• Annual Conference of KIPS
• /
• 2012.11a
• /
• pp.396-399
• /
• 2012

Topic detection is a problem of discovering the topics of online publishing documents. For topic detection, it is important to extract correct topic words and to show the topical words easily to understand. We consider a topic tree-based approach to more effectively and more briefly show the result of topic detection for online text documents. In this paper, to achieve the topic tree-based topic detection, we propose a new term weighting method, called CTF-CDF-IDF, which is simple yet effective. Moreover, we have modified a conventional clustering method, which we call incremental k-medoids algorithm. Our experimental results with Reuters-21578 and Google news collections show that the proposed method is very useful for topic detection.

• Journal of the Korea Society of Computer and Information
• /
• v.13 no.7
• /
• pp.59-66
• /
• 2008

This paper describes a novel method for identifying the main topic and detecting topic changes in a text-based dialogue as in Instant Messaging (IM). Compared to other forms of text, dialogues are uniquely characterized with the short length of text with small number of words, two or more participants, and existence of a history that affects the current utterance. Noting the characteristics, our method detects the main topic of a dialogue by considering the keywords not only the utterances of the user but also the dialogue system's responses. Dialogue histories are also considered in the detection process to increase accuracy. For topic change detection, the similarity between the former utterance's topic and the current utterance's topic is calculated. If the similarity is smaller than a certain threshold, our system judges that the topic has been changed from the current utterance. We obtained 88.2% and 87.4% accuracy in topic detection and topic change detection, respectively.

• Journal of Multimedia Information System
• /
• v.1 no.1
• /
• pp.55-67
• /
• 2014

Image topic and emotion analysis is an important component of online image retrieval, which nowadays has become very popular in the widely growing social media community. However, due to the gaps between images and texts, there is very limited work in literature to detect one image's Topics and Emotions in a unified framework, although topics and emotions are two levels of semantics that often work together to comprehensively describe one image. In this work, a unified model, Joint Topic/Emotion Multi-Modal Hierarchical Latent Dirichlet Allocation (JTE-MMHLDA) model, which extends previous LDA, mmLDA, and JST model to capture topic and emotion information at the same time from heterogeneous data, is proposed. Specifically, a two level graphical structured model is built to realize sharing topics and emotions among the whole document collection. The experimental results on a Flickr dataset indicate that the proposed model efficiently discovers images' topics and emotions, and significantly outperform the text-only system by 4.4%, vision-only system by 18.1% in topic detection, and outperforms the text-only system by 7.1%, vision-only system by 39.7% in emotion detection.

• Journal of Information Science Theory and Practice
• /
• v.1 no.3
• /
• pp.33-46
• /
• 2013

Despite limited success, today's information retrieval (IR) systems are not intelligent or reliable. IR systems return poor search results when users formulate their information needs into incomplete or ambiguous queries (i.e., weak queries). Therefore, one of the main challenges in modern IR research is to provide consistent results across all queries by improving the performance on weak queries. However, existing IR approaches such as query expansion are not overly effective because they make little effort to analyze and exploit the meanings of the queries. Furthermore, word sense disambiguation approaches, which rely on textual context, are ineffective against weak queries that are typically short. Motivated by the demand for a robust IR system that can consistently provide highly accurate results, the proposed study implemented a novel topic detection that leveraged both the language model and structural knowledge of Wikipedia and systematically evaluated the effect of query disambiguation and topic-based retrieval approaches on TREC collections. The results not only confirm the effectiveness of the proposed topic detection and topic-based retrieval approaches but also demonstrate that query disambiguation does not improve IR as expected.

• Journal of the Korean Society for information Management
• /
• v.22 no.1 s.55
• /
• pp.191-208
• /
• 2005

This study utilizes various approaches for new topic detection in the process of assigning and updating descriptors, which is a representation method of the knowledge structure. Particularly in the case of occurring changes on the knowledge structure due to the appearance and development of new topics in specific study areas, new topic detection can be applied to solving the impossibility or limitation of the existing index terms in representing subject concepts. This study confirms that the majority of newly developing topics in information science are closely associated with each other and are simultaneously in the phase of growth and development. Also, this study shows the possibility that the use of candidate descriptor lists generated by new topic detection methods can be an effective tool in assisting indexers. In particular. the provision of candidate descriptor lists to help assignment of appropriate descriptors will contribute to the improvement of the effectiveness and accuracy of indexing.

