PHP Stanford CoreNLP adapter
PHP adapter for use with Stanford CoreNLP
Features
- Connect to Stanford University CoreNLP API online
- Connect to Stanford CoreNLP 3.7.0 server
- Annotators available: tokenize,ssplit,pos, parse, depparse, ner, regexner,lemma, mention, natlog, coref, openie, kbp
- The package creates Part-Of-Speech Trees with depth, parent- and child ID
Requirements
- PHP 5.5 or higher: it also works on PHP 7
- Windows or Linux 64-bit, 8Gb memory or more recommended
- Either Guzzle HTTP Client (installed by default) or only cURL.
- Composer for PHP
https://getcomposer.org/
Update 24th February 2018
PHP7 Type hinting removed, because it was causing issues for some users.
Update 28th January 2019
Fixed issue with PHP 7.1 upwards
Installation using ZIP files
- Install Stanford CoreNLP Server. See the installation walkthrough below.
- Download and unpack the files from this package.
- Copy the files to your to your webserver directory. Usually "htdocs" or "var/www".
- Run a Composer update
Installation using Composer
- Insert the following line into the "require" of your "composer.json" file.
{
"require": {
"dennis-de-swart/php-stanford-corenlp-adapter": "*"
}
}
- Run a composer update
Using the Stanford CoreNLP online API service
The adapter by default uses Stanford's online API service. This should work right after the composer update.
Note that the online API is a public service. If you want to analyze large volumes of text or sensitive data,
please install the Java server version.
OpenIE
OpenIE creates "subject-relation-object" tuples. This is similar (but not the same) as the "Subject-Verb-Object" concept of the English language.
Notes:
- OpenIE is only available on the Java offline version, not with the "online" mode. See the installation walkthrough below
- OpenIE data is not always available. Sometimes the result array might show empty, this is not an error.
http://nlp.stanford.edu/software/openie.html
https://en.wikipedia.org/wiki/Subject-verb-object
Installation / Walkthrough for Java server version
Step 1: install Java
https://java.com/en/download/help/index_installing.xml?os=All+Platforms&j=8&n=20
Step 2: installing the Stanford CoreNLP 3.7.0 server
http://stanfordnlp.github.io/CoreNLP/index.html#download
Step 3: Port for server
Default port for the Java server is port 9000. If port 9000 is not available you can change the port in the "bootstrap.php" file. Example:
define('CURLURL' , 'http://localhost:9000/');
Step 4: Start the CoreNLP serve from the command line.
Go to the download directory, then enter the following command:
java -mx8g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer -port 9000
Important note: the Stanford manual says "-mx4g", however I found that this can lead to a Java OutOfMemory error. It is also important to use a 64-bit operating system with at enough memory (8Gb or more recommended)
Step 5: Test if the server has started by surfing to it's URL
http://localhost:9000/
When you surf to this URL, you should see the CoreNLP GUI. If you have problems with installation you can check the manual:
http://stanfordnlp.github.io/CoreNLP/corenlp-server.html
Step 6: Set ONLINE_API to FALSE
In "bootstrap.php" set define('ONLINE_API' , FALSE). This tells the Adapter to use the Java version
Usage examples
Instantiate the adapter:
$coreNLP = new CorenlpAdapter();
To process a text, call the "getOutput" method:
$text = 'The Golden Gate Bridge was designed by Joseph Strauss.';
$coreNLP->getOutput($text);
Note that the first time that you process a text, the server takes about 20 to 30 seconds extra to load definitions. All other calls to the server after that will be much faster. Small texts are usually processed within seconds.
