A homograph is a word that has the same spelling as another word but has a different sound and a different meaning, for example,lead (to go in front of) / lead (a metal) .
I was trying to use spacy word vectors to compare documents with each other by summing each word vector for each document and then finally finding cosine similarity. If for example spacy vectors have the same vector for the two 'lead' listed above , the results will be probably bad.
In the code below , why does the similarity between the two 'bank'
tokens come out as 1.00 ?
import spacy
nlp = spacy.load('en')
str1 = 'The guy went inside the bank to take out some money'
str2 = 'The house by the river bank.'
str1_tokenized = nlp(str1.decode('utf8'))
str2_tokenized = nlp(str2.decode('utf8'))
token1 = str1_tokenized[-6]
token2 = str2_tokenized[-2]
print 'token1 = {} token2 = {}'.format(token1,token2)
print token1.similarity(token2)
The output for given program is
token1 = bank token2 = bank
1.0
As kntgu already pointed out, spaCy distinguishes tokens by their characters, not by their semantic meaning. The sense2vec approach by the developers of spaCy concatenates tokens with their POS-tag and can help in the case of 'lead_VERB' vs. 'lead_NOUN'. However, it will not help in your example of 'bank (river bank)' vs. 'bank (financial institute)', as both are nouns.
SpaCy does not support any solution to this out of the box, but you can have a look at contextualized word representations like ELMo or BERT. Both generate word vectors for a given sentence, taking the context into account. Therefore, I assume the vectors for both 'bank' tokens will be substantially different.
Both are relatively recent approaches and are not as comfortable to use, but might help in your use case. For ELMo, there is a command line tool which lets you generate word embeddings for a set of sentences without having to write any code: https://github.com/allenai/allennlp/blob/master/tutorials/how_to/elmo.md#writing-contextual-representations-to-disk
I have a thousands of sentences about events happened in the past. E.g.
sentence1 = 'The Knights Templar are founded to protect Christian pilgrims in Jerusalem.'
sentence2 = 'Alfonso VI of Castile captures the Moorish Muslim city of Toledo, Spain.'
sentence3 = 'The Hindu Medang kingdom flourishes and declines.'
I want to transform them into questions of the form:
question1 = 'When were the Knights Templar founded to protect Christian pilgrims in Jerusalem?'
question2 = 'When did Alfonso VI of Castile capture the Moorish Muslim city of Toledo, Spain?'
question3 = 'When did the Hindu Medang kingdom flourish and decline?'
I realize that this is a complex problem and I am ok with a success rate of 80%.
As far as I understand from searches on the web NTLK is the way to go for this kind of problems.
I started to try some things but it is the first time I use this library and I cannot go much further than this:
import nltk
question = 'The Knights Templar are founded to protect Christian pilgrims in Jerusalem.'
tokens = nltk.word_tokenize(question)
tagged = nltk.pos_tag(tokens)
This sounds like a problem many people must have encountered and solved.
Any suggestions?
NLTK can definitely be the right tool to use here. But the quality of your tokenizer and pos-tagger output depends on your corpus and type of sentences. Also, there is usually not really an out-of-the-box solution to this (afaik), and it requires some tuning. If you don't have very much time to put into this, I doubt that your success rate will even reach 80%.
Having said that; here's a basic list instertion based example that may help you to capture and succesfully convert some of your sentences.
import nltk
question_one = 'The Knights Templar are founded to protect Christian pilgrims in Jerusalem.'
question_two = 'Alfonso VI of Castile captures the Moorish Muslim city of Toledo, Spain.'
def modify(inputStr):
tokens = nltk.PunktWordTokenizer().tokenize(inputStr)
tagged = nltk.pos_tag(tokens)
auxiliary_verbs = [i for i, w in enumerate(tagged) if w[1] == 'VBP']
if auxiliary_verbs:
tagged.insert(0, tagged.pop(auxiliary_verbs[0]))
else:
tagged.insert(0, ('did', 'VBD'))
tagged.insert(0, ('When', 'WRB'))
return ' '.join([t[0] for t in tagged])
question_one = modify(question_one)
question_two = modify(question_two)
print(question_one)
print(question_two)
Output:
When are The Knights Templar founded to protect Christian pilgrims in Jerusalem.
When did Alfonso VI of Castile captures the Moorish Muslim city of Toledo , Spain.
As you can see, you'd still need to fix correct casing ('The' is still uppercase), 'captures' is in the wrong tense now and you will want to expand on auxiliary_verbs types (probably 'VBP' alone is too limited). But it's a start. Hope this helps!
