1. Semantic Web Development in Traditional Chinese Medicine
Huajun Chen
Zhejiang Unviersity
It is my great honor to have this precious opportunity to present our work on sw technology.
Many thanks to ……

2. Semantic Web Development for Traditional Chinese Medicine
What’s TCM?
TCM Semantic Web
TCM Ontology Engineering
TCM Semantic Search Engine (DartGrid System)
Semantic Graph Mining for biomedical network analysis
Well, let’s begin with the TCM project first. I am going to do the introduction from three aspects including a TCM ontology engineering effort, a TCM semantic search engine which is developed upon the DartGrid system, and the research on semantic graph mining for biomedical network analysis, the purpose of which is to introduce the approaches from graph mining into semantic data analysis, but we do it majorly in the life science and healthcare context.

3. What’s TCM
Before moving on, I’d like to ask two questions. How many people in this room know TCM? And how many have ever tried some kinds of TCM like massage, acupuncture…, TCM herb medicine whatever.

4. What’s TCM?
Traditional Chinese Medicine (TCM) is an ancient medical system that accounts for around 40% of all health cares delivered in China.
Preventive Medicine
Take medicine as like a daily nutrition supplement or part of food to maintain the balance of the whole body system.
Personalized Medicine
Treatment can be completely different for people with respect to their gender, age, health condition although they have very similar symptoms.
Empirical Medicine
The effect of many TCM drugs are based on more one thousand years of practices, whereas they do not know the specific underlying mechanism.

5. TCM Knowledge
TCM theories derive from many knowledge sources including the theories of Yin-Yang, Chinese five elements, the human body channel system, Zang Fu organ theory, holistic connections, mind-body intervention, and many others.
TCM practice includes diagnosis and treatments theories such as herbal medicine and , massage and cupping, acupuncture and meridians.
And since TCM derive from traditional Chinese philosophy, it has a very different knowledge system. For example, the TCM theories derive from many knowledge sources including the theories of Yin-Yang, Chinese five elements, the human body channel system, Zang Fu organ theory, holistic connections, mind-body intervention, and many others.
TCM practice includes diagnosis and treatments theories such as herbal medicine and decoction, massage and cupping, acupuncture and meridians.

6. TCM Semantic Web Project
A project in collaboration with China Academy of Traditional Chinese Medicine.
Well, what do we want to do with TCM with the Semantic Web technology?
The TCM semantic web project was started around 7 years ago. It is a project in collaboration with China Academy of Traditional Chinese Medicine, which is the highest level of TCM research institute.

7. The ultimate vision of the TCM Semantic Web
TCM medical system from the perspective of modern biological medicine. This require making a connection between these two medical system at the data level first, so researcher can look into it to investigate the facts underlying it.
Figure 2 shows the architecture of the Semantic Web layer and its role in unifying and linking relational databases, to enhance the interoperability, reusability, and utilization of TCM information and knowledge resources. Particularly, We emphasize on making connections between TCM and biomedical knowledge system in purpose of facilitating cross-cultural information retrieval and data analysis. A global conceptual map of the two independent medical systems reveals potential connecting points, such as: (1) the patient at represents EMRs and Clinical Trials that record the methods and results of integrative clinical practices and
(2) the chemical that refers to bioactive compounds from Chinese Herbal Medicine, which serve for drug discovery and safety analysis. We first engineer the TCM domain ontology based on this conceptual map, and then use the ontology to glue more and more databases.

8. The Subprojects TCM Ontology Engineering (2001-current).
The DartGrid Data Integration System (first started in 2002)
Integrating legacy relational database into Semantic Web
DartMapper: Visulized relational-2-RDF Mapper ( )
DartQuery: SPARQL2SQL Query Rewriter and a Form-based SPARQL query builder ( )
DartSearch: Semantic Search. (2005-current)
Semantic Data Analysis and Data Mining for Semantic Web
DartSpora: semantic data analysis engine (2007-current)
Semantic Graph Mining for biomedical network analysis. ( current)

9. TCM Ontology Engineering

10. TCM Ontology Engineering
A effort participated by more than 100 persons from over 30 TCM research institutes located in different parts of China
Scale
More than 20,000 classes and 100,000 instances defined in the current ontology
Service Web APIs for ontology-based applications.
TCM Ontology Platform
This platform allow TCM experts from different sites (organizations, places) to edit the ontology collaboratively, and then
Provide the necessary Web services that allow the other
Domain Categorization
The current TCM ontology contains 15 major categories for each sub-domain.
Ontology Structure
A typing system as a concept hierarchy
A semantic network defining the associations between concepts
Scale
More than 20,000 classes and 100,000 instances defined in the current ontology
Access Control
layered privilege mechanism that defines users as reader, editor, checker and administrator.
Service
Web APIs for ontology-based applications.

11. The current TCM ontology contains 15 major categories for each sub-domain.
We have 15 Categories defined in the TCM Ontology corresponding to different sub-domains of TCM. Different domain experts can concentrate on their own sub-ontology.

12. Ontology visualization and query engine

13. TCM Semantic Search Engine
A semantic search engine build upon a lot of relational databases.

14. System Architecture
Search Service supports full-text search in all databases, and semantically navigating through the result, across database boundaries.
Ontology Service is used to expose the RDF/OWL ontologies.
Semantic Query Service is used to process SPARQL semantic queries.
Semantic Registration Service maintains the semantic mapping information.

