. To situate the proposal, the report first discusses how platforms propose particular affordances but leave considerable margins for the emergence of a wide variety of practices. This argument is then related to the question of how medium and sampling technique are intrinsically connected. Background Social media platforms present numerous challenges to empirical research, making it different from researching cases in offline environments, but also different from studying the “open” Web. Because of the limited access possibilities and the sheer size of platforms like Facebook or Twitter, the question of delimitation, i.e. the selection of subsets to analyse, is particularly relevant. Whilst sampling techniques have been thoroughly discussed in the context of social science research, sampling procedures in the context of social media analysis are far from being fully understood. Even for Twitter, a platform having received considerable attention from empirical researchers due to its relative openness to data collection, methodology is largely emergent. In particular the question of how smaller collections relate to the entirety of activities of the platform is quite unclear. Recent work comparing case based studies to gain a broader picture and the development of graph theoretical methods for sampling are certainly steps in the right direction, but it seems that truly large-scale Twitter studies are limited to computer science departments, where epistemic orientation can differ considerably from work done in the humanities and social sciences.

A new temporal and social PMF-based method to predict users' interests in micro-blogging

Micro-blogging is becoming an increasingly popular social media platform where users can discover interesting information about the real world and especially corporations are able to understand customers' demands. The fast diffusion of information and the convenience of micro-blogging have resulted in large audiences sharing their daily activities, exchanging opinions and establishing friendships with others. By analyzing the user-generated contents, one can explore users' potential interests, which helps micro-blogging provide users with better personalized information services. Users' behaviors are affected by opinions of their friends and changes in their interests over time. Based on these intuitions, in this paper we propose a temporal and social probabilistic matrix factorization model to predict users' potential interests in micro-blogging. By exploiting the matrix factorization technique to learn latent features of users and topics, our model analyzes the impacts of time information and users' activities, including posting of tweets and establishing friendships with others, on the latent feature space of users and topics of their interests. The proposed model provides a unified way to fuse the time information and the social network structure to predict users' future interests accurately. The experimental results on Sina-weibo, one of the most popular micro-blogging sites in China, demonstrate the efficiency and effectiveness of our proposed model.

On Confidence-Constrained Rank Recovery in Topic Models

Topic models have been proposed to model a collection of data such as text documents and images in which each object (e.g., a document) contains a set of instances (e.g., words). In many topic models, the dimension of the latent topic space (the number of topics) is assumed to be a deterministic unknown. The number of topics significantly affects the prediction performance and interpretability of the estimated topics. In this paper, we propose a confidence-constrained rank minimization (CRM) to recover the exact number of topics in topic models with theoretical guarantees on recovery probability and mean squared error of the estimation. We provide a computationally-efficient optimization algorithm for the problem to further the applicability of the proposed framework to large real world datasets. Numerical evaluations are used to verify our theoretical results. Additionally, to illustrate the applicability of the proposed framework to practical problems, we provide results in image classification for two real world datasets and text classification for three real world datasets.

ECG Analysis Using Multiple Instance Learning for Myocardial Infarction Detection

This paper presents a useful technique for totally automatic detection of myocardial infarction from patients' ECGs. Due to the large number of heartbeats constituting an ECG and the high cost of having all the heartbeats manually labeled, supervised learning techniques have achieved limited success in ECG classification. In this paper, we first discuss the rationale for applying multiple instance learning (MIL) to automated ECG classification and then propose a new MIL strategy called latent topic MIL, by which ECGs are mapped into a topic space defined by a number of topics identified over all the unlabeled training heartbeats and support vector machine is directly applied to the ECG-level topic vectors. Our experimental results on real ECG datasets from the PTB diagnostic database demonstrate that, compared with existing MIL and supervised learning algorithms, the proposed algorithm is able to automatically detect ECGs with myocardial ischemia without labeling any heartbeats. Moreover, it improves classification quality in terms of both sensitivity and specificity.

High order pLSA for indexing tagged images

This work presents a method for the efficient indexing of tagged images. Tagged images are a common resource of social networks and occupy a large portion of the social media stream. Their basic characteristic is the co-existence of two heterogeneous information modalities, i.e. visual and tag, which refer to the same abstract meaning. This multi-modal nature of tagged images makes their efficient indexing a challenging task that apart from dealing with the heterogeneity of modalities, it needs to also exploit their complementary information capacity. Towards this objective, we propose the extension of probabilistic latent semantic analysis to higher order, so as to become applicable for more than two observable variables. Then, by treating images, visual features and tags as the three observable variables of an aspect model, we learn a space of latent topics that incorporates the semantics of both visual and tag information. Our novelty is on using the cross-modal dependencies learned from a corpus of images to approximate the joint distribution of the observable variables. By penalizing the co-existence of visual content and tags that are known from experience to exhibit low dependency, we manage to filter out the effect of noisy content in the resulting latent space.

