from jyquickhelper import add_notebook_menu
add_notebook_menu()
Le traitement automatique des langues (ou Natural Language Processing) propose un ensemble de méthodes permettant (entre autres) :
Le traitement automatique des langues a fait ses premiers pas dans le contexte de la guerre froide, où la traduction automatique était devenu un enjeu geopolitique. En 1950, dans sont article « Computing machinery and intelligence », Alan Turing défini ce qui est appellé plus tartd, le test de Turing. On dit qu'un programme passe le test de Turing s'il parvient à personnifier un humain dans une conversation écrite en temps réel, de façon suffisamment convaincante pour que l'interlocuteur humain ne puisse pas distinguer avec certitude — sur la base du seul contenu de la conversation — s'il interagit avec un programme ou avec un autre humain.
Les progrès en traitement automatique des langues ont été beaucoup plus lents qu'initialement prévus. Cependant certains considèrent que pour la première fois en 2014, grâce aux progrès en machine learning une machine a passé le test en se faisant passer pour un enfant de 13 ans.
L'objet de ce TD est de présenter l'essentiel du traitement automatique des langues, selon trois approches :
L'approche bag of words : on ne tient pas compte de l'ordre des mots, ni du contexte dans lequel ils interviennent (ou alors de manière très partielle, en étudiant par exemple le mot suivant). L'idée est d'étudier la fréquence des mots d'un document et la surreprésentation des mots par rapport à un document de référence (appelé corpus). Cette approche un peu simpliste mais très efficace : on peut calculer des scores permettant par exemple de faire de classification automatique de document par thème, de comparer la similarité de deux documents. Elle est souvent utilisée en première analyse, et elle reste la référence pour l'analyse de textes mal structurés (tweets, dialogue tchat, etc.) Mot-clés : td-idf, indice de similarité cosine
L'approche contextuelle : on s'intéresse non seulement aux mots et à leur fréquence, mais aussi aux mots qui suivent. Cette approche est essentielle pour désambiguiser les homonymes. Elle permet aussi d'affiner les modèles "bag-of-words". Le calcul de n-grams (bigrams pour les co-occurences de mots deux-à-deux, tri-grams pour les co-occurences trois-à-trois, etc.) constitue la méthode la plus simple pour tenir compte du contexte.
L'approche structurelle : on s'intéresse à la structure des phrases, des mots (stemming, lemmatisation), aux règles syntaxiques, au sens des phrases. L'idée est d'introduire de la structure dans l'analyse du langage, à partir de règles connues et modélisées (par des expressions régulières, ou formalisation des règles syntaxiques), enrichies manuellement par des contributeurs, ou apprises par des méthodes de machine learning. Mots-clés : tokenisation des phrases et des mots, Part-Of-Speech tagging, extraction d'entité etc. Cette approche est beaucoup plus coûteuse et longue à mettre en place, mais c'est la seule capable de répondre à des besoins de traitement automatique des langues plus ambitieux tels que la traduction automatique, les agents conversationnels, et permet d'augmenter la performance des modèles de classifications de documents, de prédiction du sentiment, etc.
import httplib2 # pip install httplib2
import json # déjà installée, sinon : pip install json
import apiclient.discovery # pip install google-api-python-client
import bs4 # déjà ja installée, sinon : pip install bs4
import nltk # pip install nltk --> sous Windows, il faut aller à http://www.lfd.uci.edu/~gohlke/pythonlibs/
nltk.__version__
'3.4.1'
google a arrêté Google +, l'API est désactivée également vers le 1er mars 2019, le code suivant ne fonctionne pas mais les données récupérées sont toujours disponibles : échantillon googleplus.
Pour obtenir une clé d'API (google plus ou autre), il faut :
#remplacer par VOTRE clé
import os
try:
from pyquickhelper.loghelper import get_password
API_KEY = get_password("gapi", "ensae_teaching_cs,key")
except Exception as e:
print(e)
if False: # à remplacer par une autre API
# Entrer le nom d'une personne ayant un compte google plus public
Q = "Tim O'Reilly"
# Se connecter à l'API (méthode Oauth2)
service = apiclient.discovery.build('plus', 'v1', http=httplib2.Http(),
developerKey=API_KEY)
# Récupérer les feeds
people_feed = service.people().search(query=Q).execute()
# Imprimer le json récupéré
res = json.dumps(people_feed['items'], indent=1)
print(res if len(res) < 1000 else res[:1000] + "...")
