class SuggestCategoriesView(formbase.PageForm):
"""Suggest categories to the user and let him set them.
"""
implements(IPlonePageForm)
label = _(u"Suggested categories")
description = _(u"Choose among the proposed subjects. Clicking on apply "\
"will add the chosen categories to the existing ones.")
def getSuggestedSubjects(self):
"""
"""
classifier = getUtility(IContentClassifier)
uid = IClassifiable(self.context).UID
return classifier.probabilityClassify(uid)
@property
def form_fields(self):
"""
"""
ff = form.Fields(ISuggestCategories)
suggestions = self.getSuggestedSubjects()
if not suggestions:
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
IStatusMessage(self.request).addStatusMessage(
_(u"Classifier has not been trained or has "\
"not sufficient information."), type="error")
self.request.response.redirect(url)
return []
subject_prob_list = [
(suggestions.prob(subject), subject)
for subject in suggestions.samples()]
subject_prob_list = sorted(subject_prob_list, reverse=True)
vocab_terms = []
for (probability, subject) in subject_prob_list:
label = "%s %2.1f%%"%(subject, probability*100)
vocab_terms.append(SimpleTerm(value=subject,
token=b64encode(subject),
title=label))
choice = schema.Choice(vocabulary=SimpleVocabulary(vocab_terms))
ff['suggestions'].field.value_type = choice
return ff
@form.action(_(u"Apply"))
def action_submit(self, action, data):
"""
"""
obj = IClassifiable(self.context)
subjects = obj.categories
for subject in data['suggestions']:
if subject not in subjects:
subjects.append(subject)
obj.categories = subjects
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
IStatusMessage(self.request).addStatusMessage(
_(u"Categories saved."), type="info")
self.request.response.redirect(url)
return ''
class ISuggestCategories(Interface):
"""
"""
suggestions = schema.List(
title = _(u"Suggestions"),
description = _(u""),
default = [])
class IClassifierSettingsSchema(Interface):
"""Classifier settings
"""
train_after_update = schema.Bool(
title=_(u"Train after update"),
description=_(u"Enabling this will trigger training the classifier " \
"every time tagged content is added, modified or deleted. " \
"Disabling it means you will have to periodically manually " \
"retrain the classifier."))
def form_fields(self):
"""
"""
ff = form.Fields(ISuggestCategories)
suggestions = self.getSuggestedSubjects()
if not suggestions:
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
IStatusMessage(self.request).addStatusMessage(
_(u"Classifier has not been trained or has "\
"not sufficient information."), type="error")
self.request.response.redirect(url)
return []
subject_prob_list = [
(suggestions.prob(subject), subject)
for subject in suggestions.samples()]
subject_prob_list = sorted(subject_prob_list, reverse=True)
vocab_terms = []
for (probability, subject) in subject_prob_list:
label = "%s %2.1f%%"%(subject, probability*100)
vocab_terms.append(SimpleTerm(value=subject,
token=b64encode(subject),
title=label))
choice = schema.Choice(vocabulary=SimpleVocabulary(vocab_terms))
ff['suggestions'].field.value_type = choice
return ff
def form_fields(self):
"""
"""
ff = form.Fields(ISuggestCategories)
suggestions = self.getSuggestedSubjects()
if not suggestions:
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
IStatusMessage(self.request).addStatusMessage(
_(u"Classifier has not been trained or has "\
"not sufficient information."), type="error")
self.request.response.redirect(url)
return []
subject_prob_list = [
(suggestions.prob(subject), subject)
for subject in suggestions.samples()]
subject_prob_list = sorted(subject_prob_list, reverse=True)
vocab_terms = []
for (probability, subject) in subject_prob_list:
label = "%s %2.1f%%"%(subject, probability*100)
vocab_terms.append(SimpleTerm(value=subject,
token=b64encode(subject),
title=label))
choice = schema.Choice(vocabulary=SimpleVocabulary(vocab_terms))
ff['suggestions'].field.value_type = choice
return ff
class ClassifierSettings(ControlPanelForm):
form_fields = form.FormFields(IClassifierSettingsSchema)
label = _("Classification settings")
description = _("Settings for collective.classification.")
form_name = _("Classification settings")
@form.action(_(u"Save"))
def save_action(self, action, data):
form.applyChanges(self.context, self.form_fields, data, self.adapters)
self.status = _(u"Changes saved.")
@form.action(_(u"Retrain classifier"))
def retrain_classifier_action(self, action, data):
classifier = getUtility(IContentClassifier)
classifier.train()
self.status = _(u"Classifier trained.")
@form.action(_(u"Statistics"), validator=null_validator)
def stats_action(self, action, data):
"""Displays the stats view.
"""
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
self.request.response.redirect(url + '/@@classification-stats')
return ''
@form.action(_(u"Cancel"), validator=null_validator)
def cancel_action(self, action, data):
self.status = _(u"Changes cancelled.")