This question already has answers here:
How can I split a text into sentences?
(20 answers)
Closed 3 years ago.
I want to make a list of sentences from a string and then print them out. I don't want to use NLTK to do this. So it needs to split on a period at the end of the sentence and not at decimals or abbreviations or title of a name or if the sentence has a .com This is attempt at regex that doesn't work.
import re
text = """\
Mr. Smith bought cheapsite.com for 1.5 million dollars, i.e. he paid a lot for it. Did he mind? Adam Jones Jr. thinks he didn't. In any case, this isn't true... Well, with a probability of .9 it isn't.
"""
sentences = re.split(r' *[\.\?!][\'"\)\]]* *', text)
for stuff in sentences:
print(stuff)
Example output of what it should look like
Mr. Smith bought cheapsite.com for 1.5 million dollars, i.e. he paid a lot for it.
Did he mind?
Adam Jones Jr. thinks he didn't.
In any case, this isn't true...
Well, with a probability of .9 it isn't.
(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?)\s
Try this. split your string this.You can also check demo.
http://regex101.com/r/nG1gU7/27
Ok so sentence-tokenizers are something I looked at in a little detail, using regexes, nltk, CoreNLP, spaCy. You end up writing your own and it depends on the application. This stuff is tricky and valuable and people don't just give their tokenizer code away. (Ultimately, tokenization is not a deterministic procedure, it's probabilistic, and also depends very heavily on your corpus or domain, e.g. legal/financial documents vs social-media posts vs Yelp reviews vs biomedical papers...)
In general you can't rely on one single Great White infallible regex, you have to write a function which uses several regexes (both positive and negative); also a dictionary of abbreviations, and some basic language parsing which knows that e.g. 'I', 'USA', 'FCC', 'TARP' are capitalized in English.
To illustrate how easily this can get seriously complicated, let's try to write you that functional spec for a deterministic tokenizer just to decide whether single or multiple period ('.'/'...') indicates end-of-sentence, or something else:
function isEndOfSentence(leftContext, rightContext)
Return False for decimals inside numbers or currency e.g. 1.23 , $1.23, "That's just my $.02" Consider also section references like 1.2.A.3.a, European date formats like 09.07.2014, IP addresses like 192.168.1.1, MAC addresses...
Return False (and don't tokenize into individual letters) for known abbreviations e.g. "U.S. stocks are falling" ; this requires a dictionary of known abbreviations. Anything outside that dictionary you will get wrong, unless you add code to detect unknown abbreviations like A.B.C. and add them to a list.
Ellipses '...' at ends of sentences are terminal, but in the middle of sentences are not. This is not as easy as you might think: you need to look at the left context and the right context, specifically is the RHS capitalized and again consider capitalized words like 'I' and abbreviations. Here's an example proving ambiguity which : She asked me to stay... I left an hour later. (Was that one sentence or two? Impossible to determine)
You may also want to write a few patterns to detect and reject miscellaneous non-sentence-ending uses of punctuation: emoticons :-), ASCII art, spaced ellipses . . . and other stuff esp. Twitter. (Making that adaptive is even harder). How do we tell if #midnight is a Twitter user, the show on Comedy Central, text shorthand, or simply unwanted/junk/typo punctuation? Seriously non-trivial.
After you handle all those negative cases, you could arbitrarily say that any isolated period followed by whitespace is likely to be an end of sentence. (Ultimately, if you really want to buy extra accuracy, you end up writing your own probabilistic sentence-tokenizer which uses weights, and training it on a specific corpus(e.g. legal texts, broadcast media, StackOverflow, Twitter, forums comments etc.)) Then you have to manually review exemplars and training errors. See Manning and Jurafsky book or Coursera course [a].
Ultimately you get as much correctness as you are prepared to pay for.
All of the above is clearly specific to the English-language/ abbreviations, US number/time/date formats. If you want to make it country- and language-independent, that's a bigger proposition, you'll need corpora, native-speaking people to label and QA it all, etc.
All of the above is still only ASCII, which is practically speaking only 96 characters. Allow the input to be Unicode, and things get harder still (and the training-set necessarily must be either much bigger or much sparser)
In the simple (deterministic) case, function isEndOfSentence(leftContext, rightContext) would return boolean, but in the more general sense, it's probabilistic: it returns a float 0.0-1.0 (confidence level that that particular '.' is a sentence end).