[
{
"kind": "plus#person",
"etag": "\"Sh4n9u6EtD24TM0RmWv7jTXojqc/tjedXFyeIkzudZzRey5EJb8iZIk\"",
"objectType": "person",
"id": "107033731246200681024",
"displayName": "Tim O'Reilly",
"url": "https://plus.google.com/107033731246200681024",
"image": {
"url": "https://lh4.googleusercontent.com/-J8nmMwIhpiA/AAAAAAAAAAI/AAAAAAADdg4/68r2hyFUgzI/photo.jpg?sz=50"
}
},
{
"kind": "plus#person",
"etag": "\"Sh4n9u6EtD24TM0RmWv7jTXojqc/ofg-30rIv-rKw7XTBBnDA1i3I_Y\"",
"objectType": "person",
"id": "110160587587635791009",
"displayName": "TIM O'REILLY",
"url": "https://plus.google.com/110160587587635791009",
"image": {
"url": "https://lh4.googleusercontent.com/-gWq9vr_JEnc/AAAAAAAAAAI/AAAAAAAAADI/zwCXKP4QeiU/photo.jpg?sz=50"
}
},
{
"kind": "plus#person",
"etag": "\"Sh4n9u6EtD24TM0RmWv7jTXojqc/DVTuV3GDJ0h4UlM5bybS_d26Fdo\"",
"objectType": "person",
"id": "106492472890341598734",
"displayName": "Tim O'Reilly",
"url": "https://plus.google.com/10649...
if False: # à remplacer par une autre API
# Parce que l'on travaille sur un Notebook il est possible d'afficher facilement les images correspondantes
# l'identifiant unique d'avatar google plus et le nom
from IPython.core.display import HTML
html = []
for p in people_feed['items']:
html += ['<p><img src="{}" /> {}: {}</p>'.format(p['image']['url'], p['id'], p['displayName'])]
HTML(''.join(html[:5]))
if False: # à remplacer par une autre API
USER_ID = '107033731246200681024'
activity_feed = service.activities().list(
userId=USER_ID,
collection='public',
maxResults='100' # Max allowed per API
).execute()
res = json.dumps(activity_feed, indent=1)
print(res if len(res) < 1000 else res[:1000] + "...")
{
"kind": "plus#activityFeed",
"etag": "\"Sh4n9u6EtD24TM0RmWv7jTXojqc/UVhLnzZeFbRMD00k0VRD5tkC6es\"",
"nextPageToken": "ADSJ_i32R0IpxThTClWgVQ71un8FkJDHG8Pl4hLCvWIbyb6T65r6coxSlWk1svDgsrzxTQ3JHFV1CGnbjFCSaY14sttcvnb1QgiHBgXRtn3A8GjJjin7",
"title": "Google+ List of Activities for Collection PUBLIC",
"updated": "2017-09-13T15:59:45.234Z",
"items": [
{
"kind": "plus#activity",
"etag": "\"Sh4n9u6EtD24TM0RmWv7jTXojqc/Dlr_44FOo97cNjKbX7ZHrVWgen4\"",
"title": "It looks like #@CTRLLabsCo has made a real breakthrough. This is one of the advances that will take ...",
"published": "2017-09-13T15:59:31.577Z",
"updated": "2017-09-13T15:59:45.234Z",
"id": "z123e5zb4zbmcxf5004chl3pvxfbszirt5o",
"url": "https://plus.google.com/+TimOReilly/posts/TpYYyGh7pr1",
"actor": {
"id": "107033731246200681024",
"displayName": "Tim O'Reilly",
"url": "https://plus.google.com/107033731246200681024",
"image": {
"url": "https://lh4.googleusercontent.com/-J8nmMwIhpiA...
import json
with open("ressources_googleplus/107033731246200681024.json", "r", encoding="utf-8") as f:
activity_feed = json.load(f)
res = json.dumps(activity_feed, indent=1)
print(res if len(res) < 1000 else res[:1000] + "...")
[ { "kind": "plus#activity", "etag": "\"Sh4n9u6EtD24TM0RmWv7jTXojqc/Dlr_44FOo97cNjKbX7ZHrVWgen4\"", "title": "It looks like #@CTRLLabsCo has made a real breakthrough. This is one of the advances that will take ...", "published": "2017-09-13T15:59:31.577Z", "updated": "2017-09-13T15:59:45.234Z", "id": "z123e5zb4zbmcxf5004chl3pvxfbszirt5o", "url": "https://plus.google.com/+TimOReilly/posts/TpYYyGh7pr1", "actor": { "id": "107033731246200681024", "displayName": "Tim O'Reilly", "url": "https://plus.google.com/107033731246200681024", "image": { "url": "https://lh4.googleusercontent.com/-J8nmMwIhpiA/AAAAAAAAAAI/AAAAAAADdg4/68r2hyFUgzI/photo.jpg?sz=50" }, "verification": { "adHocVerified": "PASSED" } }, "verb": "post", "object": { "objectType": "note", "actor": { "verification": { "adHocVerified": "PASSED" } }, "content": "It looks like #@CTRLLabsCo <b>has</b> made a real breakthrough. This is one of the advances that...
from bs4 import BeautifulSoup
def cleanHtml(html):
if html == "":
return ""
return BeautifulSoup(html, 'html.parser').get_text()
try:
print(activity_feed[0]['object']['content'])
print("\n")
print(cleanHtml(activity_feed[0]['object']['content']))
except Exception as e:
print(e)
It looks like #@CTRLLabsCo <b>has</b> made a real breakthrough. This is one of the advances that will take us beyond the smartphone. If you've done any playing around with augmented reality, you realize that control and interaction is one of the key blocks to widespread adoption. Brain-computer interfaces are a key combinatorial innovation that will enable a completely different UI paradigm for interacting with devices and ambient computing. It looks like #@CTRLLabsCo has made a real breakthrough. This is one of the advances that will take us beyond the smartphone. If you've done any playing around with augmented reality, you realize that control and interaction is one of the key blocks to widespread adoption. Brain-computer interfaces are a key combinatorial innovation that will enable a completely different UI paradigm for interacting with devices and ambient computing.