The results
If successful the following properties will be available:
$coreNLP->serverMemory; //contains all of the server output
$coreNLP->trees; //contains processed flat trees. Each part of the tree is assigned an ID key
$coreNLP->getWordValues($coreNLP->trees[1]) // get just the words from a tree
Diagram A: Tree With Tokens
Array
(
[1] => Array
(
[parent] =>
[pennTreebankTag] => ROOT
[depth] => 0
)
[2] => Array
(
[parent] => 1
[pennTreebankTag] => S
[depth] => 2
)
[3] => Array
(
[parent] => 2
[pennTreebankTag] => NP
[depth] => 4
)
[4] => Array
(
[parent] => 3
[pennTreebankTag] => PRP
[depth] => 6
[word] => I
[index] => 1
[originalText] => I
[lemma] => I
[characterOffsetBegin] => 0
[characterOffsetEnd] => 1
[pos] => PRP
[ner] => O
[before] =>
[after] =>
[openIE] => Array
(
[0] => subject
[1] => subject
[2] => subject
)
)
[5] => Array
(
[parent] => 2
[pennTreebankTag] => VP
[depth] => 4
)
[6] => Array
(
[parent] => 5
[pennTreebankTag] => MD
[depth] => 6
[word] => will
[index] => 2
[originalText] => will
[lemma] => will
[characterOffsetBegin] => 2
[characterOffsetEnd] => 6
[pos] => MD
[ner] => O
[before] =>
[after] =>
[openIE] => Array
(
[0] => subject
[1] => subject
[2] => relation
)
)
[7] => Array
(
[parent] => 5
[pennTreebankTag] => VP
[depth] => 6
)
[8] => Array
(
[parent] => 7
[pennTreebankTag] => VB
[depth] => 8
[word] => meet
[index] => 3
[originalText] => meet
[lemma] => meet
[characterOffsetBegin] => 7
[characterOffsetEnd] => 11
[pos] => VB
[ner] => O
[before] =>
[after] =>
[openIE] => Array
(
[0] => subject
[1] => subject
[2] => relation
)
)
[9] => Array
(
[parent] => 7
[pennTreebankTag] => NP
[depth] => 8
)
[10] => Array
(
[parent] => 9
[pennTreebankTag] => NP
[depth] => 10
)
[11] => Array
(
[parent] => 10
[pennTreebankTag] => NNP
[depth] => 12
[word] => Mary
[index] => 4
[originalText] => Mary
[lemma] => Mary
[characterOffsetBegin] => 12
[characterOffsetEnd] => 16
[pos] => NNP
[ner] => PERSON
[before] =>
[after] =>
[openIE] => Array
(
[1] => subject
[2] => object
[3] => subject
[0] => subject
)
)
[12] => Array
(
[parent] => 9
[pennTreebankTag] => PP
[depth] => 10
)
[13] => Array
(
[parent] => 12
[pennTreebankTag] => IN
[depth] => 12
[word] => in
[index] => 5
[originalText] => in
[lemma] => in
[characterOffsetBegin] => 17
[characterOffsetEnd] => 19
[pos] => IN
[ner] => O
[before] =>
[after] =>
[openIE] => Array
(
[1] => relation
[3] => relation
[0] => relation
)
)
[14] => Array
(
[parent] => 12
[pennTreebankTag] => NP
[depth] => 12
)
[15] => Array
(
[parent] => 14
[pennTreebankTag] => NNP
[depth] => 14
[word] => New
[index] => 6
[originalText] => New
[lemma] => New
[characterOffsetBegin] => 20
[characterOffsetEnd] => 23
[pos] => NNP
[ner] => LOCATION
[before] =>
[after] =>
[openIE] => Array
(
[1] => relation
[3] => object
[0] => object
)
)
[16] => Array
(
[parent] => 14
[pennTreebankTag] => NNP
[depth] => 14
[word] => York
[index] => 7
[originalText] => York
[lemma] => York
[characterOffsetBegin] => 24
[characterOffsetEnd] => 28
[pos] => NNP
[ner] => LOCATION
[before] =>
[after] =>
[openIE] => Array
(
[1] => object
[3] => object
)
)
[17] => Array
(
[parent] => 7
[pennTreebankTag] => PP
[depth] => 8
)
[18] => Array
(
[parent] => 17
[pennTreebankTag] => IN
[depth] => 10
[word] => at
[index] => 8
[originalText] => at
[lemma] => at
[characterOffsetBegin] => 29
[characterOffsetEnd] => 31
[pos] => IN
[ner] => O
[before] =>
