I've tried PorterStemmer and Snowball but both don't work on all words, missing some very common ones.
My test words are: "cats running ran cactus cactuses cacti community communities", and both get less than half right.
See also:
Stemming algorithm that produces real words
Stemming - code examples or open source projects?
If you know Python, The Natural Language Toolkit (NLTK) has a very powerful lemmatizer that makes use of WordNet.
Note that if you are using this lemmatizer for the first time, you must download the corpus prior to using it. This can be done by:
>>> import nltk
>>> nltk.download('wordnet')
You only have to do this once. Assuming that you have now downloaded the corpus, it works like this:
>>> from nltk.stem.wordnet import WordNetLemmatizer
>>> lmtzr = WordNetLemmatizer()
>>> lmtzr.lemmatize('cars')
'car'
>>> lmtzr.lemmatize('feet')
'foot'
>>> lmtzr.lemmatize('people')
'people'
>>> lmtzr.lemmatize('fantasized','v')
'fantasize'
There are other lemmatizers in the nltk.stem module, but I haven't tried them myself.
I use stanford nlp to perform lemmatization. I have been stuck up with a similar problem in the last few days. All thanks to stackoverflow to help me solve the issue .
import java.util.*;
import edu.stanford.nlp.pipeline.*;
import edu.stanford.nlp.ling.*;
import edu.stanford.nlp.ling.CoreAnnotations.*;
public class example
{
public static void main(String[] args)
{
Properties props = new Properties();
props.put("annotators", "tokenize, ssplit, pos, lemma");
pipeline = new StanfordCoreNLP(props, false);
String text = /* the string you want */;
Annotation document = pipeline.process(text);
for(CoreMap sentence: document.get(SentencesAnnotation.class))
{
for(CoreLabel token: sentence.get(TokensAnnotation.class))
{
String word = token.get(TextAnnotation.class);
String lemma = token.get(LemmaAnnotation.class);
System.out.println("lemmatized version :" + lemma);
}
}
}
}
It also might be a good idea to use stopwords to minimize output lemmas if it's used later in classificator. Please take a look at coreNlp extension written by John Conwell.
I tried your list of terms on this snowball demo site and the results look okay....
cats -> cat
running -> run
ran -> ran
cactus -> cactus
cactuses -> cactus
community -> communiti
communities -> communiti
A stemmer is supposed to turn inflected forms of words down to some common root. It's not really a stemmer's job to make that root a 'proper' dictionary word. For that you need to look at morphological/orthographic analysers.
I think this question is about more or less the same thing, and Kaarel's answer to that question is where I took the second link from.
The stemmer vs lemmatizer debates goes on. It's a matter of preferring precision over efficiency. You should lemmatize to achieve linguistically meaningful units and stem to use minimal computing juice and still index a word and its variations under the same key.
See Stemmers vs Lemmatizers
Here's an example with python NLTK:
>>> sent = "cats running ran cactus cactuses cacti community communities"
>>> from nltk.stem import PorterStemmer, WordNetLemmatizer
>>>
>>> port = PorterStemmer()
>>> " ".join([port.stem(i) for i in sent.split()])
'cat run ran cactu cactus cacti commun commun'
>>>
>>> wnl = WordNetLemmatizer()
>>> " ".join([wnl.lemmatize(i) for i in sent.split()])
'cat running ran cactus cactus cactus community community'
Martin Porter's official page contains a Porter Stemmer in PHP as well as other languages.
If you're really serious about good stemming though you're going to need to start with something like the Porter Algorithm, refine it by adding rules to fix incorrect cases common to your dataset, and then finally add a lot of exceptions to the rules. This can be easily implemented with key/value pairs (dbm/hash/dictionaries) where the key is the word to look up and the value is the stemmed word to replace the original. A commercial search engine I worked on once ended up with 800 some exceptions to a modified Porter algorithm.
Based on various answers on Stack Overflow and blogs I've come across, this is the method I'm using, and it seems to return real words quite well. The idea is to split the incoming text into an array of words (use whichever method you'd like), and then find the parts of speech (POS) for those words and use that to help stem and lemmatize the words.
