An initial literature review revealed:
• Vaccine Information Statements is informally read and comprehended by patients. [1] [2]
• Interest and efforts have been made to improve the delivery of vaccine education. [1] [3] [4] • Internet misinformation campaign have had significant influence on attitudes towards vaccines. [5] [6] We introduce a prototype ontology-driven application system (Vaccine Helmsman) which employs a semantic knowledge-base, that harness the use of natural language interface of a mobile client for patients to query vaccine knowledge. This prototype intends to improve patient education of vaccines as an on-demand, ubiquitous aid to provide accurate and timely information about vaccinations.
With the development of this prototype mobile assistant, we plan on accomplishing the following initial aims.
• Specific Aim 1. Create the Vaccine Information Statement Ontology (VISO).
• Specific Aim 2. Vaccine Helmsman: Develop the prototype application with client and server components.
The Vaccine Helmsman system employs a client-server model -manifested by the mobile device and server application hosted on the authors' development server. Figure 1 shows an overview of the system and the communication between the mobile application and the server components. The UML diagram depicts the mobile application sending a SPARQL [7] query request to the application server after the natural language interface (NLI) retrieves the user's command and sends it to the translator component for query construction.
The server responds to the request and preforms the query on a published OWL2 [8] file. The server applications returns back a JSON [9] data object to the mobile application. The mobile application parses the data object with the translator component, and the NLI component constructs a semi-structured sentence to speak back to the user with an answer for the user query.
Based on the architecture design described above, we have developed a working prototype that both demonstrates the feasibility of the project and highlights some potential direction to continue to enhance this system. The prototype iOS application was developed in Objective-C (See Figure 2 ) that accepts a defined set of voice commands and sends the command to a remote application web server . The server application, a Jena Fuseki [10] install, is responsible for the queries on the published VISO OWL file and responds with a JSON object. The iOS application then constructs semistructured sentences from JSON to speak back to the user. The application also displays the text of the speech along with the returned JSON data for debugging purposes.
Also, the authors created an initial draft of the vaccine information ontology based on the CDC's Vaccine Information Statements, identified as the Vaccine Information Statement Ontology (VISO). VISO, since it is an initial draft at the time of development, only models the Rotavirus and Hepatitis B vaccine information from their respective VIS documents. VISO was serialized in OWL2 using Protégé [11].
Because it is still in the beginning stages, limitations exist with the system, which we will address as development continues. Table I outlines the limitations identified, but also describes future solutions to address them and the intended outcome. One factor is the finite vocabulary set of the system, specifically with the VISO knowledge-base, which results in semi-structured speech and limited vocabulary for query recognition. We plan on addressing this issue by utilizing Simple Knowledge Organization Systems (SKOS) [12] that will enable an ontology to add synonymous terms. SKOS may also open possibilities to utilize multi-lingual capabilities to the system. Another issue relating to VISO is the limited number of sources to generate the ontology. Only two VIS documents for Rotavirus and Hepatitis B were used for sampling, resulting in a smaller-than-ideal knowledgebase. Work is underway to improve the VISO knowledgebase by adding additional information from more VIS documents and existing ontologies, and at the time of writing, the VISO design has matured and has incorporated additional VIS knowledge. Various technical constraints exist, which is expected from a prototype. This ranges from restrictions of an offline speech recognition, latency issues relating to a clientserver architecture, undiscovered usability issues, and so on. Upcoming development will explore other options to improve some of these limitations, including generating a methodology to evaluate the user experience.
Muhammad "Tuan" Amith, M.S., Cui Tao, Ph.D.
An initial framework for an ontology-based mobile application for vaccine information retrieval and education System Architecture Background Future Direction
Since the early 90s, healthcare providers have been mandated to provide VIS (Vaccine Information Statement) from the Centers for Disease Control and Prevention (CDC) to parents and patients before their children or themselves receive any vaccination uptake. Despite the initiative, there exist issues of patients not acquiring a comprehensive understanding about the vaccines and some evidence of doubt of the safety of vaccines. In addition, a significant number of patients find vaccine information on the Internet, which may inevitably influence perceptions of vaccines. This paper introduces Vaccine Helmsman, an initial prototype of a mobile client that allows for natural language querying of semantically-driven knowledge-base of vaccine information for patients.
A literature review revealed that CDC-mandated VIS are seldom read or understood [1][2] , and a need for better delivery of vaccine education [1][3] [4] . Additionally, misinformation campaigns have influenced attitudes and decisions for uptake [5][6] [7] .
The authors introduces a prototype ontologydriven application system (Vaccine Helmsman) which employs semantic knowledge base, permitting the use of natural language interfaces for patients to query vaccine knowledge. The prototype aims to address the vaccine education gap.
The authors have developed the Vaccine Information Statement Ontology (VISO) that only models Rotavirus and Hepatitis B vaccine information, represented in OWL [8] , using Protégé [9] . Later development will include a corpus of vaccine knowledge from CDC's VIS.
In conjunction with the creation of VISO ontology, the authors have developed a prototype iOS application in Objective-C that accepts defined set of voice commands and sends the command to a remote Java web server. The server application, a Jena Fuseki [10] install, queries the published VISO OWL file and responds with a JSON [11] object. The app then constructs semistructured sentences from JSON to speak back to the user.
Because it is still in the beginning stages, limitations exist with the system, which we will address as development continues. One factor is the finite vocabulary set of the system, specifically with the VISO knowledge-base, which results in semistructured speech and limited vocabulary for query recognition. We plan on addressing this issue by utilizing Simple Knowledge Organization Systems (SKOS) [12] that will enable an ontology to add synonymous terms. SKOS may also open possibilities to utilize multilingual capabilities to the system. Another issue relating to VISO is the limited number of sources to generate the ontology. Only two VIS documents for Rotavirus and Hepatitis B were used for sampling, resulting in a smaller-than-ideal knowledgebase. Work is underway to improve the VISO knowledge-base by adding additional information from more VIS documents and existing ontologies, and at the time of writing, the VISO design has matured and has incorporated additional VIS knowledge. Various technical constraints exist, which is expected from a prototype. This ranges from restrictions of an offline speech recognition, latency issues relating to a clientserver architecture, undiscovered usability issues, and so on. Upcoming development will explore other options to improve some of these limitations, including generating a methodology to evaluate the user experience.
Plan Outcome
For more information on this poster, please contact Tuan Amith, MS at muhammad.f.amith@uth.tmc.edu, Cui Tao, PhD at Cui.Tao@uth.tmc.edu