Créer un dossier "ressources_googleplus" dans votre répertoire courant (%pwd pour le connaitre)
if False: # à remplacer par une autre API
import json
import apiclient.discovery
MAX_RESULTS = 200 # limite fixée à 100 résultats par requete => on va itérer sur une boucle pour en avoir 200
activity_feed = service.activities().list(
userId=USER_ID,
collection='public',
maxResults='100'
)
activity_results = []
while activity_feed != None and len(activity_results) < MAX_RESULTS:
activities = activity_feed.execute()
if 'items' in activities:
for activity in activities['items']:
if activity['object']['objectType'] == 'note' and activity['object']['content'] != '':
activity['title'] = cleanHtml(activity['title'])
activity['object']['content'] = cleanHtml(activity['object']['content'])
activity_results += [activity]
# list_next permet de passer à la requete suivante
activity_feed = service.activities().list_next(activity_feed, activities)
# on écrit le résultat dans un fichier json
import os
if not os.path.exists("ressources_googleplus"):
os.mkdir("ressources_googleplus")
f = open('./ressources_googleplus/' + USER_ID + '.json', 'w')
f.write(json.dumps(activity_results, indent=1))
f.close()
print(str(len(activity_results)), "activités écrites dans", f.name)
Le calcul tf-idf (term frequency–inverse document frequency) permet de calculer un score de proximité entre un terme de recherche et un document (c'est ce que font les moteurs de recherche). La partie tf calcule une fonction croissante de la fréquence du terme de recherche dans le document à l'étude, la partie idf calcule une fonction inversement proportionnelle à la fréquence du terme dans l'ensemble des documents (ou corpus). Le score total, obtenu en multipliant les deux composantes, permet ainsi de donner un score d'autant plus élevé que le terme est surréprésenté dans un document (par rapport à l'ensemble des documents). Il existe plusieurs fonctions, qui pénalisent plus ou moins les documents longs, ou qui sont plus ou moins smooth.
import json
with open("ressources_googleplus/107033731246200681024.json", "r", encoding="utf-8") as f:
activity_results = json.load(f)
corpus = {
'a' : "Mr. Green killed Colonel Mustard in the study with the candlestick. \
Mr. Green is not a very nice fellow.",
'b' : "Professor Plum has a green plant in his study.",
'c' : "Miss Scarlett watered Professor Plum's green plant while he was away \
from his office last week."
}
terms = {
'a' : [ i.lower() for i in corpus['a'].split() ],
'b' : [ i.lower() for i in corpus['b'].split() ],
'c' : [ i.lower() for i in corpus['c'].split() ]
}
from math import log
QUERY_TERMS = ['mr.', 'green']
def tf(term, doc, normalize=True):
doc = doc.lower().split()
if normalize:
return doc.count(term.lower()) / float(len(doc))
else:
return doc.count(term.lower()) / 1.0
def idf(term, corpus):
num_texts_with_term = len([True for text in corpus if term.lower() \
in text.lower().split()])
try:
return 1.0 + log(float(len(corpus)) / num_texts_with_term)
except ZeroDivisionError:
return 1.0
def tf_idf(term, doc, corpus):
return tf(term, doc) * idf(term, corpus)
for (k, v) in sorted(corpus.items()):
print(k, ':', v)
print('\n')
query_scores = {'a': 0, 'b': 0, 'c': 0}
for term in [t.lower() for t in QUERY_TERMS]:
for doc in sorted(corpus):
print('TF({}): {}'.format(doc, term), tf(term, corpus[doc]))
print('IDF: {}'.format(term, ), idf(term, corpus.values()))
print('\n')
for doc in sorted(corpus):
score = tf_idf(term, corpus[doc], corpus.values())
print('TF-IDF({}): {}'.format(doc, term), score)
query_scores[doc] += score
print('\n')
print("Score TF-IDF total pour le terme '{}'".format(' '.join(QUERY_TERMS), ))
for (doc, score) in sorted(query_scores.items()):
print(doc, score)
a : Mr. Green killed Colonel Mustard in the study with the candlestick. Mr. Green is not a very nice fellow. b : Professor Plum has a green plant in his study. c : Miss Scarlett watered Professor Plum's green plant while he was away from his office last week. TF(a): mr. 0.10526315789473684 TF(b): mr. 0.0 TF(c): mr. 0.0 IDF: mr. 2.09861228866811 TF-IDF(a): mr. 0.22090655670190631 TF-IDF(b): mr. 0.0 TF-IDF(c): mr. 0.0 TF(a): green 0.10526315789473684 TF(b): green 0.1111111111111111 TF(c): green 0.0625 IDF: green 1.0 TF-IDF(a): green 0.10526315789473684 TF-IDF(b): green 0.1111111111111111 TF-IDF(c): green 0.0625 Score TF-IDF total pour le terme 'mr. green' a 0.3261697145966431 b 0.1111111111111111 c 0.0625
Le score td-idf pour le terme "Mr. Green" est le plus élevé pour le document a.
