Contents

• Introduction

  • SPECIALIST Lexical System

  • LexBuild

• Motivation

• Objectives

• Life Cycle Development

  • System Architecture

  • LexCheck - Grammar Specification

• Status

• Conclusion

• Future Work

• Demo

Introduction - SPECIALIST Lexical Systems

• The SPECIALIST Lexicon
  • The SPECIALIST lexicon provides comprehensive coverage of biomedical terms as well as English terms
  • The SPECIALIST lexicon contains a range of linguistic knowledge, which includes syntactic categorization, variant forms, and specifications of acronyms
  • Linguistic knowledge is fundamental to text processing

• The SPECIALIST Tool Set
  • Includes the SPECIALIST Lexical Tools, text tools, spelling resource SubTerm Mapping Tools (STMT) Visual Tagging Tools (VTT)
  • This toolset dramatically furthers methods for text processing

• Provides a foundation for the NLM's research programs in Natural Language System (NLS).

• Is an unique resource to the computational linguistics and information retrieval community

• It's importance has grown in parallel with the increased demand for improved access to textual information in the biomedical domain

Back - Details

---

Introduction - LexBuild

• The SPECIALIST lexicon is an important fundamental resource for NLP

• LexBuild is an extremely important and useful tool for building the SPECIALIST lexicon at NLM

• Ease lexical entry

• Ensure quality and completeness of lexical records

• Increase lexical coverage

Back - Details

---

Motivation

An old version of LexBuild was used. However, there are some issues:

• Maintenance issue:
  • Legacy system
  • No documents
  • Not a completed system
  • Hard to fix bugs and enhance features
  • Hard to upgrade

• Efficiency issue:
  • Distributed system
  • Time consuming to compile files and synchronize data

• Other issues:
  • Access issue
  • Portable issue
  • Quality issue
  • Backup issue

Back

---

Objective

• Replace the old version of LexBuild

• Preserve features:
  • Build the SPECIALIST lexicon
  • Generate annual release of lexicon tables
  • Provide access to information of lexicon

• Improved system:
  • A centralized web-based system
  • Fully documented life-cycle design system

Back - Details

---

Life Cycle Development

• Requirements capturing

• Analysis & design

• Implementation

• QA & testing

• Deployment

• Maintenance

Back - Details

---

System Architecture

• Three types of user: user, manager, administrator

• Tomcat, JSP, web application

• MySql database

• LexCheck

Back - Details

---

LexCheck - Grammar Specification

• A well structured & defined syntax

• A Java API package

• Lexical entry: Java object, Xml, and text format

• Generate inflectional variants

• Example:

  {base=dog
  entry=E0023681
  	cat=noun
  	variants=reg
  	variants=uncount
  }
  

Back - Details

---

System Status

• Completed final test on January, 2004

• Officially deployed on January 29, 2004

• User friendly GUI, intuit logic design

• Easy transaction, short learning curve

• Machine generated records with grammar check to ensure quality

• Capable of loading batch files (March, 2004)

• Provides more flexibilities to edit/add lexical records

• Resolve data synchronization issue

• Fixed 49 SCRs after system deployed

• Only one down time in 3 months (due to network security problem)

• Performed daily backup

• Generated lexicon files daily

• Successfully generated lexicon 2004

Back

---

Conclusion

• LexBuild works great !

• Resolved maintenance issue, efficiency issue, access issue, etc.

Back

---

Future Work

• More SCRs in SCRT to be completed

• Implement post processing programs to fix pre-existed records

• Integrate with LexAccess

• Develop automatic processing tool to expand lexicon to incorporate vocabulary from genomics and proteomics domains

• Release subset of LexBuild features and specification to assist individuals in creating their own standardized lexicon

Back

---

LexBuild Web Site

---