Wonsung Lee

Efficient extraction of domain specific sentiment lexicon with active learning

Proposed a graphical model to extract a sentiment lexicon with document annotations.Applied an active learning to extract a sentiment lexicon to reduce the annotation.Suggested and experimented four distinct initialization methods for active learners.Proposed lexicon coverage analysis algorithm to initialize the active learner. Recent research indicates that a sentiment lexicon focusing on a specific domain leads to better sentiment analyses compared to a general-purpose sentiment lexicon, such as SentiWordNet. In spite of this potential improvement, the cost of building a domain-specific sentiment lexicon hinders its wider and more practical applications. To compensate for this difficulty, we propose extracting a sentiment lexicon from a domain-specific corpus by annotating an intelligently selected subset of documents in the corpus. Specifically, the subset is selected by an active learner with initializations from diverse text analytics, i.e. latent Dirichlet allocation and our proposed lexicon coverage algorithm. This active learning produces a better domain-specific sentiment lexicon which results in a higher accuracy of the sentiment classification. Subsequently, we evaluate extracted sentiment lexicons by observing (1) the increased F1 measure in sentiment classifications and (2) the increased similarity to the sentiment lexicon with the full annotation. We expect that this contribution will enable more accurate sentiment classification by domain-specific sentiment lexicons with less sentiment tagging efforts.

Augmented Variational Autoencoders for Collaborative Filtering with Auxiliary Information

Recommender systems offer critical services in the age of mass information. A good recommender system selects a certain item for a specific user by recognizing why the user might like the item. This awareness implies that the system should model the background of the items and the users. This background modeling for recommendation is tackled through the various models of collaborative filtering with auxiliary information. This paper presents variational approaches for collaborative filtering to deal with auxiliary information. The proposed methods encompass variational autoencoders through augmenting structures to model the auxiliary information and to model the implicit user feedback. This augmentation includes the ladder network and the generative adversarial network to extract the low-dimensional representations influenced by the auxiliary information. These two augmentations are the first trial in the venue of the variational autoencoders, and we demonstrate their significant improvement on the performances in the applications of the collaborative filtering.

Supervised Dynamic Topic Models for Associative Topic Extraction with A Numerical Time Series

A series of events generates multiple types of time series data, such as numeric and text data over time, and the variations of the data types capture the events from different angles. This paper aims to integrate the analyses on such numerical and text time-series data influenced by common events with a single model to better understand the events. Specifically, we present a topic model, called a supervised dynamic topic model (sDTM), which finds topics guided by a numerical time series. We applied sDTM to financial indexes and financial news articles. First, sDTM identifies topics associated with the characteristics of time-series data from the multiple types of data. Second, sDTM predicts numerical time-series data with a higher level of accuracy than does the iterative model, which is supported by lower mean squared errors.

Consumers Perceptions on Sodium Saccharin in Social Media

ABSTRACT - The purpose of this study was to investigate consumers’ perceptions of sodium saccharin in socialmedia. Data was collected from Naver blogs and Naver web communities (Korean representative portal web-site), andmedia reports including comment sections on a Yonhap news website (Korean largest news agency). The results fromNaver blogs and Naver web communities showed that it was primarily mentioned ‘sodium saccharin-no added’ prod-ucts, properties of sodium saccharin, and methods of reducing sodium saccharin in food. When media reported theexpansion of food categories permitted to use sodium saccharin, search volume for sodium saccharin has increased inboth PC and mobile search engines. Also, it was mainly commented about distrust of government, criticism of foodproduct price, and distrust of food companies below the news on the news site. The label of sodium saccharin-noadded products in market emphasized “no added-sodium saccharin”. These results suggest that consumers are inter-ested in sodium saccharin and especially when media reported the expansion of food categories permitted to use it.Consumers were able to search various information on sodium saccharin except safety or acceptable daily intakethrough social media. Therefore media or competent authority should report item on sodium saccharin with informa-tion including safety or acceptable daily intake based on scientific background and reference or experts’ interview forconsumers to get reliable information.Key words : sodium saccharin, social media, consumer perception, contents analysis, text mining

Associative topic models with numerical time series

Introduce a probabilistic graphical model extracting topics with numerical guidance.Enhance the regression performance with unified PGM of text and numbers.Tightly links the analysis on numeric and text data over time. A series of events generates multiple types of time series data, such as numeric and text data over time, and the variations of the data types capture the events from different angles. This paper aims to integrate the analyses on such numerical and text time-series data influenced by common events with a single model to better understand the events. Specifically, we present a topic model, called an associative topic model (ATM), which finds the soft cluster of time-series text data guided by time-series numerical value. The identified clusters are represented as word distributions per clusters, and these word distributions indicate what the corresponding events were. We applied ATM to financial indexes and president approval rates. First, ATM identifies topics associated with the characteristics of time-series data from the multiple types of data. Second, ATM predicts numerical time-series data with a higher level of accuracy than does the iterative model, which is supported by lower mean squared errors.

Disease-medicine topic model for prescription record mining

Analyzing patient records is important for improving the quality of medical services and for understanding each patients historical diseases. However, the huge size of the data requires statistical analysis procedures. In this paper, we proposed a probabilistic model-the disease-medicine topic model (DMTM)-to explore connected knowledge about diseases and medicines. In the model, diseases and medicines are modeled using generative process. We used the latent Dirichlet allocation, which is one of the most popular topic models, as the baseline model. Then, we compared the qualities of topic representations quantitatively and qualitatively. The comparison results showed that the topics derived from the DMTM are clearer to identify and that specific patterns were found in the diseases and medicines. In the case of topic network analysis, these specific patterns were proved using centrality measurements.