MMTx Detailed Trouble Report (Trouble Report #247) |
Field | Field Data |
---|---|
ID | 247 |
Category | mmtxT |
Synopsis | Conservative Final Mapping |
Severity | non-critical |
Priority | low |
Responsible | guy |
State | analyzed |
Class | change-request |
Arrival-Date | Thu Jun 15 13:34:00 EDT 2006 |
Closed-Date | --- |
Release | V2.4B |
Environment | Operating System: MMTx: Windows NT/2000 MySQL: OS Version: 4.0 Java Version: Java 1.4 MySQL Version: |
Description | In a recent analysis of MMTx results against our document set, we have noticed situtations when the inclusion of certain candidate maps in the final mapping would make for a better final mapping. A concrete example from our report set follows. (This is using the MMTx command line processor, and results are consistent with the MMTxAPI called from with a Java app.) Processing 00000000.tx.1: It displays a pink-tan glistening epithelial surface. Phrase: "It" Meta Candidates (0): Meta Mappings: Phrase: "displays" Meta Candidates (1) 966 Display (Display {NCI}) [Activity] Meta Mapping (966) 966 Display (Display {NCI}) [Activity] Phrase: "a pink-tan glistening epithelial surface." Meta Candidates (8) 804 SURFACE (SURF {NCI,LNC}) [Spatial Concept] 637 TAN (2,2,6,6-Tetramethyl-4-oxopiperidine-1-oxyl {MSH,MTH}) [Indicator, Reagent, or Diagnostic Aid,Organic Chemical,Pharma cologic Substance] 637 Epithelial (Epithelial {NCI,NCI-GLOSS}) [Body System] 604 Tanning (Tannings {MSH,MTH}) [Occupational Activity] 604 Tanning (Tanning {MTH}) [Individual Behavior,Individual Behavior,Organism Function] 582 Epi (Adrenalin {MSH,UWDA,NCI,NCI-GLOSS,LNC}) [Organic Chemical,Pharmacologic Substance] 582 EPI (Exocrine Pancreatic Insufficiency {MSH}) [Disease or Syndrome] 566 Epithelium (Epithelium {MSH,NCI,UWDA,NCI-GLOSS}) [Tissue] Meta Mapping (764) 637 TAN (2,2,6,6-Tetramethyl-4-oxopiperidine-1-oxyl {MSH,MTH}) [Indicator, Reagent, or Diagnostic Aid,Organic Chemical,Pharma cologic Substance] 804 SURFACE (SURF {NCI,LNC}) [Spatial Concept] For the phrase "a pink-tan glistening epithelial surface" the final mapping is "tan surface". One would hope that "epithelial tan surface" would be chosen. And it seems like the preliminary candidates make this possible. We have seen this same pattern of behavior across many sentences. It would be a great help if you could explain the algorithmic reasoning that leads to this decision and possibly suggest a configuration of options that would favour a more liberal coverage in these cases. My guess would be that MMTx is considering the POS of these terms and prefers a single ADJ N head rather than ADJ ADJ N. Thanks, Kevin |
How-To-Repeat | This should be reproducible by simply giving the example sentence to MMTx using the default configurations in mmtxRegistry.conf. |
Notes | --- |
[ Contact Us | Copyright, Privacy, and Disclaimer Notice | Accessibility ]