• Theme

The theme is to exploit the recent progress about conceptual modeling theory, principles, and applications required in conceptualizing, managing, and building Big Data and AI-based Data Design. Applications include autonomous activity recognition and ambient assisted living in a smart life, context awareness using big data gathered from various sensors, and this will be useful for not only physical world but also future cyber society.

PROGRAM(The 2nd International Workshop on Conceptual Modeling 2021)

• 17 January 2021 [Korea Standard Time (Sunday) 09:00, CST 08:00, EST (Sat.) Sat.19:00]

Session I (Chair: Prof. Il-Yeol Song, Drexel Univ., USA)

09:00 - 09:50

Invited talk1: Community Discovery over Big Graphs

Chaokun Wang (Tsinghua University, China)

09:50 - 10:10

Features Conceptual Modeling for Big Transportation Analytics Towards Smart City

Carson K.S. Leung (University of Manitoba, Canada)

10:10 - 10:30

Recent Trends in Context-aware Emotion Recognition

Sudarshan Pant, Guee sang Lee, Soo hyung Kim, Hyung-Jeong Yang (Chonnam National University, Korea)

10:30 - 10:50

An OpenMP-based Parallel Execution of NNEF

Nakhoon Baek (Kyongpook National University, Korea)

10:50 - 11:00

Coffee break

Session II (Chair: Prof. Wookey Lee, Inha Univ., Korea)

11:00 - 11:50

Invited talk2: Towards Effective Temporal RDB Keyword Search Using ORA-Semantics

Tok Wang Ling (National University of Singapore, Singapore)

11:50 - 12:40

Invited talk3: A Software Architecture Aspect Modeling for Microservice Development

Eunmi Choi (Kookmin University, Korea)

• Paper Submission

Papers can be either regular or short papers, where regular papers are limited to 8 pages and short ones to 4 pages. CMComp2020 adopts a single-blind review policy.

You can submit your papers through the easychair site: https://easychair.org/conferences/?conf=cmodel2021

The paper templates with formatting instructions:

IEEE Conference Microsoft Word Template: IEEE_MSWord_Template.doc

IEEE Conference Latex Template: IEEE_Latex_Template.zip

• Organizers

Tok Wang Ling (National University of Singapore, Singapore)

Il-Yeol Song (Drexel University, USA )

Wookey Lee (Inha University, Korea): ask questions to wookeylee@gmail.com

• Committee (Tentative)

James Geller (NJIT, USA)

Junseok Hong (Kyeonggi, Korea)

Changi Jeong (Ajou University, Korea)

Seokho Kang (Sungkyunkwan Unversity, Korea)

Young-Kuk Kim (Chungnam National University, Korea)

Sungbum Kim (Korea Unversity, Korea)

Carson Leung (University of Manitoba, Canada)

Suan Lee (AI Research Institute, Inha Univ., Korea)

Jooyeon Lee (Ajou University, Korea)

Sunghwa Lim (Namseoul University, Korea)

Mukesh Mohania (IBM India Research Laboratory, India)

Joo-Seok Park (Kyung-Hee University, Korea)

Mye Son (Sungkyunkwan Unversity, Korea)

Chaokun Wang (Tsinghua University, China)

Robert Wrembel (Poznan University of Technology, Poland)

Invited talk I:

Prof. Tok-Wang Ling (National University of Singapore)
Abstract:

Keyword search over non-temporal relational databases has been widely studied in recent years. Relational keyword search can be broadly classified into two categories: (1) data graph approach and (2) schema graph approach. In data (or schema) graph approach, the RDB is modeled as a graph where each node represents a tuple (or relation, resp.), and each edge represents a foreign key-key reference. For data graph approach, an answer of a keyword query is typically defined as a minimal connected data subgraph which contains all the keywords. For schema graph approach, it translates a keyword query into a set of SQL statements that join the relations in the schema graph with tuples matching the keywords. Research in these two approaches have been focused on efficient computation of keyword query results and strategies to rank the answers and output. However, the challenge to retrieve the intended keyword query results remains.

We present the serious limitations and errors of existing RDB keyword search methods which were identified by our earlier work, such as answers depend on the normal form of the RDB, incomplete answers, meaningless answers, etc. The main reason for causing these serious limitations is because RDB schema cannot capture the semantics of objects and relationships (ORA-semantics) in the RDB. We explain how ORA-semantics can be used to improve the effectiveness of RDB keyword search. We classify the relations in a RDB into object relations, relationship relations, mixed relations, and their component relations which contain multivalued attributes of object classes or relationship types. We refer to these semantics as Object-Relationship-Mixed (ORM) semantics. We construct Object-Relationship-Mixed (ORM) data (or schema) graph where each node represents a tuple of an object/relationship/mixed relation and its component relations (or an ORM relation, resp.), and each edge represents a foreign key-key reference. RDB keyword search methods which use the ORM data (or schema, resp.) graph are proposed to solve the serious problems of other existing RDB keyword search methods.

Many applications such as finance and health-care, need temporal databases to manage and historical data that changes over time. However, constructing correct SQL programs with time conditions as well as temporal operators are error-prone for users. Existing works on temporary keyword search did not consider ORA-semantics of the temporal databases which lead to the inability of database to store intended temporal and non-temporal ORA semantics in the databases of the applications, and the inability of the keyword query processing to capture the possible query interpretations when the time condition is applied to different ORA-semantics.

We extend our works in keyword search in non-temporal RDBs to temporal RDBs. For temporal database schema design, we propose a framework to help users to generate temporal database schemas from a traditional ER diagram. Our temporal database schema supports both temporal and non-temporal ORA semantics with minimum data redundancy, and avoids updating anomalies.

For temporal keyword query processing, we propose solutions to process the time condition, negation, aggregate and group-by in temporal keyword query. In all the solutions, we exploit an ORM schema graph to capture the temporal and non-temporal ORA semantics of the temporal database, which facilitates identifying objects and relationships involved in query and whether they are temporal or not.

When handling the time condition in temporal keyword query, we first identify all the temporal ORA-semantics involved in query, and apply the time condition to all or each temporal semantics to generate a set of temporal constraints. Each temporal constraint represents one possible interpretation of the time condition. Temporal join operators are used in this process to guarantee correct search results. A prototype system is designed to support interactive keyword search with a two-level ranking scheme, which allows users to choose required query interpretation step-by-step.

When handling the negation in temporal keyword query, we differentiate the cases when the negation in query is applied to temporal or non-temporal ORA semantics in the database. Depending on whether the negation should be tested on single tuple or multiple tuples, and whether the negation is applied on temporal semantics or not, we interpret the negation as a logical NOT, an anti-join operator or a temporal anti-join operator, in order to guarantee correct search results.

When handling the aggregates and group-by in temporal keyword query, we observed that the data redundancy in an intermediate joined relation would lead to incorrect temporal aggregate results. We utilize the ORA semantics to identify the unique objects/relationships in the intermediate relation and remove the data duplicates, and compute the correct answers.