Willie Rogers

Purpose

MetaMap maps terms occuring in text to UMLS Metathesaurus concepts. As part of this mapping process, MetaMap tokenizes text into sections, sentences, phrases, terms, and words. MetaMap maps the noun phrases of the text to the best matching UMLS concept or set of concepts that best cover each phrase. The MetaMap Java API provides java programs with programmatic access to MetaMap mapping engine.

MetaMap API's Underlying Architecture

MetaMap mapping engine is written primarily in SICStus Prolog; to facilitate its use by Java programs, the system uses PrologBeans to provide a loose coupling between the Java API and the mapping engine. See the SICStus Prolog PrologBeans documentation for more information (http://www.sics.se/sicstus/docs/latest4/html/sicstus.html/lib_002dprologbeans.html).

Pre-requisites

The full MetaMap download and installation is required to use the MetaMap Java API (see \http://metamap.nlm.nih.gov/#Downloads). Also, Java 1.6 SDK or greater is required. (It should work Java 1.5 but it has not been tested with Java 1.5)

Downloading, Extracting and Installing the API distribution

In the directory where you installed the Public Metamap (the directory containing the public_mm directory) extract the javaapi archive:

$ bzip2 -dc /home/piro/public_mm_linux_javaapi_{four-digit-year}.tar.bz2 | tar xvf -

If you plan on modifying the sources to the prolog-based MetaMap server (mmserver) you will need to download and extract the source archive (http://metamap.nlm.nih.gov/download/public_mm_src_{four-digit-year}.tar.bz2) as well:

$ bzip2 -dc /home/piro/public_mm_src_{four-digit-year}.tar.bz2 | tar xvf -

Linux Install

You will need to re-run ./bin/install.sh from the public_mm directory to setup the files for javaapi.

$ ./bin/install.sh

Windows Graphical Install

Using Windows Explorer find the "public_mm" directory extracted from the distribution archive. Click on the icon with the name "Install MetaMap". The dialog will display an input box with the label "Location of the Public MetaMap Directory:" filled in with the probable location where you've installed the distribution.


If the location is incorrent, you can change it by clicking the "Browse" button and using the file dialog to locate the directory or typing in the location in the input text box. When the "public_mm" location is set click the "Next" button to go to the next dialog.

The next dialog will display one of possible locations where the Java Runtime Environment (JRE) is installed. This information is obtained from the Windows Registry which was set when one of the JREs was installed.


If the JRE location displayed is not the one you wish to use, change it by clicking the "Browse" button and using the file dialog to locate the proper directory or type in the location in the input text box. When the proper JRE location is set click the "Next" button to go to the next dialog.

The next dialog will display a results of install.


If no errors are present in the log you can close the installation program by pressing the "Finish" button. After that, skip to the section on running MetaMap.

Using the MetaMap server

Starting supporting servers

The MetaMap server (mmserver) must first be running to use the Java API. If the SKR/Medpost Tagger is not already running start it using the following command:

$ ./bin/skrmedpostctl start

Under Windows use the following (In a command prompt window):

\public_mm> bin\skrmedpostctl_start

If you wish to the Word Sense Disambiguation (WSD) Server (optional), start it also.

$ ./bin/wsdserverctl start

Under Windows use the following:

public_mm> bin\wsdserverctl_start

Running the MetaMap server

Then start the MetaMap server:

$ ./bin/mmserver
/home/piro/public_mm/bin/SKRrun -L {year} \
     -w /home/piro/public_mm/lexicon /home/piro/public_mm/bin/mmserver{year}.BINARY.Linux -Z {two-digit-year}
Server options: [port(8888),accepted_hosts(['127.0.0.1','130.14.111.76','130.14.110.82'])]
Berkeley DB databases (normal {two-digit-year} strict model) are open.
Static variants will come from table varsan in /home/piro/public_mm/DB/BDB4/DB.normal.{two-digit-year}.strict.
Derivational Variants: Adj/noun ONLY.
Accessing lexicon /home/piro/public_mm/lexicon/data/BDB4/lexiconStatic{four-digit-year}.
Variant generation mode: static.

If you're using Windows XP or Windows 7 then use the command bin\mmserver11 in a "Command Prompt" or "Windows PowerShell" window.

