You are on page 1of 1

Anthropomorphic Question Answering System for the -Omic Era

Michael A. Bauer1,2, Daniel Berleant1, Robert Belford1


1University of Arkansas at Little Rock
2University of Arkansas for Medical Sciences

INTRODUCTION RESEARCH DESIGN AND METHODS EXAMPLE CASE

An example interaction with the system:


In clinical and biomedical settings researchers use specialized search The system's architecture can be divided into four subsystems: 1)The user asks the system a question that is sent to each of the agents to
engines to acquire answers to technical questions or to quickly get z Knowledge base - Create and maintain the knowledge base process in its own way.
verification of experimental results. The outcome of such queries result in z Question analyzer - Multiple software agents will find possible 2)In this case more information was needed to reduce the number of
the reading and scanning of multiple Web pages and documents. Question answers to questions. A dialog established with the user can obtain possible answers so the system asked if the user could provide any
answering (QA) is a specialized type of information retrieval with the aim of feedback to refine the query. additional terms.
returning precise short answers to queries posed as natural language z Answer identifier - From those possible answers the most relevant 3)The additional information allows the system to find an answer that is
questions. We propose a QA system that creates a dialog with the user that will be presented to the user. processed by the HyperGlossary.
mimics human interaction to answer biological questions supplemented with z Answer presenter - Answers will be automatically marked up and 4)4) The term ‘acetone’ was matched in the chosen glossary.
multimedia information. linked to semantically relevant content in other databases. The 5)A definition of acetone is presented with a Spanish translation and the 3D
additional information will be presented in a popup window that structure of the molecule.
appears when a marked term is clicked.

1
BASIC SYSTEM

4 5

KEY FEATURES

z Query Formation - This is the process of building a query. The


user has an information need and creates an initial query either in Enhanced Answers CONCLUSIONS AND IMPLICATIONS
natural language or Boolean form. The query may be modified
further through the use of the knowledge space and dialog with the
user.
z Query Weighting - In this process weights are given to the terms The proposed system addresses two current requirement gaps in QA:
in the query. The weights are often identical to the weights given to
the terms in the document. z Incorporating relevant multimedia information
z Document Representation - This is the process by which the z Ability to interact with the user
documents are converted to a secondary representation. In this
process the tokens (words and phrases) are identified and indexed. There is a lack of systems that allow the user to choose his or her
The tokens can be all the terms in a document or a subset based on a environment, establish context, have the system take that
controlled vocabulary. information into account, and automatically return the appropriate
z Term Document Weighting - Document representation tokens answer with relevant multimedia information. The QA system is
are given a value relative to their perceived level of importance. intended to return short answers to biological questions which will
z Similarity Measure - The similarity measure is used to The multi-model approach takes advantage Answers will be automatically marked up reduce the need to read entire documents.
determines similarity between the query and the document. of the fact that different models will and linked to semantically relevant content
retrieve different subsets of the relevant In other databases using the
information for any particular HyperGlossary.
query/question.
This work is supported by the NSF Division of Undergraduate Education Award number
0840830 and the IDeA Networks of Biomedical Research Excellence (INBRE) Program of the
National Center for Research Resources.

You might also like