[after] =>
[openIE] => Array
(
[1] => object
)
)
[19] => Array
(
[parent] => 17
[pennTreebankTag] => NP
[depth] => 10
)
[20] => Array
(
[parent] => 19
[pennTreebankTag] => CD
[depth] => 12
[word] => 10pm
[index] => 9
[originalText] => 10pm
[lemma] => 10pm
[characterOffsetBegin] => 32
[characterOffsetEnd] => 36
[pos] => CD
[ner] => TIME
[normalizedNER] => T22:00
[before] =>
[after] =>
[timex] => Array
(
[tid] => t1
[type] => TIME
[value] => T22:00
)
[openIE] => Array
(
[0] => object
[1] => object
)
)
)
Diagram B: The ServerMemory contains all the server data
Array
(
[0] => Array
(
[sentences] => Array
(
[0] => Array
(
[index] => 0
[parse] => (ROOT
(S
(NP (PRP I))
(VP (MD will)
(VP (VB meet)
(NP
(NP (NNP Mary))
(PP (IN in)
(NP (NNP New) (NNP York))))
(PP (IN at)
(NP (CD 10pm)))))))
[basic-dependencies] => Array
(
[0] => Array
(
[dep] => ROOT
[governor] => 0
[governorGloss] => ROOT
[dependent] => 3
[dependentGloss] => meet
)
[1] => Array
(
[dep] => nsubj
[governor] => 3
[governorGloss] => meet
[dependent] => 1
[dependentGloss] => I
)
[2] => Array
(
[dep] => aux
[governor] => 3
[governorGloss] => meet
[dependent] => 2
[dependentGloss] => will
)
[3] => Array
(
[dep] => dobj
[governor] => 3
[governorGloss] => meet
[dependent] => 4
[dependentGloss] => Mary
)
[4] => Array
(
[dep] => case
[governor] => 7
[governorGloss] => York
[dependent] => 5
[dependentGloss] => in
)
[5] => Array
(
[dep] => compound
[governor] => 7
[governorGloss] => York
[dependent] => 6
[dependentGloss] => New
)
[6] => Array
(
[dep] => nmod
[governor] => 4
[governorGloss] => Mary
[dependent] => 7
[dependentGloss] => York
)
[7] => Array
(
[dep] => case
[governor] => 9
[governorGloss] => 10pm
[dependent] => 8
[dependentGloss] => at
)
[8] => Array
(
[dep] => nmod
[governor] => 3
[governorGloss] => meet
[dependent] => 9
[dependentGloss] => 10pm
)
)
[collapsed-dependencies] => Array
(
[0] => Array
(
[dep] => ROOT
[governor] => 0
[governorGloss] => ROOT
[dependent] => 3
[dependentGloss] => meet
)
[1] => Array
(
[dep] => nsubj
[governor] => 3
[governorGloss] => meet
[dependent] => 1
[dependentGloss] => I
)
[2] => Array
(
[dep] => aux
[governor] => 3
[governorGloss] => meet
[dependent] => 2
[dependentGloss] => will
)
[3] => Array
(
[dep] => dobj
[governor] => 3
[governorGloss] => meet
[dependent] => 4
[dependentGloss] => Mary
)
[4] => Array
(
[dep] => case
[governor] => 7
[governorGloss] => York
[dependent] => 5
[dependentGloss] => in
)
[5] => Array
(
[dep] => compound
[governor] => 7
[governorGloss] => York
[dependent] => 6
[dependentGloss] => New
)
[6] => Array
(
[dep] => nmod:in
[governor] => 4
[governorGloss] => Mary
[dependent] => 7
[dependentGloss] => York
)
[7] => Array
(
[dep] => case
[governor] => 9
[governorGloss] => 10pm
[dependent] => 8
[dependentGloss] => at
)
[8] => Array
(
[dep] => nmod:at
[governor] => 3
[governorGloss] => meet
[dependent] => 9
[dependentGloss] => 10pm
)
)
[collapsed-ccprocessed-dependencies] => Array
(
[0] => Array
(
[dep] => ROOT
[governor] => 0
[governorGloss] => ROOT
[dependent] => 3
[dependentGloss] => meet
)
[1] => Array
(
[dep] => nsubj
[governor] => 3
[governorGloss] => meet
[dependent] => 1
[dependentGloss] => I
)
[2] => Array
(
[dep] => aux
[governor] => 3
[governorGloss] => meet
[dependent] => 2
[dependentGloss] => will
)
[3] => Array
(
[dep] => dobj
[governor] => 3
[governorGloss] => meet
[dependent] => 4
[dependentGloss] => Mary