References: [a] Coursera video: "Basic Text Processing 2-5 - Sentence Segmentation - Stanford NLP - Professor Dan Jurafsky & Chris Manning" [UPDATE: an unofficial version used to be on YouTube, was taken down]
Try to split the input according to the spaces rather than a dot or ?, if you do like this then the dot or ? won't be printed in the final result.
>>> import re
>>> s = """Mr. Smith bought cheapsite.com for 1.5 million dollars, i.e. he paid a lot for it. Did he mind? Adam Jones Jr. thinks he didn't. In any case, this isn't true... Well, with a probability of .9 it isn't."""
>>> m = re.split(r'(?<=[^A-Z].[.?]) +(?=[A-Z])', s)
>>> for i in m:
... print i
...
Mr. Smith bought cheapsite.com for 1.5 million dollars, i.e. he paid a lot for it.
Did he mind?
Adam Jones Jr. thinks he didn't.
In any case, this isn't true...
Well, with a probability of .9 it isn't.
sent = re.split('(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?)(\s|[A-Z].*)',text)
for s in sent:
print s
Here the regex used is : (?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?)(\s|[A-Z].*)
First block: (?<!\w\.\w.) : this pattern searches in a negative feedback loop (?<!) for all words (\w) followed by fullstop (\.) , followed by other words (\.)
Second block: (?<![A-Z][a-z]\.): this pattern searches in a negative feedback loop for anything starting with uppercase alphabets ([A-Z]), followed by lower case alphabets ([a-z]) till a dot (\.) is found.
Third block: (?<=\.|\?): this pattern searches in a feedback loop of dot (\.) OR question mark (\?)
Fourth block: (\s|[A-Z].*): this pattern searches after the dot OR question mark from the third block. It searches for blank space (\s) OR any sequence of characters starting with a upper case alphabet ([A-Z].*).
This block is important to split if the input is as
Hello world.Hi I am here today.
i.e. if there is space or no space after the dot.
Naive approach for proper english sentences not starting with non-alphas and not containing quoted parts of speech:
import re
text = """\
Mr. Smith bought cheapsite.com for 1.5 million dollars, i.e. he paid a lot for it. Did he mind? Adam Jones Jr. thinks he didn't. In any case, this isn't true... Well, with a probability of .9 it isn't.
"""
EndPunctuation = re.compile(r'([\.\?\!]\s+)')
NonEndings = re.compile(r'(?:Mrs?|Jr|i\.e)\.\s*$')
parts = EndPunctuation.split(text)
sentence = []
for part in parts:
if len(part) and len(sentence) and EndPunctuation.match(sentence[-1]) and not NonEndings.search(''.join(sentence)):
print(''.join(sentence))
sentence = []
if len(part):
sentence.append(part)
if len(sentence):
print(''.join(sentence))
False positive splitting may be reduced by extending NonEndings a bit. Other cases will require additional code. Handling typos in a sensible way will prove difficult with this approach.
You will never reach perfection with this approach. But depending on the task it might just work "enough"...
I'm not great at regular expressions, but a simpler version, "brute force" actually, of above is
sentence = re.compile("([\'\"][A-Z]|([A-Z][a-z]*\. )|[A-Z])(([a-z]*\.[a-z]*\.)|([A-Za-z0-9]*\.[A-Za-z0-9])|([A-Z][a-z]*\. [A-Za-z]*)|[^\.?]|[A-Za-z])*[\.?]")
which means
start acceptable units are '[A-Z] or "[A-Z]
please note, most regular expressions are greedy so the order is very important when we do |(or). That's, why I have written i.e. regular expression first, then is come forms like Inc.
Try this:
(?<!\b(?:[A-Z][a-z]|\d|[i.e]))\.(?!\b(?:com|\d+)\b)
I wrote this taking into consideration smci's comments above. It is a middle-of-the-road approach that doesn't require external libraries and doesn't use regex. It allows you to provide a list of abbreviations and accounts for sentences ended by terminators in wrappers, such as a period and quote: [.", ?', .)].
abbreviations = {'dr.': 'doctor', 'mr.': 'mister', 'bro.': 'brother', 'bro': 'brother', 'mrs.': 'mistress', 'ms.': 'miss', 'jr.': 'junior', 'sr.': 'senior', 'i.e.': 'for example', 'e.g.': 'for example', 'vs.': 'versus'}
terminators = ['.', '!', '?']
