• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.12
• /
• pp.3274-3292
• /
• 2013

Unsupervised methods for image segmentation are recently drawing attention because most images do not have labels or tags. A topic model is such an unsupervised probabilistic method that captures latent aspects of data, where each latent aspect, or a topic, is associated with one homogeneous region. The results of topic models, however, usually have noises, which decreases the overall segmentation performance. In this paper, to improve the performance of image segmentation using topic models, we propose two topic masks applicable to topic assignments of homogeneous regions obtained from topic models. The topic masks capture the noises among the assigned topic assignments or topic labels, and remove the noises by replacements, just like image masks for pixels. However, as the nature of topic assignments is different from image pixels, the topic masks have properties that are different from the existing image masks for pixels. There are two contributions of this paper. First, the topic masks can be used to reduce the noises of topic assignments obtained from topic models for image segmentation tasks. Second, we test the effectiveness of the topic masks by applying them to segmented images obtained from the Latent Dirichlet Allocation model and the Spatial Latent Dirichlet Allocation model upon the MSRC image dataset. The empirical results show that one of the masks successfully reduces the topic noises.

• Journal of Information Processing Systems
• /
• v.15 no.5
• /
• pp.1096-1107
• /
• 2019

In this paper, we present an ontology-based approach to labeling influential topics of scientific articles. First, to look for influential topics from scientific article, topic modeling is performed, and then social network analysis is applied to the selected topic models. Abstracts of research papers related to data mining published over the 20 years from 1995 to 2015 are collected and analyzed in this research. Second, to interpret and to explain selected influential topics, the UniDM ontology is constructed from Wikipedia and serves as concept hierarchies of topic models. Our experimental results show that the subjects of data management and queries are identified in the most interrelated topic among other topics, which is followed by that of recommender systems and text mining. Also, the subjects of recommender systems and context-aware systems belong to the most influential topic, and the subject of k-nearest neighbor classifier belongs to the closest topic to other topics. The proposed framework provides a general model for interpreting topics in topic models, which plays an important role in overcoming ambiguous and arbitrary interpretation of topics in topic modeling.

• Journal of the Korean Society for information Management
• /
• v.32 no.3
• /
• pp.397-412
• /
• 2015

The main goal of this study is to investigate how to route a question to some relevant users who have interest in the topic of the question based on users' topic interest. In order to assess users' topic interest, archived question-answer pairs in the community were used to identify latent topics in the chosen categories using LDA. Then, these topic models were used to identify users' topic interest. Furthermore, the topics of newly submitted questions were analyzed using the topic models in order to recommend relevant answerers to the question. This study introduces the process of topic modeling to investigate relevant users based on their topic interest.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.1
• /
• pp.151-159
• /
• 2015

This article analyzes English abstracts of the articles published in Journal of the Korean Data & Information Science Society using text mining techniques. At first, term-document matrices are formed by various methods and then visualized by social network analysis. LDA (latent Dirichlet allocation) and CTM (correlated topic model) are also employed in order to extract topics from the abstracts. Performances of the topic models are compared via entropy for several numbers of topics and weighting methods to form term-document matrices.

• Journal of the Korean Society for Library and Information Science
• /
• v.58 no.1
• /
• pp.5-30
• /
• 2024

The purpose of this study is to extract topics from experimental data using the topic modeling methods(LDA, Top2Vec, and BERTopic) and compare the characteristics and differences between these models. The experimental data consist of 55,442 papers published in 85 academic journals in the field of library and information science, which are indexed in the Web of Science(WoS). The experimental process was as follows: The first topic modeling results were obtained using the default parameters for each model, and the second topic modeling results were obtained by setting the same optimal number of topics for each model. In the first stage of topic modeling, LDA, Top2Vec, and BERTopic models generated significantly different numbers of topics(100, 350, and 550, respectively). Top2Vec and BERTopic models seemed to divide the topics approximately three to five times more finely than the LDA model. There were substantial differences among the models in terms of the average and standard deviation of documents per topic. The LDA model assigned many documents to a relatively small number of topics, while the BERTopic model showed the opposite trend. In the second stage of topic modeling, generating the same 25 topics for all models, the Top2Vec model tended to assign more documents on average per topic and showed small deviations between topics, resulting in even distribution of the 25 topics. When comparing the creation of similar topics between models, LDA and Top2Vec models generated 18 similar topics(72%) out of 25. This high percentage suggests that the Top2Vec model is more similar to the LDA model. For a more comprehensive comparison analysis, expert evaluation is necessary to determine whether the documents assigned to each topic in the topic modeling results are thematically accurate.

• Journal of KIISE:Software and Applications
• /
• v.31 no.4
• /
• pp.488-497
• /
• 2004

This paper describes the hybrid document summarization using the indicative summarization and the query-based summarization. The learning models are built from teaming documents in order to extract topic phrases. We use Naive Bayesian, Decision Tree and Supported Vector Machine as the machine learning algorithm. The system extracts topic phrases automatically from new document based on these models and outputs the summary of the document using query-based summarization which considers the extracted topic phrases as queries and calculates the locality-based similarity of each topic phrase. We examine how the topic phrases affect the summarization and how many phrases are proper to summarization. Then, we evaluate the extracted summary by comparing with manual summary, and we also compare our summarization system with summarization mettled from MS-Word.

