Difference between revisions of "Cohen and Hersh Briefings in Bioinformatics 2005"

From Cohen Courses
Jump to navigationJump to search
(Created page with '== Citation == Aaron M. Cohen and William R. Hersh. 2005. A Survey of Current Work in Biomedical Text Mining. Briefings in Bioinformatics. Vol 6. No 1. 57-71. == Online version…')
 
 
(4 intermediate revisions by the same user not shown)
Line 8: Line 8:
  
 
== Summary ==
 
== Summary ==
 +
This is a survey paper about biomedical text mining in 2005.
  
−
The paper presents a MEDical Information Extraction (MedIE) system, which extracts patient information from free-text clinical records.  
+
* Named entity recognition
 +
** Problems
 +
*** No complete dictionary for most types of biological named entities
 +
*** ambiguous words and phrases
 +
*** multi names
 +
** approaches are mainly categorized into three below
 +
*** lexicon based
 +
*** rule based
 +
*** statistically based
 +
** performance
 +
*** overall, the performance of gene and protein NER systems is F-scores between 75 and 85 percent.
  
−
They divided their extraction job into three tasks below.
+
* Text classification
−
* extraction of medical terms
 
−
* relation extraction
 
−
** extraction of associated medical concepts
 
−
** e.g. Blood pressure & 144/90 in the sentence, "Blood pressure is 144/90"
 
−
* text classification  
 
−
** e.g. a patient can be classified as a former smoker, a current smoker, or a non-smoker
 
  
−
Their approaches are:
 
−
* An ontology-based approach for extracting medical terms of interest
 
−
** they used Unified Medical Language System (UMLS)
 
−
** About terms that are not defined in UMLS, they predicted categories of some terms using sentence structures.
 
−
* A graph-based approach which uses the parsing result of link-grammar parser for relation-extraction
 
−
** They included the processing of negation.
 
−
** When the parser fails, they used a pattern-based approach.
 
−
** Because the parser did not process multi-word terms, they replaced the terms with placeholders.
 
−
* an NLP-based feature extraction method coupled with an ID3-based decision tree for text classification
 
  
 +
* Synonym and abbreviation extraction
 +
** Synonym
 +
*** use dictionary
 +
*** automatic extraction of gene name synonyms from biomedical free text
 +
*** SVM classifier-based
 +
*** pattern-based
 +
** abbreviation
 +
*** either the full form or the abbreviation is often enclosed in parentheses.
 +
*** a variety of alignment and scoring methods
  
−
This approach was fairly successful mostly showing over 80% of precision and recall. However, the system was tested on the data written by only a clinician, which means that the style of free-text records was consistent. Nevertheless, the research is worth in that they applied various IE techniques to the free-text clinical records, explain about the problems they encountered.
+
* Relationship extraction
 +
** detect occurrences of a prespecified type of relationship between a pair of entities of given types
 +
** manually generated template-based methods
 +
** automatic template methods
 +
** statistical methods
 +
** NLP-based methods
 +
mostly are about the relationships between genes and proteins
  
−
== Related papers ==
+
* Hypothesis generation
 +
** uncover relationships that are not present in the text but instead are inferred by the presence of other more explicit relationships. uncover previously unrecognized relationships
  
−
The widely cited [[RelatedPaper::Pang et al EMNLP 2002]] paper was influenced by this paper - but considers supervised learning techniques.  The choice of movie reviews as the domain was suggested by the (relatively) poor performance of Turney's method on movies.
+
* Integration frameworks
 +
** integrated text-mining frameworks
 +
** still in the research and development phrase.
  
−
An interesting follow-up paper is [[RelatedPaper::Turney and Littman, TOIS 2003]] which focuses on evaluation of the technique of using PMI for predicting the [[semantic orientation of words]].
+
* The authors' suggestions
 +
** Access to full text is required
 +
** Additional analytical methods with possible features are required for a particular application
 +
** Researchers should consider actual users' needs. The performance of a system with certain metrics does not guarantee users' satisfaction.
 +
** Shared challenge tasks should be continued

Latest revision as of 00:23, 1 October 2010

Citation

Aaron M. Cohen and William R. Hersh. 2005. A Survey of Current Work in Biomedical Text Mining. Briefings in Bioinformatics. Vol 6. No 1. 57-71.

Online version

oxfordjournals

Summary

This is a survey paper about biomedical text mining in 2005.

  • Named entity recognition
    • Problems
      • No complete dictionary for most types of biological named entities
      • ambiguous words and phrases
      • multi names
    • approaches are mainly categorized into three below
      • lexicon based
      • rule based
      • statistically based
    • performance
      • overall, the performance of gene and protein NER systems is F-scores between 75 and 85 percent.
  • Text classification


  • Synonym and abbreviation extraction
    • Synonym
      • use dictionary
      • automatic extraction of gene name synonyms from biomedical free text
      • SVM classifier-based
      • pattern-based
    • abbreviation
      • either the full form or the abbreviation is often enclosed in parentheses.
      • a variety of alignment and scoring methods
  • Relationship extraction
    • detect occurrences of a prespecified type of relationship between a pair of entities of given types
    • manually generated template-based methods
    • automatic template methods
    • statistical methods
    • NLP-based methods

mostly are about the relationships between genes and proteins

  • Hypothesis generation
    • uncover relationships that are not present in the text but instead are inferred by the presence of other more explicit relationships. uncover previously unrecognized relationships
  • Integration frameworks
    • integrated text-mining frameworks
    • still in the research and development phrase.
  • The authors' suggestions
    • Access to full text is required
    • Additional analytical methods with possible features are required for a particular application
    • Researchers should consider actual users' needs. The performance of a system with certain metrics does not guarantee users' satisfaction.
    • Shared challenge tasks should be continued
Retrieved from "http://curtis.ml.cmu.edu/w/courses/index.php?title=Cohen_and_Hersh_Briefings_in_Bioinformatics_2005&oldid=1977"