Yong-Jin Han

A natural language interface of thorough coverage by concordance with knowledge bases

One of the critical problems in natural language interfaces is the discordance between the expressions covered by the interface and those by the knowledge base. In the graph-based knowledge base such as an ontology, all possible queries can be prepared in advance. As a solution of the discordance problem in natural language interfaces, this paper proposes a method that translates a natural language query into a formal language query such as SPARQL. In this paper, a user query is translated into a formal language by choosing the most appropriate query from the prepared queries. The experimental results show a high accuracy and coverage for the given knowledge base.

Experience Search: Accessing the Emergent Knowledge from Annotated Blog Postings

Emergent knowledge does not come from a particular document or a particular knowledge source, but comes from a collection of documents or knowledge sources. This paper proposes a system which combines the social web contents and the semantic web technology to process the emergent knowledge from the blogosphere. The proposed system regards blog postings as experiences of people on particular topics. By annotating postings in the selected domains with ontology vocabularies, the system collects experiences from various people into an ontology about people and experiences. The system processes this ontology with semantic rules to find the emergent knowledge. Users can access previously unavailable facts, concepts and trends which are emerging from system.

Ranking entities similar to an entity for a given relationship

This paper proposes a similarity ranking method for entities in the real world. Real world entities like people or objects often have some relationship between themselves. Finding such relationships from real world data can greatly enhance recognition of real world situations. However, it is difficult to capture such relationships from real world sensors alone. Nowadays, activities of people are often shared via Web. The activities can be represented as a relationship between people with shared items such as books, movies or other items. In semantic Web research, such relational information has been modeled in ontologies. The proposed ranking method of this paper is a method that finds meaningful relationships between entities in ontologies. In the first step, the method discovers pairs of entities which have meaningful connections in an ontology. Then it ranks the pairs according to similarities between entities. Unlike previous work, the proposed method assumes not only instance level connections, but also ontology schema level connections. This approach enables machines to access previously hidden indirect relationships into the similarity rankings. The experiments using an existing people-experience ontology show that the proposed method outperforms previous methods.

Time Variant Event Ontology for Temporal People Information

The people information is distributed in various forms such as database, web page, text, and so on, where the world wide web is one of the main sources of publicly-available people information. It has a characteristic that the information on people is intrinsically temporal. Therefore, the reconstruction of the information is needed for an individual or a company to use it efficiently. In order to maintain or manage the temporal people information, it must distinguish the variable information from invariable information of people. In this paper, we propose a method that constructs an ontology based on events to manage the variable people information efficiently. In addition, we present a system which reconstructs people information that satisfies the users’ demand with the ontology.

A single-directional influence topic model using call and proximity logs simultaneously

Understanding social interactions is one of the key factors in the development of context-aware ubiquitous applications. Identifying interaction patterns sensed by a mobile device is one possible way for understanding social interactions. Most previous studies on this problem have employed call and proximity logs to represent social interactions. Because these interactions can be characterized by topics, the studies have applied topic models based on latent Dirichlet allocation (LDA) to identifying interaction patterns from social interactions. However, these previous studies regarded calls and proximities as independent interaction types. As a result, they lost the information obtainable when calls and proximities were analyzed simultaneously. This paper proposes a topic-based method that simultaneously considers calls and proximities, allowing interaction patterns to be identified from a mobile log. For this purpose, the proposed method regards calls and proximities as a homogeneous information type that are drawn from the same temporal space expressed by the same distribution, but with different parameters. From the observation that the number of proximities in a mobile log usually overwhelms that of calls and the proximities are observed regularly, the proposed method models a single-directional influence from proximities to calls, where both call and proximity are modeled by LDA. The experiments with three different data sets from the Massachusetts Institute of Technology’s Reality Mining project show that the proposed method outperforms the method that considers calls and proximities independently; this proves the plausibility of the proposed method.

Processing of Korean Natural Language Queries Using Local Grammars

For casual web users, a natural language is more accessible than formal query languages. However, understanding of a natural language query is not trivial for computer systems. This paper proposes a method to parse and understand Korean natural language queries with local grammars. A local grammar is a formalism that can model syntactic structures and synonymous phrases. With local grammars, the system directly extracts users intentions from natural language queries. With 163 hand-crafted local grammar graphs, the system could attain a good level of accuracy and meaningful coverage over IT company/people domain.

Finding social interaction patterns using call and proximity logs simultaneously

This paper proposes a topic-based method to reflect calls and proximities simultaneously into finding interaction patterns from a mobile log. For this purpose, the proposed method regards calls and proximities as a homogeneous information type that are drawn from the same temporal space expressed by the same distribution, but with different parameters. The number of proximities in a mobile log usually overwhelms that of calls and the proximities are observed regularly. Therefore, the proposed method models a single directional influence from proximities to calls, where both call and proximity are modeled by the Latent Dirichlet Allocation (LDA). According to the experiments on the data set from MITs Reality Mining project, the proposed method outperforms the method that treats calls and proximities independently, which proves the plausibility of the proposed method.

Detection of Error-Prone Cases for Word Sense Disambiguation

Word sense disambiguation (WSD) is essential for natural language understanding applications such as machine translation, question & answering, and natural language interface, since the performance of such applications depends on the senses of lexicons. Thus, it is natural to consider lexicons as the most crucial features in WSD. However, due to the extensiveness of lexical space, WSD methods based on machine learning techniques with lexical features suffer from the sparse data problem. To tackle this problem, this paper proposes a hybrid approach which separately copes with an error-prone data due to sparsity. A data is regarded as error-prone if its nearest neighbors are relatively distant and their senses are uniformly distributed. Then, our hybrid approach focuses on such an error-prone data without tuning of a base method. In the experiments, the k-nearest neighbor method is used as a base method. If a data is determined as an error-prone case, it is processed by a prototype based method. The prototype based method takes an advantage from overall training examples rather than depends on only several neighbors. The experimental results on Senseval-3 nouns show that an error-prone data is effectively detected by the proposed method and our hybrid approach outperforms the ordinary k-nearest neighbor method and the prototype based method.

Determining Gender of Korean Names with Context

Machine translation systems have various problems although they have been developed continuously. Especially, in Korean-English translation system, zero pronoun problem is an important problem, since omitted subject or object Korean are must be restored in English. In order to solve this problem, various methods have been proposed. In this paper, we focus on the gender determination problem in Korean names as a first-step for solving a zero pronoun problem in Korean. Since this problem can be viewed as a binary classification problem, we adopt support vector machines which are well-known for solving binary classification. The bag-of-words model is used to represent a name with context as a vector and information entropy of words is adopted for selecting features. An evaluation of the proposed method shows about 86% of accuracy. This method achieves higher accuracy than baseline which determines the gender of a name by its majority and additionally resolves the limitation of memory based and statistical method which use only names.

Personalized app recommendation using spatio-temporal app usage log

This paper proposes a probabilistic method to recommend apps appropriate to current time and location of a user. The proposed method regards an app as a distribution of topics discovered from a large number of app descriptions. A user preference is then modeled, using spatio-temporal app usage log of a user, as a topic distribution that is affected by time and location. Since time and location can be regarded as two continuous random variables that are independent of each other, the proposed method is in contrast to conventional methods in that the conventional methods are based on a limited number of discrete contexts and assume that locations are dependent on time. Therefore, the proposed method captures user-specific contexts and is robust even in unseen time and location. Our experiments show that the proposed method outperforms two baseline methods in NDCG, which implies that the proposed method is effective in personalized app recommendation. An app recommendation method at a users current time and location is proposed.Time and location are regarded as continuous and independent random variables.The proposed method is robust even in unseen time and location.