15. Visualized Mapper
Visualized mapper for defining mappings from relational schema to RDF ontology.
Map relational schemas into domain ontology, and define each relational table as a Semantic View over classes defined in the ontology. The database administrators use DartMapping to define semantic mapping between relational schema and ontology, and to specify table-to-class mappings in a drag-and-drop fashion aided by the intelligent agent that reasons on the database schema. This paradigm can improve efficiency and reduce errors compared with editing configuration files.
User Interface Overview
If you have a relational database, it is very easy to use this tool to publish it as RDF store. Configure the connection to the database using this tool, import an semantic ontology that may be published on the web somewhere, define a mapping, and then deploy it using the linking data deployment functionality. Each table is defined as a view over the ontology: Drag and Drop table/column and ontology/class from the resource tree to the central panel.
Support one-table-2-multiple-class mapping; Semantics for Join relations can be configured; URI function.
Backend Query-Rewriting Engine
At the backend of the data integration system, is a View-based query rewriting server.Translate a Sparql query into a series of SQL queries against underlying relational schemas, based on mapping rules derived from semantic views.
Use Cases
A full list of mapping cases that have been considered can be found at:

16. Semantic Search Portal Version 1
Background
we have developed a semantic-based query and search portal to assist in user interaction with the system. Basically, this system consists of two components. The search component enables users to perform full-text searching through all of the integrated data sources using keywords that is similar to common Internet search engines, while the query component supplies with a means for handling more complex semantic queries posed against the semantic web ontology.
Semantic Search
In response to a keyword search, the portal displays content entries in the middle, and a recommended list of synonyms and associated concepts on the right. Disparate pieces of data about the same concept/topic are synergized.
Semantic Query
Using the classes and concepts in the visualized ontology or the searching results, the user can start to specify an accurate semantic query in an interactive query customization environment.
Semantic Navigation
Based on the semantic associations defined at the ontological level, the user can keep navigating information and knowledge unaware of database boundaries.
The Role of Ontology
The ontology plays an important role in the mediation of the query, search and navigation. At first, it serves as a logic layer for users in constructing semantic queries. The form-based query interface is automatically generated based on the ontological structure. The constructed semantic query will then be translated into SQL queries based on the mapping rules generated by the mapping tool. At second, it enables semantic navigation across database boundaries during query and search. At third, it also serves as a control vocabulary to facilitate search by making semantic suggestions such as synonyms, and related concepts.

17. Semantic Search Portal Version 1

18. Semantic Data Analysis for TCM
What kinds of new connections can be discovered or mined from this huge web of data?

19. Graph vs Semantic Graph
Conventional Graph Model
Semantic Graph Model
Node
All nodes are identical
Nodes are labeled, different
Edge
All edges are identical
Edges are labeled, different
Basic
Element
Nodes stand for entities
RDF statement stands for facts.
Semantic Graph as a Knowledge Base
Semantic Graph as a Complex Network
Reasoning
Network Analysis
Biological network analysis is built upon many good graph models such as random graph model, scale free graph model. However, the semantic web provide us with a very brand-new graph. It is a semantic graph. Here I make a list of comparsons between conventional graph and the semantic graph.
Semantic Graph Mining

20. An example.
A semantic graph can connect data from different sources and domains while preserving the provenance of data.

21. An example Frequent Semantic Sub-graph Discovery Problem Descriptions:
Semantic Sub-graph. In a semantic graph G, every transaction can be represented as a knowledge base consisting of statements. One graph A is a sub-graph of graph B iff. A is subsumed by B.
Frequent Semantic Sub-Graph. Give a graph g, and a semantic graph G. g is a frequent sub-graph with respect to G, iff. there are more than i|K| minimum subsumed sub-graphs in G with respect to g, where i is a user- specified minimum support threshold, and |K| is the total number of graphs in K.
Applicatisions:
Network motifs identification in biological networks
Drug Efficacy Analysis
This is a example of the study of frequent patterns of compositional use of herbs in TCM.
SGM can help indentify frequently appeared knowledge from a number of TCM formulas, help reveal the underlying

22. Semantic data analysis
Semantic graph contains richer information than normal graph.
It is based upon the integration capability of semantic web.
Much more meaningful mining results:
Discover the facts directly.
Find more meaningful associations among entities.
Calculate the network parameters in a more accurate way.
Ontological reasoning can be leveraged to further facilitate the mining process.
We need good tools to help do so.

23. DartSpora: a interactive mining engine for TCM
DartSpora: Interactive Mining for TCM Knowledge Discovery
The Spora system facilitates users to perform knowledge discovery experiments based on semantic graph mining.
Background
In TCM domain, knowledge discovery experiments often involves various interactions between software tools, datasets, and researchers, which requires datasets and algorithms from inter-related domains to be integrated and re-used on demand in an ad-hoc manner.
User Interface
In this Ajax-based interface, The Spora system facilitates TCM domain experts to perform knowledge discovery experiments in an intuitive manner. Users can create an Experiment by specifying a knowledge discovery process as a tree of operators with customizable properties, and then execute the process and review the visualized results. The user interface contains four units: (1) Login, Introduction, and Navigation panels; (2) Explorer panel that manages Semantic Web resources(including experiments and
operators) that is visible and controllable by users; (3) Operator and Operator Parameter Panels that display operators and their customizable parameters; (4) Result panel that visualizes data mining results through interactive tables, histograms, etc.
Use Cases
Rule-based Text/Data Preprocessing
TCM Formulae System Analysis

24. Summary A Web of Data means a lot to us.
It can enable fancy ways of searching and browsing the daunting online information space.
It can also finally unleash the potential underlying disparate data sources to greatly facilitate and advance the data mining and knowledge discovery technology.
But we need powerful tools to help us to achieve the goal.

25. Summary: Key Benefits of Semantic Web for TCM
Fusion of data across many scientific discipline
Easier recombination of data
Querying of data at different levels of granularity
Capture provenance of data through annotation
Data can be assessed for inconsistencies
Integrative knowledge discovery from large-scale semantic graph formed by integrating cross-institutional, cross-dispinaries data sources.

26. Thanks for your time!