Testing the API

Using another terminal, you can verify that api is running using the program testapi.sh (testapi.bat on Windows) which takes a query as an argument:

$ ./testapi.sh laboratory culture
options: []
terms: laboratory culture 
input text: 
 laboratory culture 
Utterance:
 Id: 00000000.tx.1
 Utterance text: laboratory culture 
 Position: (0, 19)
Phrase:
 text: laboratory culture
 Minimal Commitment Parse: [head([lexmatch([laboratory culture]),inputmatch([laboratory,culture]),tag(noun),tokens([laboratory,culture])])]
Candidates:
Mappings:
 Map Score: -1000
   Score: -1000
   Concept Id: C0430400
   Concept Name: Laboratory culture
   Preferred Name: Laboratory culture
   Matched Words: [laboratory, culture]
   Semantic Types: [lbpr]
   MatchMap: [[[1, 2], [1, 2], 0]]
   MatchMap alt. repr.: [[[phrase start: 1, phrase end: 2], [concept start: 1, concept end: 2], lexical variation: 0]]
   is Head?: true
   is Overmatch?: false
   Sources: [MTH, LNC, MDR, NCI, RCD, MEDCIN, CCPSS, SNOMEDCT]
   Positional Info: [(0, 18)]
$

The source to java class used testapi.sh is in ./src/javaapi/sources/gov/nih/nlm/nls/metamap/MetaMapApiTest.java.

Using the API

Instantiating the API

The following sections expose the source code used to produce the example output shown in previous section

MetaMapApi api = new MetaMapApiImpl();

If one is running the MetaMap server (mmserver) of a machine other than the one the Java api client is running, one can specify the hostname of the MetaMap server when instantiating the api:

If the MetaMap server is running on the host: "resource.example.org" then the instantiation would be as follows:

MetaMapApi api = new MetaMapApiImpl("resource.example.org");

Setting MetaMap options

    api.setOptions("-y");  // turn on Word Sense Disambiguation

Multiple options

    api.setOptions("-yk dsyn");  // turn on Word Sense Disambiguation and
                                 // exclude concepts with dsyn
                                 // (disease or syndrome)

Or alternatively:

    api.setOptions("-y -k dsyn");  // turn on Word Sense Disambiguation and
                                   // exclude concepts with dsyn
                                   // (disease or syndrome)

Or a more complicated method:

    List<String> theOptions = new ArrayList<String>();
    theOptions.append("-y");  // turn on Word Sense Disambiguation
    if (theOptions.size() > 0) {
      api.setOptions(theOptions);
    }

MetaMap options available in the api:

   -@ --WSD <hostname>                   : Which WSD server to use.
   -8 --dynamic_variant_generation       : dynamic variant generation
   -A --strict_model                     : use strict model 
   -C --relaxed_model                    : use relaxed model 
   -D --all_derivational_variants        : all derivational variants
   -J --restrict_to_sts <semtypelist>    : restrict to semantic types
   -K --ignore_stop_phrases              : ignore stop phrases.
   -R --restrict_to_sources <sourcelist> : restrict to sources
   -S --tagger <sourcelist>              : Which tagger to use.
   -V --mm_data_version <name>           : version of MetaMap data to use.
   -X --truncate_candidates_mappings     : truncate candidates mapping
   -Y --prefer_multiple_concepts         : prefer multiple concepts
   -Z --mm_data_year <name>              : year of MetaMap data to use.
   -a --all_acros_abbrs                  : allow Acronym/Abbreviation variants
   -b --compute_all_mappings             : compute/display all mappings
   -d --no_derivational_variants         : no derivational variants
   -e --exclude_sources <sourcelist>     : exclude sources
   -g --allow_concept_gaps               : allow concept gaps
   -i --ignore_word_order                : ignore word order
   -k --exclude_sts <semtypelist>        : exclude semantic types
   -l --allow_large_n                    : allow Large N
   -o --allow_overmatches                : allow overmatches 
   -r --threshold <integer>              : Threshold for displaying candidates. 
   -y --word_sense_disambiguation        : use WSD 
   -z --term_processing                  : use term processing 

Performing a query using the api

List<Result> resultList = api.processCitationsFromString(terms);