Quel document est le plus proche du terme "green plant ? Calculer les scores TF-IDF pour le terme "green plan". Cela correspond-il à vos attentes ? Que se passe-t-il si vous inversez les termes "green" et "plant" ? Que se passe-t-il avec "green" seul ?
import nltk
# nltk donne accès a des methodes, mais aussi à des données, qui faut télécharge grâce à la commande .download()
nltk.download('stopwords')
[nltk_data] Downloading package stopwords to [nltk_data] C:\Users\xavie\AppData\Roaming\nltk_data... [nltk_data] Package stopwords is already up-to-date!
True
from pprint import pprint
len(activity_results)
273
if len(activity_results) > 0:
pprint(activity_results[0])
{'access': {'description': 'Public', 'items': [{'type': 'public'}], 'kind': 'plus#acl'}, 'actor': {'displayName': "Tim O'Reilly", 'id': '107033731246200681024', 'image': {'url': 'https://lh4.googleusercontent.com/-J8nmMwIhpiA/AAAAAAAAAAI/AAAAAAADdg4/68r2hyFUgzI/photo.jpg?sz=50'}, 'url': 'https://plus.google.com/107033731246200681024', 'verification': {'adHocVerified': 'PASSED'}}, 'etag': '"Sh4n9u6EtD24TM0RmWv7jTXojqc/Dlr_44FOo97cNjKbX7ZHrVWgen4"', 'id': 'z123e5zb4zbmcxf5004chl3pvxfbszirt5o', 'kind': 'plus#activity', 'object': {'actor': {'verification': {'adHocVerified': 'PASSED'}}, 'attachments': [{'content': 'This startup lets you control ' 'machines with your mind—no implants ' 'required.', 'displayName': "The Brain-Machine Interface Isn't " 'Sci-Fi Anymore | Backchannel', 'fullImage': {'type': 'image/jpeg', 'url': 'https://media.wired.com/photos/59b81acc9365592813946567/191:100/pass/2lead.jpg'}, 'image': {'height': 910, 'type': 'image/jpeg', 'url': 'https://lh3.googleusercontent.com/proxy/L96mdlI6FizC1ijKXUpxf_u6JjcJdl79sEzYVwyWmdeeJfiBvHMVvWFMn8kvL4sq8kG82ST8lEqjhW9-j9KdQuTYh9lscVMkKb-IgK0j_s-PKZ84xceA2OPHTcQJ4g=w506-h910', 'width': 506}, 'objectType': 'article', 'url': 'https://www.wired.com/story/brain-machine-interface-isnt-sci-fi-anymore/'}], 'content': 'It looks like #@CTRLLabsCo <b>has</b> made a real ' 'breakthrough. This is one of the advances that will ' "take us beyond the smartphone. If you've done any " 'playing around with augmented reality, you realize ' 'that control and interaction is one of the key blocks ' 'to widespread adoption. Brain-computer interfaces are ' 'a key combinatorial innovation that will enable a ' 'completely different UI paradigm for interacting with ' 'devices and ambient computing. ', 'objectType': 'note', 'plusoners': {'selfLink': 'https://www.googleapis.com/plus/v1/activities/z123e5zb4zbmcxf5004chl3pvxfbszirt5o/people/plusoners', 'totalItems': 101}, 'replies': {'selfLink': 'https://www.googleapis.com/plus/v1/activities/z123e5zb4zbmcxf5004chl3pvxfbszirt5o/comments', 'totalItems': 14}, 'resharers': {'selfLink': 'https://www.googleapis.com/plus/v1/activities/z123e5zb4zbmcxf5004chl3pvxfbszirt5o/people/resharers', 'totalItems': 16}, 'url': 'https://plus.google.com/+TimOReilly/posts/TpYYyGh7pr1'}, 'provider': {'title': 'Google+'}, 'published': '2017-09-13T15:59:31.577Z', 'title': 'It looks like #@CTRLLabsCo has made a real breakthrough. This is ' 'one of the advances that will take ...', 'updated': '2017-09-13T15:59:45.234Z', 'url': 'https://plus.google.com/+TimOReilly/posts/TpYYyGh7pr1', 'verb': 'post'}
if len(activity_results) > 0:
pprint(activity_results[0]['object']['content'])
('It looks like #@CTRLLabsCo <b>has</b> made a real breakthrough. This is one ' "of the advances that will take us beyond the smartphone. If you've done any " 'playing around with augmented reality, you realize that control and ' 'interaction is one of the key blocks to widespread adoption. Brain-computer ' 'interfaces are a key combinatorial innovation that will enable a completely ' 'different UI paradigm for interacting with devices and ambient computing. ')
all_content = " ".join([ a['object']['content'] for a in activity_results ])
print("Nombre de caractères : ",len(all_content))
print('\n')
#Tokenisation naïve sur les espaces entre les mots => on obtient une liste de mots
tokens = all_content.split()
#On transforme cette liste en objet nltk "Text" (objet chaine de caractère qui conserve la notion de tokens, et qui
#comprend un certain nombre de méthodes utiles pour explorer les données.