• Journal of the Korean Society of Industry Convergence
• /
• v.27 no.3
• /
• pp.635-644
• /
• 2024

Although the business model(BM) patents act as a creative bridge between technology and the marketplace, limited scholarly attention has been paid to the content analysis of BM patents. This study aims to contextualize converging BM patents by employing topic modeling technique and clustering highly marketable topics, which are expressed through a topic-market impact matrix. We relied on BM patent data filed between 2010 and 2022 to derive empirical insights into the commercial potential of emerging business models. Subsequently, nine topics were identified, including but not limited to "Data Analytics and Predictive Modeling" and "Mobile-Based Digital Services and Advertising." The 2x2 matrix allows to position topics based on the variables of topic growth rate and market impact, which is useful for prioritizing areas that require attention or are promising. This study differentiates itself by going beyond simple topic classification based on topic modeling, reorganizing the findings into a matrix format. T he results of this study are expected to serve as a valuable reference for companies seeking to innovate their business models and enhance their competitive positioning.

• Journal of KIISE:Software and Applications
• /
• v.34 no.12
• /
• pp.1045-1055
• /
• 2007

Latent topic models are statistical models which automatically captures salient patterns or correlation among features underlying a data collection in a probabilistic way. They are gaining an increased popularity as an effective tool in the application of automatic semantic feature extraction from text corpus, multimedia data analysis including image data, and bioinformatics. Among the important issues for the effectiveness in the application of latent topic models to the massive data set is the efficient learning of the model. The paper proposes an accelerated learning technique for PLSA model, one of the popular latent topic models, by an incremental EM algorithm instead of conventional EM algorithm. The incremental EM algorithm can be characterized by the employment of a series of partial E-steps that are performed on the corresponding subsets of the entire data collection, unlike in the conventional EM algorithm where one batch E-step is done for the whole data set. By the replacement of a single batch E-M step with a series of partial E-steps and M-steps, the inference result for the previous data subset can be directly reflected to the next inference process, which can enhance the learning speed for the entire data set. The algorithm is advantageous also in that it is guaranteed to converge to a local maximum solution and can be easily implemented just with slight modification of the existing algorithm based on the conventional EM. We present the basic application of the incremental EM algorithm to the learning of PLSA and empirically evaluate the acceleration performance with several possible data partitioning methods for the practical application. The experimental results on a real-world news data set show that the proposed approach can accomplish a meaningful enhancement of the convergence rate in the learning of latent topic model. Additionally, we present an interesting result which supports a possible synergistic effect of the combination of incremental EM algorithm with parallel computing.

• Journal of Information Processing Systems
• /
• v.16 no.2
• /
• pp.329-342
• /
• 2020

In recent years, emotional text classification is one of the essential research contents in the field of natural language processing. It has been widely used in the sentiment analysis of commodities like hotels, and other commentary corpus. This paper proposes an improved W-LDA (weighted latent Dirichlet allocation) topic model to improve the shortcomings of traditional LDA topic models. In the process of the topic of word sampling and its word distribution expectation calculation of the Gibbs of the W-LDA topic model. An average weighted value is adopted to avoid topic-related words from being submerged by high-frequency words, to improve the distinction of the topic. It further integrates the highest classification of the algorithm of support vector machine based on the extracted high-quality document-topic distribution and topic-word vectors. Finally, an efficient integration method is constructed for the analysis and extraction of emotional words, topic distribution calculations, and sentiment classification. Through tests on real teaching evaluation data and test set of public comment set, the results show that the method proposed in the paper has distinct advantages compared with other two typical algorithms in terms of subject differentiation, classification precision, and F1-measure.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.1
• /
• pp.81-98
• /
• 2013

Since Probabilistic Latent Semantic Analysis (PLSA) and Latent Dirichlet Allocation (LDA) were introduced, many revised or extended topic models have appeared. Due to the intractable likelihood of these models, training any topic model requires to use some approximation algorithm such as variational approximation, Laplace approximation, or Markov chain Monte Carlo (MCMC). Although these approximation algorithms perform well, training a topic model is still computationally expensive given the large amount of data it requires. In this paper, we propose a new method, called non-simultaneous sampling deactivation, for efficient approximation of parameters in a topic model. While each random variable is normally sampled or obtained by a single predefined burn-in period in the traditional approximation algorithms, our new method is based on the observation that the random variable nodes in one topic model have all different periods of convergence. During the iterative approximation process, the proposed method allows each random variable node to be terminated or deactivated when it is converged. Therefore, compared to the traditional approximation ways in which usually every node is deactivated concurrently, the proposed method achieves the inference efficiency in terms of time and memory. We do not propose a new approximation algorithm, but a new process applicable to the existing approximation algorithms. Through experiments, we show the time and memory efficiency of the method, and discuss about the tradeoff between the efficiency of the approximation process and the parameter consistency.