Learn to Personalized Image Search From the Photo Sharing Websites

Increasingly developed social sharing websites like Flickr and Youtube allow users to create, share, annotate, and comment medias. The large-scale user-generated metadata not only facilitate users in sharing and organizing multimedia content, but provide useful information to improve media retrieval and management. Personalized search serves as one of such examples where the web search experience is improved by generating the returned list according to the modified user search intents. In this paper, we exploit the social annotations and propose a novel framework simultaneously considering the user and query relevance to learn to personalized image search. The basic premise is to embed the user preference and query-related search intent into user-specific topic spaces. Since the users' original annotation is too sparse for topic modeling, we need to enrich users' annotation pool before user-specific topic spaces construction. The proposed framework contains two components: 1) a ranking-based multicorrelation tensor factorization model is proposed to perform annotation prediction, which is considered as users' potential annotations for the images; 2) we introduce user-specific topic modeling to map the query relevance and user preference into the same user-specific topic space. For performance evaluation, two resources involved with users' social activities are employed. Experiments on a large-scale Flickr dataset demonstrate the effectiveness of the proposed method.

Query expansion based on contextual meaning of the query terms

In this paper, a new method for query expansion is proposed which expands the original query based on the contextual meaning of it. User provides the method with an original query and a set of relevant documents. The term space extracted from the relevant documents by the vector space model is transformed to a topic space by means of latent semantic analysis. The terms’ projections in the topic space are grouped to some clusters called topic clusters. Expansion terms are the terms which have the same topic as the query terms. The results are evaluated by users in Google search engines and evaluation results show that the proposed method improves significantly the user satisfaction. Key words: Query expansion, term clustering, contextual topic, latent semantic analysis, cluster analysis.

TopicNets

We present TopicNets , a Web-based system for visual and interactive analysis of large sets of documents using statistical topic models. A range of visualization types and control mechanisms to support knowledge discovery are presented. These include corpus- and document-specific views, iterative topic modeling, search, and visual filtering. Drill-down functionality is provided to allow analysts to visualize individual document sections and their relations within the global topic space. Analysts can search across a dataset through a set of expansion techniques on selected document and topic nodes. Furthermore, analysts can select relevant subsets of documents and perform real-time topic modeling on these subsets to interactively visualize topics at various levels of granularity, allowing for a better understanding of the documents. A discussion of the design and implementation choices for each visual analysis technique is presented. This is followed by a discussion of three diverse use cases in which TopicNets enables fast discovery of information that is otherwise hard to find. These include a corpus of 50,000 successful NSF grant proposals, 10,000 publications from a large research center, and single documents including a grant proposal and a PhD thesis.

An efficient approach to suggesting topically related web queries using hidden topic model

Keyword-based Web search is a widely used approach for locating information on the Web. However, Web users usually suffer from the difficulties of organizing and formulating appropriate input queries due to the lack of sufficient domain knowledge, which greatly affects the search performance. An effective tool to meet the information needs of a search engine user is to suggest Web queries that are topically related to their initial inquiry. Accurately computing query-to-query similarity scores is a key to improve the quality of these suggestions. Because of the short lengths of queries, traditional pseudo-relevance or implicit-relevance based approaches expand the expression of the queries for the similarity computation. They explicitly use a search engine as a complementary source and directly extract additional features (such as terms or URLs) from the top-listed or clicked search results. In this paper, we propose a novel approach by utilizing the hidden topic as an expandable feature. This has two steps. In the offline model-learning step, a hidden topic model is trained, and for each candidate query, its posterior distribution over the hidden topic space is determined to re-express the query instead of the lexical expression. In the online query suggestion step, after inferring the topic distribution for an input query in a similar way, we then calculate the similarity between candidate queries and the input query in terms of their corresponding topic distributions; and produce a suggestion list of candidate queries based on the similarity scores. Our experimental results on two real data sets show that the hidden topic based suggestion is much more efficient than the traditional term or URL based approach, and is effective in finding topically related queries for suggestion.