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
self.request.response.redirect(url + '/plone_control_panel')
return ''
def retrain_classifier_action(self, action, data):
storage = getUtility(INounPhraseStorage)
classifier = getUtility(IContentClassifier)
classifier.clear()
catalog = getToolByName(self.context, "portal_catalog")
types_to_search = storage.friendlyTypes or self._friendlyContentTypes()
trainContent = catalog.searchResults(portal_type=types_to_search)
for item in trainContent:
if item.Subject:
classifier.addTrainingDocument(item["UID"], item["Subject"])
classifier.train()
self.status = _(u"Classifier trained.")
def action_submit(self, action, data):
"""
"""
obj = IClassifiable(self.context)
subjects = obj.categories
for subject in data["suggestions"]:
if subject not in subjects:
subjects.append(subject)
obj.categories = subjects
url = getMultiAdapter((self.context, self.request), name="absolute_url")()
IStatusMessage(self.request).addStatusMessage(_(u"Categories saved."), type="info")
self.request.response.redirect(url)
return ""
def action_submit(self, action, data):
"""
"""
obj = IClassifiable(self.context)
subjects = obj.categories
for subject in data['suggestions']:
if subject not in subjects:
subjects.append(subject)
obj.categories = subjects
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
IStatusMessage(self.request).addStatusMessage(
_(u"Categories saved."), type="info")
self.request.response.redirect(url)
return ''
def retrain_termextractor_action(self, action, data):
storage = getUtility(INounPhraseStorage)
storage.clear()
catalog = getToolByName(self.context, "portal_catalog")
types_to_search = storage.friendlyTypes or self._friendlyContentTypes()
trainContent = catalog.searchResults(portal_type=types_to_search)
for item in trainContent:
# NOTE: Why can't I obtain item.SearchableText?
# Is it too big to be returned in catalog brains?
obj = item.getObject()
uid = obj.UID()
text = convertHtmlToWebIntelligentPlainText(obj.SearchableText())
storage.addDocument(uid, text)
self.status = _(
u"Term extractor trained and NP storage updated." " You will need to re-train the classifier as well."
)
def save_action(self, action, data):
form.applyChanges(self.context, self.form_fields, data, self.adapters)
extractor = getUtility(ITermExtractor)
# Check if user has changed the tagger...
ttype = data["tagger_type"]
tcategories = data["brown_categories"]
if extractor.tagger_metadata["type"] != ttype or extractor.tagger_metadata["categories"] != tcategories:
if ttype == "N-Gram":
tagged_sents = brown.tagged_sents(categories=tcategories)
tagger = getUtility(IPOSTagger, name="collective.classification.taggers.NgramTagger")
tagger.train(tagged_sents)
extractor.setTagger(tagger, {"type": "N-Gram", "categories": tcategories})
else:
tagger = getUtility(IPOSTagger, name="collective.classification.taggers.PennTreebankTagger")
extractor.setTagger(tagger, {"type": "Pen TreeBank", "categories": []})
self.status = _(u"Changes saved. You will need to reparse the " "content and then retrain the classifier.")
class ClassificationStatsView(formbase.PageForm):
form_fields = form.Fields(IStats)
template = ViewPageTemplateFile('classificationstats.pt')
def __init__(self, *args, **kwargs):
"""
"""
super(ClassificationStatsView, self).__init__(*args,**kwargs)
catalog = getToolByName(self.context, 'portal_catalog')
self.classifier = getUtility(IContentClassifier)
self.informativeFeatures = self.classifier.informativeFeatures()
self.parsedDocs = len(catalog._catalog.getIndex('noun_terms')._unindex)
@form.action(_(u"Apply"))
def action_apply(self, action, data):
"""
"""
self.informativeFeatures = \
self.classifier.informativeFeatures(data['no_features'])
class ClusterizeView(formbase.PageForm):
form_fields = form.Fields(IClusterize)
template = ViewPageTemplateFile('clusterize.pt')
@form.action(_(u"Clusterize"))
def action_clusterize(self, action, data):
"""
"""
catalog = getToolByName(self.context, 'portal_catalog')
clusterer = KMeans()
clusters = clusterer.clusterize(data['no_clusters'],
data['no_noun_ranks'],
repeats=data['repeats'])
result = []
for cluster in clusters.values():
clusterlist = []
for uid in cluster:
item = catalog.unrestrictedSearchResults(UID=uid)[0]
clusterlist.append(
(item.getURL(), item.Title, item.Description))
result.append(clusterlist)
self.clusters = result
class IClusterize(Interface):
no_clusters = schema.Int(
title=_(u"Number of clusters"),
description=_(u""),
required=True,
)
no_noun_ranks = schema.Int(
title=_(u"Important nouns to keep"),
description=_(u"Indicates how many nouns to keep when building the" \
"list of most frequent nouns in the text."),
default=20,
required=True)
repeats = schema.Int(
title=_(u"Number of runs"),
description=_(u""),
default=10,
required=True,
)
def save_action(self, action, data):
form.applyChanges(self.context, self.form_fields, data, self.adapters)
self.status = _(u"Changes saved.")
class IStats(Interface):
no_features = schema.Int(
title=_(u"Number of informative features to show"),
required=True,
default=10)
def retrain_classifier_action(self, action, data):
classifier = getUtility(IContentClassifier)
classifier.train()
self.status = _(u"Classifier trained.")
def cancel_action(self, action, data):
self.status = _(u"Changes cancelled.")
url = getMultiAdapter((self.context, self.request),
name='absolute_url')()
self.request.response.redirect(url + '/plone_control_panel')
return ''