You're sample above doesn't work too well, because the POS can't be determined. However, if we use a real sentence, things work much better.
import nltk
from nltk.corpus import wordnet
lmtzr = nltk.WordNetLemmatizer().lemmatize
def get_wordnet_pos(treebank_tag):
if treebank_tag.startswith('J'):
return wordnet.ADJ
elif treebank_tag.startswith('V'):
return wordnet.VERB
elif treebank_tag.startswith('N'):
return wordnet.NOUN
elif treebank_tag.startswith('R'):
return wordnet.ADV
else:
return wordnet.NOUN
def normalize_text(text):
word_pos = nltk.pos_tag(nltk.word_tokenize(text))
lemm_words = [lmtzr(sw[0], get_wordnet_pos(sw[1])) for sw in word_pos]
return [x.lower() for x in lemm_words]
print(normalize_text('cats running ran cactus cactuses cacti community communities'))
# ['cat', 'run', 'ran', 'cactus', 'cactuses', 'cacti', 'community', 'community']
print(normalize_text('The cactus ran to the community to see the cats running around cacti between communities.'))
# ['the', 'cactus', 'run', 'to', 'the', 'community', 'to', 'see', 'the', 'cat', 'run', 'around', 'cactus', 'between', 'community', '.']
http://wordnet.princeton.edu/man/morph.3WN
For a lot of my projects, I prefer the lexicon-based WordNet lemmatizer over the more aggressive porter stemming.
http://wordnet.princeton.edu/links#PHP has a link to a PHP interface to the WN APIs.
Look into WordNet, a large lexical database for the English language:
http://wordnet.princeton.edu/
There are APIs for accessing it in several languages.
This looks interesting:
MIT Java WordnetStemmer:
http://projects.csail.mit.edu/jwi/api/edu/mit/jwi/morph/WordnetStemmer.html
Take a look at LemmaGen - open source library written in C# 3.0.
Results for your test words (http://lemmatise.ijs.si/Services)
cats -> cat
running
ran -> run
cactus
cactuses -> cactus
cacti -> cactus
community
communities -> community
The top python packages (in no specific order) for lemmatization are: spacy, nltk, gensim, pattern, CoreNLP and TextBlob. I prefer spaCy and gensim's implementation (based on pattern) because they identify the POS tag of the word and assigns the appropriate lemma automatically. The gives more relevant lemmas, keeping the meaning intact.
If you plan to use nltk or TextBlob, you need to take care of finding the right POS tag manually and the find the right lemma.
Lemmatization Example with spaCy:
# Run below statements in terminal once.
pip install spacy
spacy download en
import spacy
# Initialize spacy 'en' model
nlp = spacy.load('en', disable=['parser', 'ner'])
sentence = "The striped bats are hanging on their feet for best"
# Parse
doc = nlp(sentence)
# Extract the lemma
" ".join([token.lemma_ for token in doc])
#> 'the strip bat be hang on -PRON- foot for good'
Lemmatization Example With Gensim:
from gensim.utils import lemmatize
sentence = "The striped bats were hanging on their feet and ate best fishes"
lemmatized_out = [wd.decode('utf-8').split('/')[0] for wd in lemmatize(sentence)]
#> ['striped', 'bat', 'be', 'hang', 'foot', 'eat', 'best', 'fish']
The above examples were borrowed from in this lemmatization page.
If I may quote my answer to the question StompChicken mentioned:
The core issue here is that stemming algorithms operate on a phonetic basis with no actual understanding of the language they're working with.
As they have no understanding of the language and do not run from a dictionary of terms, they have no way of recognizing and responding appropriately to irregular cases, such as "run"/"ran".
If you need to handle irregular cases, you'll need to either choose a different approach or augment your stemming with your own custom dictionary of corrections to run after the stemmer has done its thing.
The most current version of the stemmer in NLTK is Snowball.
You can find examples on how to use it here:
http://nltk.googlecode.com/svn/trunk/doc/api/nltk.stem.snowball2-pysrc.html#demo
You could use the Morpha stemmer. UW has uploaded morpha stemmer to Maven central if you plan to use it from a Java application. There's a wrapper that makes it much easier to use. You just need to add it as a dependency and use the edu.washington.cs.knowitall.morpha.MorphaStemmer class. Instances are threadsafe (the original JFlex had class fields for local variables unnecessarily). Instantiate a class and run morpha and the word you want to stem.
new MorphaStemmer().morpha("climbed") // goes to "climb"
Do a search for Lucene, im not sure if theres a PHP port but I do know Lucene is available for many platforms. Lucene is an OSS (from Apache) indexing and search library. Naturally it and community extras might have something interesting to look at. At the very least you can learn how it's done in one language so you can translate the "idea" into PHP.