wrappers = ['"', "'", ')', ']', '}']
def find_sentences(paragraph):
end = True
sentences = []
while end > -1:
end = find_sentence_end(paragraph)
if end > -1:
sentences.append(paragraph[end:].strip())
paragraph = paragraph[:end]
sentences.append(paragraph)
sentences.reverse()
return sentences
def find_sentence_end(paragraph):
[possible_endings, contraction_locations] = [[], []]
contractions = abbreviations.keys()
sentence_terminators = terminators + [terminator + wrapper for wrapper in wrappers for terminator in terminators]
for sentence_terminator in sentence_terminators:
t_indices = list(find_all(paragraph, sentence_terminator))
possible_endings.extend(([] if not len(t_indices) else [[i, len(sentence_terminator)] for i in t_indices]))
for contraction in contractions:
c_indices = list(find_all(paragraph, contraction))
contraction_locations.extend(([] if not len(c_indices) else [i + len(contraction) for i in c_indices]))
possible_endings = [pe for pe in possible_endings if pe[0] + pe[1] not in contraction_locations]
if len(paragraph) in [pe[0] + pe[1] for pe in possible_endings]:
max_end_start = max([pe[0] for pe in possible_endings])
possible_endings = [pe for pe in possible_endings if pe[0] != max_end_start]
possible_endings = [pe[0] + pe[1] for pe in possible_endings if sum(pe) > len(paragraph) or (sum(pe) < len(paragraph) and paragraph[sum(pe)] == ' ')]
end = (-1 if not len(possible_endings) else max(possible_endings))
return end
def find_all(a_str, sub):
start = 0
while True:
start = a_str.find(sub, start)
if start == -1:
return
yield start
start += len(sub)
I used Karl's find_all function from this entry: Find all occurrences of a substring in Python
My example is based on the example of Ali, adapted to Brazilian Portuguese. Thanks Ali.
ABREVIACOES = ['sra?s?', 'exm[ao]s?', 'ns?', 'nos?', 'doc', 'ac', 'publ', 'ex', 'lv', 'vlr?', 'vls?',
'exmo(a)', 'ilmo(a)', 'av', 'of', 'min', 'livr?', 'co?ls?', 'univ', 'resp', 'cli', 'lb',
'dra?s?', '[a-z]+r\(as?\)', 'ed', 'pa?g', 'cod', 'prof', 'op', 'plan', 'edf?', 'func', 'ch',
'arts?', 'artigs?', 'artg', 'pars?', 'rel', 'tel', 'res', '[a-z]', 'vls?', 'gab', 'bel',
'ilm[oa]', 'parc', 'proc', 'adv', 'vols?', 'cels?', 'pp', 'ex[ao]', 'eg', 'pl', 'ref',
'[0-9]+', 'reg', 'f[ilí]s?', 'inc', 'par', 'alin', 'fts', 'publ?', 'ex', 'v. em', 'v.rev']
ABREVIACOES_RGX = re.compile(r'(?:{})\.\s*$'.format('|\s'.join(ABREVIACOES)), re.IGNORECASE)
def sentencas(texto, min_len=5):
# baseado em https://stackoverflow.com/questions/25735644/python-regex-for-splitting-text-into-sentences-sentence-tokenizing
texto = re.sub(r'\s\s+', ' ', texto)
EndPunctuation = re.compile(r'([\.\?\!]\s+)')
# print(NonEndings)
parts = EndPunctuation.split(texto)
sentencas = []
sentence = []
for part in parts:
txt_sent = ''.join(sentence)
q_len = len(txt_sent)
if len(part) and len(sentence) and q_len >= min_len and \
EndPunctuation.match(sentence[-1]) and \
not ABREVIACOES_RGX.search(txt_sent):
sentencas.append(txt_sent)
sentence = []
if len(part):
sentence.append(part)
if sentence:
sentencas.append(''.join(sentence))
return sentencas
Full code in: https://github.com/luizanisio/comparador_elastic
If you want to break up sentences at 3 periods (not sure if this is what you want) you can use this regular expresion:
import re
text = """\
Mr. Smith bought cheapsite.com for 1.5 million dollars, i.e. he paid a lot for it. Did he mind? Adam Jones Jr. thinks he didn't. In any case, this isn't true... Well, with a probability of .9 it isn't.
"""
sentences = re.split(r'\.{3}', text)
for stuff in sentences:
print(stuff)
I never really dealt with NLP but had an idea about NER which should NOT have worked and somehow DOES exceptionally well in one case. I do not understand why it works, why doesn't it work or weather it can be extended.