Interrogating the result

Getting Acronyms and Abbreviations

To get a list of all the acronyms and abbreviations occurring in the input text use the instance method Result.getAcronymsAbbrevs:

Result result = resultList.get(0);
List<AcronymsAbbrevs> aaList = result.getAcronymsAbbrevs();
if (aaList.size() > 0) {
  System.out.println("Acronyms and Abbreviations:");
  for (AcronymsAbbrevs e: aaList) {
    System.out.println("Acronym: " + e.getAcronym());
    System.out.println("Expansion: " + e.getExpansion());
    System.out.println("Count list: " + e.getCountList());
    System.out.println("CUI list: " + e.getCUIList());
  }
} else {
  System.out.println(" None.");
}

Getting Negations

To get a list of all the negated concepts in the input text use the instance method Result.getNegations:

List<Negation> negList = result.getNegations();
if (negList.size() > 0) {
  System.out.println("Negations:");
  for (Negation e: negList) {
    System.out.println("type: " + e.getType());
    System.out.print("Trigger: " + e.getTrigger() + ": [");
    for (Position pos: e.getTriggerPositionList()) {
      System.out.print(pos  + ",");
    }
    System.out.println("]");
    System.out.print("ConceptPairs: [");
    for (ConceptPair pair: e.getConceptPairList()) {
      System.out.print(pair + ",");
    }
    System.out.println("]");
    System.out.print("ConceptPositionList: [");
    for (Position pos: e.getConceptPositionList()) {
      System.out.print(pos + ",");
    }
    System.out.println("]");
  }
} else {
	System.out.println(" None.");
}

Getting Utterances and Associated Phrases, Candidates, and Mappings

The instance method Result.getUtteranceList() produces a list of the utterances present in the result:

for (Utterance utterance: result.getUtteranceList()) {
	System.out.println("Utterance:");
	System.out.println(" Id: " + utterance.getId());
	System.out.println(" Utterance text: " + utterance.getString());
	System.out.println(" Position: " + utterance.getPosition());

To get the list of phrases, candidates, and mappings associated with an utterance use the instance method Utterance.getPCMList:

	for (PCM pcm: utterance.getPCMList()) {

Each phrase, and the list of candidates and mappings associated with the phrase are encapsulated within a PCM instance. Use PCM.getPhrase to get the phrase instance residing within the PCM instance:

	  System.out.println("Phrase:");
	  System.out.println(" text: " + pcm.getPhrase().getPhraseText());

Similarly, get the candidate list using PCM.getCandidateList():

          System.out.println("Candidates:");
          for (Ev ev: pcm.getCandidateList()) {
            System.out.println(" Candidate:");
            System.out.println("  Score: " + ev.getScore());
            System.out.println("  Concept Id: " + ev.getConceptId());
            System.out.println("  Concept Name: " + ev.getConceptName());
            System.out.println("  Preferred Name: " + ev.getPreferredName());
            System.out.println("  Matched Words: " + ev.getMatchedWords());
            System.out.println("  Semantic Types: " + ev.getSemanticTypes());
            System.out.println("  MatchMap: " + ev.getMatchMap());
            System.out.println("  MatchMap alt. repr.: " + ev.getMatchMapList());
            System.out.println("  is Head?: " + ev.isHead());
            System.out.println("  is Overmatch?: " + ev.isOvermatch());
            System.out.println("  Sources: " + ev.getSources());
            System.out.println("  Positional Info: " + ev.getPositionalInfo());
          }

One can get the mappings list from the PCM instance using PCM.getMappingList:

          System.out.println("Mappings:");
          for (Mapping map: pcm.getMappingList()) {
            System.out.println(" Map Score: " + map.getScore());
            for (Ev mapEv: map.getEvList()) {
              System.out.println("   Score: " + mapEv.getScore());
              System.out.println("   Concept Id: " + mapEv.getConceptId());
              System.out.println("   Concept Name: " + mapEv.getConceptName());
              System.out.println("   Preferred Name: " + mapEv.getPreferredName());
              System.out.println("   Matched Words: " + mapEv.getMatchedWords());
              System.out.println("   Semantic Types: " + mapEv.getSemanticTypes());
              System.out.println("   MatchMap: " + mapEv.getMatchMap());
              System.out.println("   MatchMap alt. repr.: " + mapEv.getMatchMapList());
              System.out.println("   is Head?: " + mapEv.isHead());
              System.out.println("   is Overmatch?: " + mapEv.isOvermatch());
              System.out.println("   Sources: " + mapEv.getSources());
              System.out.println("   Positional Info: " + mapEv.getPositionalInfo());
            }
          }
    }
}

Refer to the API javadoc for more information on the available methods for each interface.