.Net lucene has an inbuilt porter stemmer. You can try that. But note that porter stemming does not consider word context when deriving the lemma. (Go through the algorithm and its implementation and you will see how it works)
Martin Porter wrote Snowball (a language for stemming algorithms) and rewrote the "English Stemmer" in Snowball. There are is an English Stemmer for C and Java.
He explicitly states that the Porter Stemmer has been reimplemented only for historical reasons, so testing stemming correctness against the Porter Stemmer will get you results that you (should) already know.
From http://tartarus.org/~martin/PorterStemmer/index.html (emphasis mine)
The Porter stemmer should be regarded as ‘frozen’, that is, strictly defined, and not amenable to further modification. As a stemmer, it is slightly inferior to the Snowball English or Porter2 stemmer, which derives from it, and which is subjected to occasional improvements. For practical work, therefore, the new Snowball stemmer is recommended. The Porter stemmer is appropriate to IR research work involving stemming where the experiments need to be exactly repeatable.
Dr. Porter suggests to use the English or Porter2 stemmers instead of the Porter stemmer. The English stemmer is what's actually used in the demo site as #StompChicken has answered earlier.
In Java, i use tartargus-snowball to stemming words
Maven:
<dependency>
<groupId>org.apache.lucene</groupId>
<artifactId>lucene-snowball</artifactId>
<version>3.0.3</version>
<scope>test</scope>
</dependency>
Sample code:
SnowballProgram stemmer = new EnglishStemmer();
String[] words = new String[]{
"testing",
"skincare",
"eyecare",
"eye",
"worked",
"read"
};
for (String word : words) {
stemmer.setCurrent(word);
stemmer.stem();
//debug
logger.info("Origin: " + word + " > " + stemmer.getCurrent());// result: test, skincar, eyecar, eye, work, read
}
Try this one here: http://www.twinword.com/lemmatizer.php
I entered your query in the demo "cats running ran cactus cactuses cacti community communities" and got ["cat", "running", "run", "cactus", "cactus", "cactus", "community", "community"] with the optional flag ALL_TOKENS.
Sample Code
This is an API so you can connect to it from any environment. Here is what the PHP REST call may look like.
// These code snippets use an open-source library. http://unirest.io/php
$response = Unirest\Request::post([ENDPOINT],
array(
"X-Mashape-Key" => [API KEY],
"Content-Type" => "application/x-www-form-urlencoded",
"Accept" => "application/json"
),
array(
"text" => "cats running ran cactus cactuses cacti community communities"
)
);
I highly recommend using Spacy (base text parsing & tagging) and Textacy (higher level text processing built on top of Spacy).
Lemmatized words are available by default in Spacy as a token's .lemma_ attribute and text can be lemmatized while doing a lot of other text preprocessing with textacy. For example while creating a bag of terms or words or generally just before performing some processing that requires it.
I'd encourage you to check out both before writing any code, as this may save you a lot of time!