The idea was to extract names of the main characters in a story through:
Building a dictionary for each word
Filling for each word a list with the words that appear right next to it in the text
Finding for each word a word with the max correlation of lists (meaning that the words are used similarly in the text)
Given that one name of a character in the story, the words that are used like it, should be as well (Bogus, that is what should not work but since I never dealt with NLP until this morning I started the day naive)
I ran the overly simple code (attached below) on Alice in Wonderland, which for "Alice" returns:
21 ['Mouse', 'Latitude', 'William', 'Rabbit', 'Dodo', 'Gryphon', 'Crab', 'Queen', 'Duchess', 'Footman', 'Panther', 'Caterpillar', 'Hearts', 'King', 'Bill', 'Pigeon', 'Cat', 'Hatter', 'Hare', 'Turtle', 'Dormouse']
Though it filters for upper case words (and receives "Alice" as the word to cluster around), originally there are ~500 upper case words, and it's still pretty spot on as far as main characters goes.
It does not work that well with other characters and in other stories, though gives interesting results.
Any idea if this idea is usable, extendable or why does it work at all in this story for "Alice" ?
Thanks!
#English Name recognition
import re
import sys
import random
from string import upper
def mimic_dict(filename):
dict = {}
f = open(filename)
text = f.read()
f.close()
prev = ""
words = text.split()
for word in words:
m = re.search("\w+",word)
if m == None:
continue
word = m.group()
if not prev in dict:
dict[prev] = [word]
else :
dict[prev] = dict[prev] + [word]
prev = word
return dict
def main():
if len(sys.argv) != 2:
print 'usage: ./main.py file-to-read'
sys.exit(1)
dict = mimic_dict(sys.argv[1])
upper = []
for e in dict.keys():
if len(e) > 1 and e[0].isupper():
upper.append(e)
print len(upper),upper
exclude = ["ME","Yes","English","Which","When","WOULD","ONE","THAT","That","Here","and","And","it","It","me"]
exclude = [ x for x in exclude if dict.has_key(x)]
for s in exclude :
del dict[s]
scores = {}
for key1 in dict.keys():
max = 0
for key2 in dict.keys():
if key1 == key2 : continue
a = dict[key1]
k = dict[key2]
diff = []
for ia in a:
if ia in k and ia not in diff:
diff.append( ia)
if len(diff) > max:
max = len(diff)
scores[key1]=(key2,max)
dictscores = {}
names = []
for e in scores.keys():
if scores[e][0]=="Alice" and e[0].isupper():
names.append(e)
print len(names), names
if __name__ == '__main__':
main()
From the looks of your program and previous experience with NER, I'd say this "works" because you're not doing a proper evaluation. You've found "Hare" where you should have found "March Hare".
The difficulty in NER (at least for English) is not finding the names; it's detecting their full extent (the "March Hare" example); detecting them even at the start of a sentence, where all words are capitalized; classifying them as person/organisation/location/etc.
Also, Alice in Wonderland, being a children's novel, is a rather easy text to process. Newswire phrases like "Microsoft CEO Steve Ballmer" pose a much harder problem; here, you'd want to detect
[ORG Microsoft] CEO [PER Steve Ballmer]
What you are doing is building a distributional thesaurus-- finding words which are distributionally similar to a query (e.g. Alice), i.e. words that appear in similar contexts. This does not automatically make them synonyms, but means they are in a way similar to the query. The fact that your query is a named entity does not on its own guarantee that the similar words that you retrieve will be named entities. However, since Alice, the Hare and the Queen tend to appear is similar context because they share some characteristics (e.g. they all speak, walk, cry, etc-- the details of Alice in wonderland escape me) they are more likely to be retrieved. It turns out whether a word is capitalised or not is a very useful piece of information when working out if something is a named entity. If you do not filter out the non-capitalised words, you will see many other neighbours that are not named entities.
Have a look at the following papers to get an idea of what people do with distributional semantics:
Lin 1998
Grefenstette 1994
Schuetze 1998
To put your idea in the terminology used in these papers, Step 2 is building a context vector for the word with from a window of size 1. Step 3 resembles several well-known similarity measures in distributional semantics (most notably the so-called Jaccard coefficient).
As larsmans pointed out, this seems to work so well because you are not doing a proper evaluation. If you ran this against a hand-annotated corpus you will find it is very bad at identifying the boundaries of names entities and it does not even attempt to guess if they are people or places or organisations... Nevertheless, it is a great first attempt at NLP, keep it up!