• Journal of Korea Multimedia Society
• /
• v.22 no.8
• /
• pp.874-886
• /
• 2019

To delve into the semantic gap problem of the automatic video summarization, we focused on an endogenous ERP responses at around 400ms and 600ms after the on-set of audio-visual stimulus. Our experiment included two factors: the topic exploration of experimental conditions (Topic Given vs. Topic Exploring) as a between-subject factor and the topic relevance of the shots (Topic-Relevant vs. Topic-Irrelevant) as a within-subject factor. For the Topic Given condition of 22 subjects, 6 short historical documentaries were shown with their video titles and written summaries, while in the Topic Exploring condition of 25 subjects, they were asked instead to explore topics of the same videos with no given information. EEG data were gathered while they were watching videos in real time. It was hypothesized that the cognitive activities to explore topics of videos while watching individual shots increase the amplitude of endogenous ERP at around 600 ms after the onset of topic relevant shots. The amplitude of endogenous ERP at around 400ms after the onset of topic-irrelevant shots was hypothesized to be lower in the Topic Given condition than that in the Topic Exploring condition. The repeated measure MANOVA test revealed that two hypotheses were acceptable.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.12
• /
• pp.5905-5926
• /
• 2017

Sentiments can profoundly affect individual behavior as well as decision-making. Confronted with the ever-increasing amount of review information available online, it is desirable to provide an effective sentiment model to both detect and organize the available information to improve understanding, and to present the information in a more constructive way for consumers. This study developed a unified phrase-based topic and sentiment detection model, combined with a tracking model using incremental hierarchical dirichlet allocation (PTSM_IHDP). This model was proposed to discover the evolutionary trend of topic-based sentiments from online reviews. PTSM_IHDP model firstly assumed that each review document has been composed by a series of independent phrases, which can be represented as both topic information and sentiment information. PTSM_IHDP model secondly depended on an improved time-dependency non-parametric Bayesian model, integrating incremental hierarchical dirichlet allocation, to estimate the optimal number of topics by incrementally building an up-to-date model. To evaluate the effectiveness of our model, we tested our model on a collected dataset, and compared the result with the predictions of traditional models. The results demonstrate the effectiveness and advantages of our model compared to several state-of-the-art methods.

• Journal of the Korean Society for information Management
• /
• v.30 no.1
• /
• pp.285-302
• /
• 2013

This study identified topic shifts and patterns over time by analyzing an enormous amount of Twitter data whose characteristics are high accessibility and briefness. First, we extracted keywords for a certain product and used them for representing the topic network allows for intuitive understanding of keywords associated with topics by nodes and edges by co-word analysis. We conducted temporal analysis of term co-occurrence as well as topic modeling to examine the results of network analysis. In addition, the results of comparing topic shifts on Twitter with the corresponding retrieval results from newspapers confirm that Twitter makes immediate responses to news media and spreads the negative issues out quickly. Our findings may suggest that companies utilize the proposed technique to identify public's negative opinions as quickly as possible and to apply for the timely decision making and effective responses to their customers.

• Journal of Korean Library and Information Science Society
• /
• v.46 no.1
• /
• pp.321-344
• /
• 2015

This paper presents a fusion approach to sentiment detection that combines multiple sources of evidence to retrieve blogs that contain opinions on a specific topic. Our approach to finding opinionated blogs on topic consists of first applying traditional information retrieval methods to retrieve blogs on a given topic and then boosting the ranks of opinionated blogs based on the opinion scores computed by multiple sentiment detection methods. Our sentiment detection strategy, whose central idea is to rely on a variety of complementary evidences rather than trying to optimize the utilization of a single source of evidence, includes High Frequency module, which identifies opinions based on the frequency of opinion terms (i.e., terms that occur frequently in opinionated documents), Low Frequency module, which makes use of uncommon/rare terms (e.g., "sooo good") that express strong sentiments, IU Module, which leverages n-grams with IU (I and you) anchor terms (e.g., I believe, You will love), Wilson's lexicon module, which uses a collection-independent opinion lexicon constructed from Wilson's subjectivity terms, and Opinion Acronym module, which utilizes a small set of opinion acronyms (e.g., imho). The results of our study show that combining multiple sources of opinion evidence is an effective method for improving opinion detection performance.

• The KIPS Transactions:PartD
• /
• v.14D no.7
• /
• pp.819-828
• /
• 2007

In this paper, we propose a topic maps matching and merging approach based on the syntactic or semantic characteristics and constraints of the topic maps. Previous schema matching approaches have been developed to enhance effectiveness and generality of matching techniques. However they are inefficient because the approaches should transform input ontologies into graphs and take into account all the nodes and edges of the graphs, which ended up requiring a great amount of processing time. Now, standard languages for developing ontologies are RDF/OWL and Topic Maps. In this paper, we propose an enhanced version of matching and merging technique based on topic partitioning, several matching operations and merging conflict detection.