df_plots = pd.read_excel("Plot Summary.xlsx", index_col = 0)
df_plots
# Printing first sentence of first row and last sentence of last row
nltk.sent_tokenize(df_plots.loc[1].Plot)[0] + nltk.sent_tokenize(df_plots.loc[len(df)].Plot)[-1]
# Calculating length of all plots by words
df_plots["Length"] = df_plots.Plot.apply(lambda x :
len(nltk.word_tokenize(x)))
print("Longest plot is for season"),
print(df_plots.Length.idxmax())
print("Shortest plot is for season"),
print(df_plots.Length.idxmin())
#What is this show about? (What are the top 3 words used , excluding the #stop words, in all the #seasons combined)
word_sample = list(["struggled", "died"])
word_list = nltk.pos_tag(word_sample)
[wnl.lemmatize(str(word_list[index][0]), pos = word_list[index][1][0].lower()) for index in range(len(word_list))]
# Figure out the stop words
stop = (stopwords.words('english'))
# Tokenize all the plots
df_plots["Tokenized"] = df_plots.Plot.apply(lambda x : nltk.word_tokenize(x.lower()))
# Remove the stop words
df_plots["Filtered"] = df_plots.Tokenized.apply(lambda x : (word for word in x if word not in stop))
# Lemmatize each word
wnl = WordNetLemmatizer()
df_plots["POS"] = df_plots.Filtered.apply(lambda x : nltk.pos_tag(list(x)))
# df_plots["POS"] = df_plots.POS.apply(lambda x : ((word[1] = word[1][0] for word in word_list) for word_list in x))
df_plots["Lemmatized"] = df_plots.POS.apply(lambda x : (wnl.lemmatize(x[index][0], pos = str(x[index][1][0]).lower()) for index in range(len(list(x)))))
#Which Season had the highest screenplay of "Jesse" compared to "Walt"
#Screenplay of Jesse =(Occurences of "Jesse")/(Occurences of "Jesse"+ #Occurences of "Walt")
df_plots.groupby("Season").Tokenized.sum()
df_plots["Share"] = df_plots.groupby("Season").Tokenized.sum().apply(lambda x : float(x.count("jesse") * 100)/float(x.count("jesse") + x.count("walter") + x.count("walt")))
print("The highest times Jesse was mentioned compared to Walter/Walt was in season"),
print(df_plots["Share"].idxmax())
#float(df_plots.Tokenized.sum().count('jesse')) * 100 / #float((df_plots.Tokenized.sum().count('jesse') + #df_plots.Tokenized.sum().count('walt') + #df_plots.Tokenized.sum().count('walter')))
Related
I'm searching for a Lemmatizer/PoS-tagger for the Italian language, that works on Python. I tried with Spacy, it works but it's not very precise, expecially for verbs it often returns the wrong lemma. NLKT has only english as language. I'm searching for an optimized tool for the Italian language, does it exists?
If it doesn't exist, is it possible, given a corpus, to create it? Whats the work needed to create it?
I also found myself into this problem. I found that one of the best italian lemmatizers is TreeTagger. I preferred it to Spacy's lemmatizer for some projects (I also think that it could be better at POS-tagging). You can also test it online to find out if it is ok for your use case.
I found very useful to use it inside my Spacy pipeline, just for lemmatization, to keep the infrastructure that Spacy provides. This is how you can replace Spacy's lemmatizer with TreeTagger in Python thanks to treetaggerwrapper (you could easily do the same with the POS-tagger):
from treetaggerwrapper import TreeTagger
...
nlp = spacy.load("it_core_news_lg")
TREETAGGER = TreeTagger(TAGDIR="path_to_treetagger", TAGLANG="it")
#Language.component("treetagger")
def treetagger(doc):
tokens = [token.text for token in doc if not token.is_space]
tags = TREETAGGER.tag_text(tokens, tagonly=True)
lemmas = [tag.split("\t")[2].split("|")[0] for tag in tags]
j = 0
for token in doc:
if not token.is_space:
token.lemma_ = lemmas[j]
j += 1
else:
token.lemma_ = " "
return doc
nlp.replace_pipe("lemmatizer", "treetagger")
This could be a useful temporaneous solution.
I am using Spacy lemmatization for preprocessing texts.
doc = 'ups'
for i in nlp(doc):
print(i.lemma_)
>> up
I understand why spacy remove the 's', but it is important for me that in that case, it won't do it. Is there a way to add specific rules to spacy or do I have to use If statements outside the process (which is something I don't want to do )
In Spacy 3 the accepted solution throws an error:
KeyError: "[E159] Can't find table 'lemma_exc' in lookups. Available tables: ['lexeme_norm']"
As the lemmatizer is now a dedicated Spacy component the lookups have to be modified directly at the component (this at least works for me):
nlp.get_pipe('lemmatizer').lookups.get_table("lemma_exc")["noun"]["data"] = ["data"]
Hope this is helpful to someone!
