def __init__(self,
selector,
mapper,
score,
mapperIsFactory=True,
timer=False,
useTopic=False):
'''
:param selector: returns list of documents for the topic
:param mapper: a mapper function on (selected) topic documents,
returning vectors or scalar,
or alternatively a 'factory' - callable that accepts 'dictionary',
'text2tokens' and 'corpus' parameters (attributes of
a topic model to which processed topic belongs) and
builds such a mapper.
This is for creating a customized mapper for each model.
:param mapperIsFactory: if True treat mapper as a factory as described above
:param score: score function on the matrix/vector of transformed documents
:param useTopic: if True, topic is also sent as parameter to mapper
'''
self.selector, self.score = selector, deduceId(score)
self.mapper = resolveId(mapper)
IdComposer.__init__(self)
self.__factory = mapperIsFactory
self.__mapper = resolve(mapper)
self.__score = score
self.__timer = timer
self.__useTopic = useTopic
def __init__(self,
paramSet,
scorerBuilder,
ltopics,
posClass,
folder,
verbose=True,
cache=False):
'''
:param paramSet: list-like of paramsets passed to scorerBuilder
each 'paramset' is a dict of parameters
:param scorerBuilder: callable building scoring functions from a paramset
:param ltopics: list of (topic, labels) - topic is (modelId, topicId),
labels is a string or dict of 'label':{0 or 1}
:param posClass: string or a list of strings defining which label is positive class
:param verbose: if true, print results during experiment runtime
:param cache: if True, force scorerBuilder to cache function results
'''
self.setOfParams = paramSet
self.scorerBuilder = scorerBuilder.__name__
self.__builder = scorerBuilder
self.ltopics = ltopics
self.posClass = posClass
IdComposer.__init__(self, class_='TSE')
self.verbose = verbose
self.__log = createLogger(fullClassName(self), INFO)
self.baseFolder = folder
self.folder = path.join(folder, self.id)
if not path.exists(self.folder): os.mkdir(self.folder)
self.cache = cache
if cache:
self.cacheFolder = path.join(self.baseFolder, 'function_cache')
if not path.exists(self.cacheFolder): os.mkdir(self.cacheFolder)
self.__rSetUp = False
def __init__(self,
selector,
mapper,
metric,
weightFilter=None,
weighted=True,
algorithm='clustering',
center='mean',
mapperIsFactory=True):
'''
:param selector: selector of topic items (nodes of the graph)
:param mapper: mapper of topic items to vectors
:param metric: weight metric on pairs of vectors, for forming weight matrix
:param weightFilter: if not None - pair/list of two elements
weights in weight matrix outside this interval will be set to 0
:param weighted: if True, graph algorithms will work with weighted graphs
:param algorithm: graph-based measure the coherence is based on
'clustering' - clustering coefficient
'closeness' - closeness centrality
'communicability' - communicability centrality
:param center: 'mean' or 'median' - to use when averaging a measure over nodes
'''
self.selector, self.mapper = selector, resolveId(mapper)
self.metric, self.weightFilter = metric, weightFilter
self.center = center
self.algorithm = algorithm
self.weighted = weighted
IdComposer.__init__(self)
self.__tes = TopicDocumentsScore(selector, mapper, None,
mapperIsFactory)
def __init__(self, fname):
'''
:param fname: file with vectors stored in txt format
:return:
'''
self.file = self.__idFromFname(fname)
IdComposer.__init__(self)
self.__fname = fname
def __init__(self, fname):
'''
:param fname: file with vectors stored in original google format,
with optional .bin extension
:return:
'''
self.file = self.__idFromFname(fname)
IdComposer.__init__(self)
self.__fname = fname
def __init__(self, corpus, txt2tok, dict, context=None, initSize=100):
self.__cr = ContextResolver(context)
corpus, txt2tok, dict = self.__cr.resolve(corpus, txt2tok, dict)
self.corpus, self.txt2tok, self.dict = \
resolveIds(corpus, txt2tok, dict)
IdComposer.__init__(self, attributes=['corpus', 'txt2tok', 'dict'])
self.__corpus = np.empty(initSize, dtype=np.object)
self.__nd = 0 # number of documents
self.__index = None
def __init__(self, threshold):
'''
:param threshold: integer (for top K selction),
a number between 0 and 1 (as topic-weight threshold)
or 'above-random' that sats topic-weight threshold to 1/num_topics
'''
