An Empirical Study of Chinese Name matching and Applications

An Empirical Study of Chinese Name matching and Applications

1. Introduction

   • Name matching
     • Important
       • Downstream tasks
         • Entity linking
           • Includes context of mentions
         • Entity clustering
           • Includes context of mentions
         • ?
         • Entity coreference
           • ?
         • Name transliteration
         • Identifying names for mining paraphrases
           • ?
         • Standalone name matching
           • Context independen
       • Entity disambiguation
         • Determine if two mentioned strings refer to the same entity
     • Methods
       • Language type
         • Alphabetic languages
           • Focused on
             • Example
               • English
             • Indo-European
         • Logogram languages
           • Example
             • Chinese
               • Hanzi
                 • Challenge
                   • A small number of hanzi represents an entire name
                   • There are X*10.000 hanzi in use
                   • Current methods
                     • Largely UNTESTED
                     • Coreference resolution errors
                       • Caused by
                         • Chinese name matching errors
       • Focus on persons names
     • Challenge
       • Issue: Name variations
         • Nicknames
         • Aliases
         • Acronyms
         • Differences in translation
       • Exact string matching
         • POOR results!
     • Determine whether two strings refer to the same entity based on the strings above.

2. Research

   • Evaluate Name Matching methods
     • In Chineese
     • Approaches
       • Existing
         • String matching
           • ?
         • Learnig
           • ?
       • New
         • New Representation for Chinese
     • Experiments
       • New Representation for Chinese
         • Improves
           • name matching
           • Entity clustering
         • No details?!
   • Newly developed data sets
     • Matched Chinese name pairs
   • Mingpipe
     • Name matching tool
       • Python package
         • Usage
           • As stand alone
           • Integrated in a larger system