For spacy v2:
Depending on whether you have a tagger, you can customize the rule-based lemmatizer exceptions or the lookup table:
import spacy
# original
nlp = spacy.load("en_core_web_sm", disable=["parser", "ner"])
# may be "up" or "ups" depending on exact version of spacy/model because it
# depends on the POS tag
assert nlp("ups")[0].lemma_ in ("ups", "up")
# 1. Exception for rule-based lemmatizer (with tagger)
# reload to start with a clean lemma cache
nlp = spacy.load("en_core_web_sm", disable=["parser", "ner"])
# add an exception for "ups" as with POS NOUN or VERB
nlp.vocab.lookups.get_table("lemma_exc")["noun"]["ups"] = ["ups"]
nlp.vocab.lookups.get_table("lemma_exc")["verb"]["ups"] = ["ups"]
assert nlp("ups")[0].lemma_ == "ups"
# 2. New entry for lookup lemmatizer (without tagger)
nlp = spacy.load("en_core_web_sm", disable=["tagger", "parser", "ner"])
nlp.vocab.lookups.get_table("lemma_lookup")["ups"] = "ups"
assert nlp("ups")[0].lemma_ == "ups"
If you are processing words in isolation, the tagger is not going to be very reliable (you might get NOUN, PROPN, or VERB for something like ups), so it might be easier to deal with customizing the lookup lemmatizer. The quality of the rule-based lemmas is better overall, but you need at least full phrases, preferably full sentences, to get reasonable results.
The below code breaks the sentence into individual tokens and the output is as below
"cloud" "computing" "is" "benefiting" " major" "manufacturing" "companies"
import en_core_web_sm
nlp = en_core_web_sm.load()
doc = nlp("Cloud computing is benefiting major manufacturing companies")
for token in doc:
print(token.text)
What I would ideally want is, to read 'cloud computing' together as it is technically one word.
Basically I am looking for a bi gram. Is there any feature in Spacy that allows Bi gram or Tri grams ?
Spacy allows the detection of noun chunks. So to parse your noun phrases as single entities do this:
Detect the noun chunks
https://spacy.io/usage/linguistic-features#noun-chunks
Merge the noun chunks
Do dependency parsing again, it would parse "cloud computing" as single entity now.
>>> import spacy
>>> nlp = spacy.load('en')
>>> doc = nlp("Cloud computing is benefiting major manufacturing companies")
>>> list(doc.noun_chunks)
[Cloud computing, major manufacturing companies]
>>> for noun_phrase in list(doc.noun_chunks):
... noun_phrase.merge(noun_phrase.root.tag_, noun_phrase.root.lemma_, noun_phrase.root.ent_type_)
...
Cloud computing
major manufacturing companies
>>> [(token.text,token.pos_) for token in doc]
[('Cloud computing', 'NOUN'), ('is', 'VERB'), ('benefiting', 'VERB'), ('major manufacturing companies', 'NOUN')]
If you have a spacy doc, you can pass it to textacy:
ngrams = list(textacy.extract.basics.ngrams(doc, 2, min_freq=2))
Warning: This is just an extension of the right answer made by Zuzana.
My reputation does not allow me to comment so I am making this answer just to answer the question of Adit Sanghvi above: "How do you do it when you have a list of documents?"
First you need to create a list with the text of the documents
Then you join the text lists in just one document
now you use the spacy parser to transform the text document in a Spacy document
You use the Zuzana's answer's to create de bigrams
This is the example code:
Step 1
doc1 = ['all what i want is that you give me back my code because i worked a lot on it. Just give me back my code']
doc2 = ['how are you? i am just showing you an example of how to make bigrams on spacy. We love bigrams on spacy']
doc3 = ['i love to repeat phrases to make bigrams because i love make bigrams']
listOfDocuments = [doc1,doc2,doc3]
textList = [''.join(textList) for text in listOfDocuments for textList in text]
print(textList)
This will print this text:
['all what i want is that you give me back my code because i worked a lot on it. Just give me back my code', 'how are you? i am just showing you an example of how to make bigrams on spacy. We love bigrams on spacy', 'i love to repeat phrases to make bigrams because i love make bigrams']
then step 2 and 3:
doc = ' '.join(textList)
spacy_doc = parser(doc)
print(spacy_doc)
and will print this:
all what i want is that you give me back my code because i worked a lot on it. Just give me back my code how are you? i am just showing you an example of how to make bigrams on spacy. We love bigrams on spacy i love to repeat phrases to make bigrams because i love make bigrams
Finally step 4 (Zuzana's answer)
ngrams = list(textacy.extract.ngrams(spacy_doc, 2, min_freq=2))
print(ngrams)
will print this:
[make bigrams, make bigrams, make bigrams]
I had a similar problem (bigrams, trigrams, like your "cloud computing"). I made a simple list of the n-grams, word_3gram, word_2grams etc., with the gram as basic unit (cloud_computing).