text = nltk.Text(tokens)
#Comme par exemple "concordance" : montre les occurences d'un mot dans son contexte
print("Exemples d'occurences du terme 'open' :")
text.concordance("open")
print('\n')
# Analyse de la fréquence des termes d'intérêt
fdist = text.vocab()
Nombre de caractères : 103145 Exemples d'occurences du terme 'open' : Displaying 13 of 13 matches: , and how at least one team is using open source to let others see inside the ear that computational biologist and open science advocate (UC Berkeley profes : magazine slogan say, "If you can't open it, you don't own it." Predictive po I'm proud to be a signatory to this open letter calling for this key policy i st, I've focused a lot on areas like open source software and the implications opic at greater length in my article Open Data and Algorithmic Regulation: htt cessful participatory projects, from open source software to wikis to social m ere isn't one (except that it's only open to US students - sorry. If anyone ha new contract that conformed with the open data mandate. If it were consistent If it were consistent with the Obama open data guidance, that RFP would requir ut of step with the administration’s open data policy.The founder of Hipcamp, is is a really important piece about open data and platforms. Work on sh-t tha r. An excellent demonstration of why Open Access lowers the barriers to knowle
print("Co-occurences fréquentes :")
colloc = text.collocation_list()
print(colloc)
Co-occurences fréquentes : ['Silicon Valley', "O'Reilly Media", 'New York', 'Common Core', '+Jennifer Pahlka', 'Next:Economy Summit', 'Brett Goldstein', 'Cabo Pulmo', 'Humble Bundle', 'Bay Mini', 'East Bay', 'White House', 'on-demand economy,', 'Maker Faire', 'Mini Maker', 'granite workers', 'Real businesses', 'Well worth', 'worth reading.', 'Barre Historical']
print('\n')
print("Nombre de mots :", len(tokens))
print('\n')
print("Nombre de mots uniques :",len(fdist.keys()))
print('\n')
print("Nombre de mots uniques v2 :",len(set(tokens)))
Nombre de mots : 17104 Nombre de mots uniques : 5561 Nombre de mots uniques v2 : 5561
print("Nombre d'occurences du terme 'open' :",fdist["open"])
print("Nombre d'occurences du terme 'source' :", fdist["source"])
print("Nombre d'occurences du terme 'web' :", fdist["web"])
print("Nombre d'occurences du terme 'API' :",fdist["API"])
print('\n')
#100 tokens les plus fréquents
top100_items = sorted(fdist.items(),key=lambda x: x[1],reverse=True)[:100]
#sans les fréquences
top100 = [t[0] for t in top100_items]
print("Top 100 :", top100)
print('\n')
#sans les termes trop frequents ("stopwords")
top100_without_stopwords = [w for w in top100 if w.lower() \
not in nltk.corpus.stopwords.words('english')]
print("Top 100 sans les mots fréquents :", top100_without_stopwords)
print('\n')
long_words_not_urls = [w for w in fdist.keys() if len(w) > 15 and not w.startswith("http")]
print("Longs mots sans les urls :", long_words_not_urls)
print('\n')
# Nombre d'urls
print("Nombre d'urls :", len([w for w in fdist.keys() if w.startswith("http")]))
print('\n')
# Enumerate the frequency distribution
for rank, word in enumerate(sorted(fdist.items(),key=lambda x: x[1],reverse=True)):
print(rank, word)
if rank > 75:
print("....")