)
[4] => Array
(
[dep] => case
[governor] => 7
[governorGloss] => York
[dependent] => 5
[dependentGloss] => in
)
[5] => Array
(
[dep] => compound
[governor] => 7
[governorGloss] => York
[dependent] => 6
[dependentGloss] => New
)
[6] => Array
(
[dep] => nmod:in
[governor] => 4
[governorGloss] => Mary
[dependent] => 7
[dependentGloss] => York
)
[7] => Array
(
[dep] => case
[governor] => 9
[governorGloss] => 10pm
[dependent] => 8
[dependentGloss] => at
)
[8] => Array
(
[dep] => nmod:at
[governor] => 3
[governorGloss] => meet
[dependent] => 9
[dependentGloss] => 10pm
)
)
[openie] => Array
(
[0] => Array
(
[subject] => I
[subjectSpan] => Array
(
[0] => 0
[1] => 1
)
[relation] => will meet Mary at
[relationSpan] => Array
(
[0] => 1
[1] => 3
)
[object] => 10pm
[objectSpan] => Array
(
[0] => 8
[1] => 9
)
)
[1] => Array
(
[subject] => I
[subjectSpan] => Array
(
[0] => 0
[1] => 1
)
[relation] => will meet
[relationSpan] => Array
(
[0] => 1
[1] => 3
)
[object] => Mary in New York
[objectSpan] => Array
(
[0] => 3
[1] => 7
)
)
[2] => Array
(
[subject] => I
[subjectSpan] => Array
(
[0] => 0
[1] => 1
)
[relation] => will meet
[relationSpan] => Array
(
[0] => 1
[1] => 3
)
[object] => Mary
[objectSpan] => Array
(
[0] => 3
[1] => 4
)
)
[3] => Array
(
[subject] => Mary
[subjectSpan] => Array
(
[0] => 3
[1] => 4
)
[relation] => is in
[relationSpan] => Array
(
[0] => 4
[1] => 5
)
[object] => New York
[objectSpan] => Array
(
[0] => 5
[1] => 7
)
)
)
[tokens] => Array
(
[0] => Array
(
[index] => 1
[word] => I
[originalText] => I
[lemma] => I
[characterOffsetBegin] => 0
[characterOffsetEnd] => 1
[pos] => PRP
[ner] => O
[before] =>
[after] =>
)
[1] => Array
(
[index] => 2
[word] => will
[originalText] => will
[lemma] => will
[characterOffsetBegin] => 2
[characterOffsetEnd] => 6
[pos] => MD
[ner] => O
[before] =>
[after] =>
)
[2] => Array
(
[index] => 3
[word] => meet
[originalText] => meet
[lemma] => meet
[characterOffsetBegin] => 7
[characterOffsetEnd] => 11
[pos] => VB
[ner] => O
[before] =>
[after] =>
)
[3] => Array
(
[index] => 4
[word] => Mary
[originalText] => Mary
[lemma] => Mary
[characterOffsetBegin] => 12
[characterOffsetEnd] => 16
[pos] => NNP
[ner] => PERSON
[before] =>
[after] =>
)
[4] => Array
(
[index] => 5
[word] => in
[originalText] => in
[lemma] => in
[characterOffsetBegin] => 17
[characterOffsetEnd] => 19
[pos] => IN
[ner] => O
[before] =>
[after] =>
)
[5] => Array
(
[index] => 6
[word] => New
[originalText] => New
[lemma] => New
[characterOffsetBegin] => 20
[characterOffsetEnd] => 23
[pos] => NNP
[ner] => LOCATION
[before] =>
[after] =>
)
[6] => Array
(
[index] => 7
[word] => York
[originalText] => York
[lemma] => York
[characterOffsetBegin] => 24
[characterOffsetEnd] => 28
[pos] => NNP
[ner] => LOCATION
[before] =>
[after] =>
)
[7] => Array
(
[index] => 8
[word] => at
[originalText] => at
[lemma] => at
[characterOffsetBegin] => 29
[characterOffsetEnd] => 31
[pos] => IN
[ner] => O
[before] =>
[after] =>
)
[8] => Array
(
[index] => 9
[word] => 10pm
[originalText] => 10pm
[lemma] => 10pm
[characterOffsetBegin] => 32
[characterOffsetEnd] => 36
[pos] => CD
[ner] => TIME
[normalizedNER] => T22:00
[before] =>
[after] =>
[timex] => Array
(
[tid] => t1
[type] => TIME
[value] => T22:00
)
)
)
)
)
)
Any questions?
Please let me know.
Credits
Some functions are forked from this "Stanford parser" package:
https://github.com/agentile/PHP-Stanford-NLP