Assume I have the sentence "I like cloud computing because it's cheap". The sentence_2gram is: "I_like", "like_cloud", "cloud_computing", "computing_because" ... Comparing that your bigram list only "cloud_computing" is recognized as a valid bigram; all other bigrams in the sentence are artificial. To recover all other words you just take the first part of the other words,
"I_like".split("_")[0] -> I;
"like_cloud".split("_")[0] -> like
"cloud_computing" -> in bigram list, keep it.
skip next bi-gram "computing_because" ("computing" is already used)
"because_it's".split("_")[0]" -> "because" etc.
To also capture the last word in the sentence ("cheap") I added the token "EOL". I implemented this in python, and the speed was OK (500k words in 3min), i5 processor with 8G. Anyway, you have to do it only once. I find this more intuitive than the official (spacy-style) chunk approach. It also works for non-spacy frameworks.
I do this before the official tokenization/lemmatization, as you would get "cloud compute" as possible bigram. But I'm not certain if this is the best/right approach.
I'm looking for a way to pos_tag a French sentence like the following code is used for English sentences:
def pos_tagging(sentence):
var = sentence
exampleArray = [var]
for item in exampleArray:
tokenized = nltk.word_tokenize(item)
tagged = nltk.pos_tag(tokenized)
return tagged
here is the full code source it works very well
download link for Standford NLP https://nlp.stanford.edu/software/tagger.shtml#About
from nltk.tag import StanfordPOSTagger
jar = 'C:/Users/m.ferhat/Desktop/stanford-postagger-full-2016-10-31/stanford-postagger-3.7.0.jar'
model = 'C:/Users/m.ferhat/Desktop/stanford-postagger-full-2016-10-31/models/french.tagger'
import os
java_path = "C:/Program Files/Java/jdk1.8.0_121/bin/java.exe"
os.environ['JAVAHOME'] = java_path
pos_tagger = StanfordPOSTagger(model, jar, encoding='utf8' )
res = pos_tagger.tag('je suis libre'.split())
print (res)
The NLTK doesn't come with pre-built resources for French. I recommend using the Stanford tagger, which comes with a trained French model. This code shows how you might set up the nltk for use with Stanford's French POS tagger. Note that the code is outdated (and for Python 2), but you could use it as a starting point.
Alternately, the NLTK makes it very easy to train your own POS tagger on a tagged corpus, and save it for later use. If you have access to a (sufficiently large) French corpus, you can follow the instructions in the nltk book and simply use your corpus in place of the Brown corpus. You're unlikely to match the performance of the Stanford tagger (unless you can train a tagger for your specific domain), but you won't have to install anything.
Right now I am using similar() method from nltk.
But is is not working as expected. Please see below piece of code:
from nltk import word_tokenize;
import nltk;
text = """
The girl is very pretty.
""";
text = nltk.Text(word_tokenize(text));
text.similar('beautiful'); #it returns "no matches" but pretty is synonym of beautiful.
Am I using wrong approach? Or is there any other? Please help me.
The NLTK Text class' similar() method uses Distributional Similarity.
The help() on the method states:
similar(word, num=20) method of nltk.text.Text instance
Distributional similarity: find other words which appear in the
same contexts as the specified word; list most similar words first.
Looking in the source code, similar() uses an instantiation of the ContextIndex class to find words with similar semantic windows. By default, it uses a +/- 1 word window.
If we extend your example with additional words to give similar semantic windows for "pretty" and "beautiful", we will get the result you are looking for.
from nltk import word_tokenize
import nltk
text = "The girl is pretty isn't she? The girl is beautiful isn't she?"
text = nltk.Text(word_tokenize(text))
text.similar('pretty')
# prints beautiful
So it seems you need to have more context in your text to give meaningful results.