break
Nombre d'occurences du terme 'open' : 11 Nombre d'occurences du terme 'source' : 5 Nombre d'occurences du terme 'web' : 1 Nombre d'occurences du terme 'API' : 2 Top 100 : ['the', 'to', 'of', 'and', 'a', 'in', 'is', 'for', 'that', 'I', 'on', 'with', 'about', 'it', 'are', 'this', 'you', 'at', 'from', 'as', 'have', 'be', 'my', 'how', 'an', 'by', 'we', 'what', 'but', 'was', 'This', 'not', 'The', 'they', 'their', 'his', 'has', 'than', 'so', 'more', 'new', 'can', 'do', 'like', 'or', 'who', 'out', 'one', 'good', 'our', 'make', '-', 'will', 'work', 'people', 'should', 'when', 'all', 'just', 'see', 'if', 'It', 'which', 'way', 'We', 'me', 'up', 'data', 'get', 'why', 'us', "I'm", 'them', 'piece', 'If', "O'Reilly", 'better', 'its', 'also', 'And', 'he', 'now', 'technology', 'government', 'many', 'some', 'been', 'your', 'great', 'love', "It's", 'think', 'no', 'into', 'business', 'had', 'other', 'only', 'much', 'So'] Top 100 sans les mots fréquents : ['new', 'like', 'one', 'good', 'make', '-', 'work', 'people', 'see', 'way', 'data', 'get', 'us', "I'm", 'piece', "O'Reilly", 'better', 'also', 'technology', 'government', 'many', 'great', 'love', 'think', 'business', 'much'] Longs mots sans les urls : ['impossible.Technology', 'financialization,', 'collusion...could', 'transformative.)', 'public-relations', 'self-assessment!', 'forward-thinking', '“attend-listen-embarrass”', 'algorithmically,', 'you!https://www.crowdpac.com/campaigns/100604', '#OReillySecurity', '#makesecurityeasy', 'doing.Individually,', 'producers....Like', 'October.http://www.inc.com/magazine/201606/sheila-marikar/lola-paul-english-uber.html', 'entrepreneurship', '(http://conferences.oreilly.com/nextcon/money-fintech-us)', 'life-threatening', "Thursday.”That's", 'post.)http://www.pressheretv.com/tim-oreilly/', 'humansofnewyork.com', '(https://medium.com/the-wtf-economy/the-wtf-economy-a3bd5f52ef00)', '(https://medium.com/the-wtf-economy/networks-and-the-nature-of-the-firm-28790b6afdcc).', '(http://conferences.oreilly.com/next-economy)', 'underrepresented', 'Republican-controlled', 'philanthropists,', 'Foundation.http://www.wsj.com/articles/sean-parker-philanthropy-for-hackers-1435345787', 'Michener)quoteinvestigator.com/2010/08/27/master/Thanks', '#FutureCrimesConvo,', 'www.futurecrimes.com', 'statistics:"Typically,', 'Super-interesting.', 'Brothers...."Their', 'innovation-resistant', 'software...."Get', 'Ireland.Incidentally,', 'level-headedness.', 'administration’s', 'jasonmking@fs.fed.us', 'available:http://ebmakerfaire2014.eventbrite.com', 'Makers/Exhibitors/Performers,', 'Regulations?Evgeny', "backwards.Here's", 'opens:"Regulation', 'results."Consider,', 'outcome"Real-time', 'achieved"Algorithms', 'quasi-governmental', 'everyone"Adjustments', 'achieved"Contrast', 'outcome?"(http://beyondtransparency.org/chapters/part-5/open-data-and-algorithmic-regulation/).', 'condition."Lovely.And', 'opens:"Something’s', 'codeforamerica.org.', 'interoperability', 'customizability.', '(drnkwines.com),', 'stream-processing', 'knowledge-sharing', 'Healthcare.govDraw', 'IndependenceThis', 'Palestinians:"This', 'national-authority'] Nombre d'urls : 28 0 ('the', 844) 1 ('to', 512) 2 ('of', 465) 3 ('and', 429) 4 ('a', 362) 5 ('in', 264) 6 ('is', 238) 7 ('for', 208) 8 ('that', 191) 9 ('I', 170) 10 ('on', 156) 11 ('with', 122) 12 ('about', 118) 13 ('it', 110) 14 ('are', 105) 15 ('this', 101) 16 ('you', 100) 17 ('at', 100) 18 ('from', 85) 19 ('as', 84) 20 ('have', 81) 21 ('be', 76) 22 ('my', 73) 23 ('how', 72) 24 ('an', 72) 25 ('by', 69) 26 ('we', 61) 27 ('what', 60) 28 ('but', 57) 29 ('was', 56) 30 ('This', 54) 31 ('not', 52) 32 ('The', 50) 33 ('they', 50) 34 ('their', 49) 35 ('his', 46) 36 ('has', 46) 37 ('than', 45) 38 ('so', 44) 39 ('more', 43) 40 ('new', 43) 41 ('can', 42) 42 ('do', 42) 43 ('like', 41) 44 ('or', 39) 45 ('who', 38) 46 ('out', 38) 47 ('one', 37) 48 ('good', 35) 49 ('our', 35) 50 ('make', 35) 51 ('-', 33) 52 ('will', 32) 53 ('work', 29) 54 ('people', 29) 55 ('should', 29) 56 ('when', 29) 57 ('all', 29) 58 ('just', 29) 59 ('see', 28) 60 ('if', 28) 61 ('It', 27) 62 ('which', 27) 63 ('way', 27) 64 ('We', 27) 65 ('me', 27) 66 ('up', 27) 67 ('data', 26) 68 ('get', 26) 69 ('why', 26) 70 ('us', 25) 71 ("I'm", 25) 72 ('them', 25) 73 ('piece', 25) 74 ('If', 24) 75 ("O'Reilly", 24) 76 ('better', 24) ....
fdist = text.vocab()
%matplotlib inline
import matplotlib.pyplot as plt
fig, ax = plt.subplots(1, 1, figsize=(16,4))
fdist.plot(100,cumulative=True);
Les notes de Tim O'Reilly parlent-elles des élections américaines ? Avec quelle fréquence ? Essayer "Hillary","Trump", "vote", d'autres idées ?