A complete example of this code is in src/javaapi/sources/gov/nih/nlm/nls/metamap/MetaMapApiTest.java.

Getting Raw MetaMap Machine Output

A copy of the raw MetaMap machine output can be obtained by using the instance method Result.getMachineOutput:

List<Result> resultList = api.processCitationsFromString(terms);
Result result = resultList.get(0);
String machineOutput = result.getMachineOutput();

Advanced Configuration

Accepting the client connections from remote hosts

To allow the MetaMap server to accept connections from clients on a host other than localhost (127.0.0.1, the default) modify the environment variable ACCEPTED_HOSTS the script public_mm/bin/SKRrun.11 to contain all of the client hosts you wish to have access to the MetaMap server.

For example, change the entry:

ACCEPTED_HOSTS="['127.0.0.1']"
export ACCEPTED_HOSTS

to:

ACCEPTED_HOSTS="['127.0.0.1','192.168.111.27','192.168.111.61','192.168.111.76']"
export ACCEPTED_HOSTS

all of the entries must be ip-addresses, hostnames will not work.

Running the MetaMap server on an alternate port

To run the MetaMap server on a port other than the default port (8066) modify the environment variable "MMSERVER_PORT" in the script public_mm/bin/SKRrun.{2-digit-year}.

E.G. to change the port to 8888:

MMSERVER_PORT=8888
export MMSERVER_PORT

Specifying alternate MetaMap Server hosts and ports in the API

The MetaMap Java API now includes the methods setHost and setPort to specify the host and port locations of the MetaMap server. The source to the class gov.nih.nlm.nls.metamap.MetaMapApiTest (public_mm/src/javaapi/sources/gov/nih/nlm/nls/metamap/ MetaMapApiTest.java) contains an example of the use of these methods.

Location of jar files

The jar files used by the Java API are in the following locations:

./public_mm/src/javaapi/dist/MetaMapApi.jar
./public_mm/src/javaapi/dist/prologbeans.jar

Notes for Maven Users

To install the MetaMap api and its supporting libraries use the following commands:

First install PrologBeans:

$ mvn install:install-file \
  -Dfile=<parent>/public_mm/src/javaapi/dist/prologbeans.jar \
  -DgroupId=se.sics -DartifactId=prologbeans -Dversion=4.2.1 \
  -Dpackaging=jar

Then install MetaMap API:

$ cd <parent>/public_mm/src/javaapi
$ mvn install

What is missing from the API

The server sorely needs support for gracefully dealing with internal exceptions when processing options and input text including a mechanism for returning to a stable state. Some mechanism for reporting error conditions back to the client would be nice as well.

The MetaMap UIMA Annotator

A UIMA Annotator that uses an underlying MetaMapApi instance is available (see http://metamap.nlm.nih.gov/#MetaMapUIMA). More information about UIMA project can be found at the Apache UIMA project Page, http://uima.apache.org/ .

Possible enhancements

Possible enchancments include providing a factory method for instantiating MetaMapApi instances, and providing instantiation of instances through JNDI. A mechanism for querying the Prolog Server to determine which options are available and how to use them.

Differences compared to MMTx API

The MetaMap Java API only provides access to components available through machine output. That includes The Final mappings, the Candidates, Phrases, Utterances, Negated Concepts, and Acronyms and Abbreviations. Access to structures such as lexical elements and tokenization that were available in the eariler MMTx API are not currently available in the MetaMap API.

Windows XP/7 differences

The primary differences are:

For more information

SICStus Prolog
http://www.sics.se/sicstus/
PrologBeans Docs
http://www.sics.se/sicstus/docs/latest4/html/sicstus.html
PrologBeans Java API Docs
http://www.sics.se/sicstus/docs/3.12.9/html/prologbeans/
MetaMap
http://metamap.nlm.nih.gov/



This document was generated using AFT v5.097