Quelle distribution des termes sans les stopwords ? Vérifie-t-on la loi Zipf selon laquelle la fréquence des mots est inversement proportionnelle à son rang (le 10 ème mot est 10 fois moins présent que le premier) ? Et si l'on laisse les "stopwords" ?
Proposer un indice du calcul de la diversité du vocabulaire de Tim O'Reilly.
Le code ci-après permet d'indexer des textes en fonction d'un score de pertinence pour une requête donnée. En d'autres termes, vous avez un petit moteur de recherche :)
import json
import nltk
path = 'ressources_googleplus/107033731246200681024.json'
text_data = json.loads(open(path).read())
QUERY_TERMS = ['open','data']
activities = [activity['object']['content'].lower().split() \
for activity in text_data \
if activity['object']['content'] != ""]
# Le package TextCollection contient un module tf-idf
tc = nltk.TextCollection(activities)
relevant_activities = []
for idx in range(len(activities)):
score = 0
for term in [t.lower() for t in QUERY_TERMS]:
score += tc.tf_idf(term, activities[idx])
if score > 0:
relevant_activities.append({'score': score, 'title': text_data[idx]['title'],
'url': text_data[idx]['url']})
# Tri par score et présentation des résultats
relevant_activities = sorted(relevant_activities,
key=lambda p: p['score'], reverse=True)
c=0
for activity in relevant_activities:
if c < 6:
print(activity['title'])
print('\tLink: {}'.format(activity['url']))
print('\tScore: {}'.format(activity['score']))
c+=1
This is a really important piece about open data and platforms. Link: https://plus.google.com/+TimOReilly/posts/fo9uxWTctHb Score: 0.5498599632119789 I love new sources of trend data about technology adoption. We've used variations of this for years ... Link: https://plus.google.com/+TimOReilly/posts/FetXVRJeJFv Score: 0.17368671875174563 If you love Hamilton, as I do, and you're interested in data visualization, you'll find this fascinating... Link: https://plus.google.com/+TimOReilly/posts/NNsiSo8K7B7 Score: 0.16687547487912816 Data can play a great role in advancing sustainability. I'm quoted in this short video from Planet Labs... Link: https://plus.google.com/+TimOReilly/posts/45KX41Q2LN4 Score: 0.15760461516362104 Mark Cuban's tweet about data science in the NBA, featuring the image of his screen and an O'Reilly ... Link: https://plus.google.com/+TimOReilly/posts/2hCQhfTaX5g Score: 0.14184415364725894 An excellent demonstration of why Open Access lowers the barriers to knowledge-sharing in science. This... Link: https://plus.google.com/+TimOReilly/posts/iQ4RdspWxbY Score: 0.13381568843277453
import json
import nltk
path = 'ressources_googleplus/107033731246200681024.json'
data = json.loads(open(path).read())
# Sélection des textes qui ont plus de 1000 mots
data = [ post for post in json.loads(open(path).read()) \
if len(post['object']['content']) > 1000 ]
all_posts = [post['object']['content'].lower().split()
for post in data ]
tc = nltk.TextCollection(all_posts)
# Calcul d'une matrice terme de recherche x document
# Renvoie un score tf-idf pour le terme dans le document
td_matrix = {}
for idx in range(len(all_posts)):
post = all_posts[idx]
fdist = nltk.FreqDist(post)
doc_title = data[idx]['title']
url = data[idx]['url']
td_matrix[(doc_title, url)] = {}
for term in fdist.keys():
td_matrix[(doc_title, url)][term] = tc.tf_idf(term, post)
distances = {}
for (title1, url1) in td_matrix.keys():
distances[(title1, url1)] = {}
(min_dist, most_similar) = (1.0, ('', ''))
for (title2, url2) in td_matrix.keys():
#copie des valeurs (un dictionnaire étant mutable)
terms1 = td_matrix[(title1, url1)].copy()
terms2 = td_matrix[(title2, url2)].copy()
#on complete les gaps pour avoir des vecteurs de même longueur
for term1 in terms1:
if term1 not in terms2:
terms2[term1] = 0
for term2 in terms2:
if term2 not in terms1:
terms1[term2] = 0
#on créé des vecteurs de score pour l'ensemble des terms de chaque document
v1 = [score for (term, score) in sorted(terms1.items())]
v2 = [score for (term, score) in sorted(terms2.items())]
#calcul des similarité entre documents : distance cosine entre les deux vecteurs de scores tf-idf
distances[(title1, url1)][(title2, url2)] = \
nltk.cluster.util.cosine_distance(v1, v2)
import pandas as p
df_dist=p.DataFrame(distances)
df_dist.iloc[:5,:5]
From an article about Walmart, their move to pay more, and the lessons for the broader economy: http... | Nassau, The Bahamas Airport Travel Advice\n\nIf anyone happens to travel to Nassau, the Bahamas, I thought... | Amazing story about digital transformation http://www.codeforamerica.org/blog/2015/11/30/a-new-approach... | "Surely Democrats and Republicans could agree to cut billions from a failed program like this!" you ... | How fragile life is, even for the best of us. We heard this morning that our friend Jake Brewer was ... | ||
---|---|---|---|---|---|---|
https://plus.google.com/+TimOReilly/posts/bqErtyYp6co | https://plus.google.com/+TimOReilly/posts/dpQDew7sPbu | https://plus.google.com/+TimOReilly/posts/BRmKh2ycaPe | https://plus.google.com/+TimOReilly/posts/1Lcxb3b8VPH | https://plus.google.com/+TimOReilly/posts/jV8jeKeWWyf | ||
"Surely Democrats and Republicans could agree to cut billions from a failed program like this!" you ... | https://plus.google.com/+TimOReilly/posts/1Lcxb3b8VPH | 9.415217e-01 | 0.984552 | 0.965728 | 0.000000 | 0.969433 |
Absolutely fascinating exploration of the microbiome of a city.\n\nFor those who don't know what the microbiome... | https://plus.google.com/+TimOReilly/posts/7EaHeYc1BiB | 9.699011e-01 | 0.976170 | 0.973205 | 0.983031 | 0.974682 |
Amazing story about digital transformation http://www.codeforamerica.org/blog/2015/11/30/a-new-approach... | https://plus.google.com/+TimOReilly/posts/BRmKh2ycaPe | 9.862850e-01 | 0.980943 | 0.000000 | 0.965728 | 0.987102 |
Can We Use Data to Make Better Regulations?\n\nEvgeny Morozov either misunderstands or misrepresents the... | https://plus.google.com/+TimOReilly/posts/gboAUahQwuZ | 9.551053e-01 | 0.975855 | 0.967001 | 0.897357 | 0.964860 |
From an article about Walmart, their move to pay more, and the lessons for the broader economy: http... | https://plus.google.com/+TimOReilly/posts/bqErtyYp6co | -2.220446e-16 | 0.963338 | 0.986285 | 0.941522 | 0.982210 |
Les approches bag-of-words, bien que simplistes, permettent de créer d'indexer et de comparer des documents. La prise en compte des suites de 2, 3 ou plus mots serait un moyen d'affiner de tels modèles. Cela permet aussi de mieux comprendre le sens des homonymes, et des phrases (d'une manière générale, la sémantique).
nltk offre des methodes pour tenir compte du contexte : pour ce faire, nous calculons les n-grams, c'est-à-dire l'ensemble des co-occurrences successives de mots deux-à-deux (bigrams), trois-à-trois (tri-grams), etc.
En général, on se contente de bi-grams, au mieux de tri-grams :
import nltk
sentence = "Mr. Green killed Colonel Mustard in the study with the " + \
"candlestick. Mr. Green is not a very nice fellow."
print(list(nltk.ngrams(sentence.split(), 2)))
txt = nltk.Text(sentence.split())
txt.collocation_list()
[('Mr.', 'Green'), ('Green', 'killed'), ('killed', 'Colonel'), ('Colonel', 'Mustard'), ('Mustard', 'in'), ('in', 'the'), ('the', 'study'), ('study', 'with'), ('with', 'the'), ('the', 'candlestick.'), ('candlestick.', 'Mr.'), ('Mr.', 'Green'), ('Green', 'is'), ('is', 'not'), ('not', 'a'), ('a', 'very'), ('very', 'nice'), ('nice', 'fellow.')]
['Mr. Green']
import json
import nltk
path = 'ressources_googleplus/107033731246200681024.json'
data = json.loads(open(path).read())
# Nombre de co-occurrences à trouver
N = 25
all_tokens = [token for activity in data for token in \
activity['object']['content'].lower().split()]
finder = nltk.BigramCollocationFinder.from_words(all_tokens)
finder.apply_freq_filter(2)
#filtre des mots trop fréquents
finder.apply_word_filter(lambda w: w in nltk.corpus.stopwords.words('english'))
bigram_measures = nltk.collocations.BigramAssocMeasures()
collocations = finder.nbest(bigram_measures.jaccard, N)
for collocation in collocations:
c = ' '.join(collocation)
print(c)
bottom, “copyright brett goldstein cabo pulmo nbc press:here nick hanauer press:here tv wood fired yuval noah silicon valley +jennifer pahlka barre historical computational biologist mikey dickerson saul griffith bay mini child welfare credit card east bay on-demand economy, white house drm-free ebooks humble bundle inca trail italian granite private sector