self.threshold = threshold2str(threshold)
IdComposer.__init__(self)
self.__threshold = threshold
def __init__(self, corpus=None, text2tokens=None, label=None):
self._tok2ind = dict()
self._ind2tok = dict()
self._maxind = None
corpus, text2tokens = resolveIds(corpus, text2tokens)
if corpus is not None: self.corpus = corpus
if text2tokens is not None: self.text2tokens = text2tokens
if label is not None: self.label = label
IdComposer.__init__(self)
def __init__(self, text2tokens, filter, transformer, lowercase=False):
'''
:param filter: callable, accepts a string token, returns True (filter out) or False (keep token)
:param transformer: callable, accepts a string token, returns a string
'''
self.text2tokens = text2tokens
self.filter, self.transformer = filter, transformer
if lowercase: self.lowercase = True
IdComposer.__init__(self)
self.lowercase = lowercase
def __init__(self, distance, center='mean', exp=1.0):
self.distance = distance
self.center = center
if center == 'mean':
self.exp = exp
IdComposer.__init__(self)
elif center == 'median':
# raising to positive power does not influence median
IdComposer.__init__(self)
else:
raise Exception('Invalid measure of center %s' % center)
def __init__(self, corpus, models=None):
'''
:param corpus:
:param models: list of model Ids, or None
'''
self.corpus = resolveIds(corpus)
self.models = '_'.join(str(m) for m in models) if models else None
IdComposer.__init__(self)
self.__modelTopicOrder = {}
self.__models = models
self.__ctiCache = {}
def __init__(self, dimReduce=None, covariance='diag',
score = 'll', seed=12345):
'''
:param dimReduce: if True,
:param covariance: full, diag, spherical
:param score: ll (log likelihood), aic, bic
:param seed: random seed for initialization
'''
self.seed = seed
self.covariance = covariance
self.dimReduce = dimReduce
self.score = score
IdComposer.__init__(self)
def __init__(self, docLowerLimit, docUpperLimit, words2keep):
'''
:param docLowerLimit: words are removed if they don't occur
in at least this many documents, if None filter is not applied
:param docUpperLimit: words are removed if they occur in
more than this many documents (integer or fraction),
if None filter is not applied
:param words2keep: after applying document filters, keep only this many words
'''
self.docLowerLimit = docLowerLimit
self.docUpperLimit = docUpperLimit
self.words2keep = words2keep
IdComposer.__init__(self)
def __init__(self, model, id=None, cacheTopics=True):
'''
:param model: gensim LdaModel or compatible
:param label: additional data used for model identification
:param cacheTopics: if True topic matrix of gensim model is copied
and not calculated on every access
'''
IdComposer.__init__(self,
['corpus', 'dictionary', 'text2tokens', 'options'])
self.id = id
if model is not None:
if not isinstance(model, LdaModel_mod):
raise TypeError('model must be of type ldamodel')
self.__init_gensim_data(model)
self.cacheTopics = cacheTopics
def __init__(self, distance, center='mean', exp=1.0):
'''
:param distance: distance function
:param exp: each distance is raised to power exp > 0
:param center: 'mean' or 'median'
'''
self.distance = distance
self.center = center
if center == 'mean':
self.exp = exp
IdComposer.__init__(self)
elif center == 'median':
# raising to positive power does not influence median
IdComposer.__init__(self)
else:
raise Exception('Invalid measure of center %s' % center)
def __init__(self, vectorizer, corpus=None, verbose=False):
'''
:param vectorizer: callable accepting text objects or text ids and
returning vectors of the same type (ndarray or scipy sparse), shape and stored scalar type.
:param corpus: if None and if vectorizer has a 'corpus' property it is used as corpus
'''
self.vectorizer = vectorizer
if corpus is None:
if hasattr(vectorizer, 'corpus'):
self.corpus = resolveIds(vectorizer.corpus)
else:
raise Exception('corpus must be specified')
else:
self.corpus = resolveIds(corpus)
IdComposer.__init__(self, attributes=['vectorizer'])
self.verbose = verbose
def __init__(self, measure, index, topWords, windowSize=0, wordTransform=None, standard=True):
'''
:param measure: string describing coherence measure
:param index: pytopia id of Palmetto Lucene index
:param topWords: number of top topic words used to calculate coherence
'''
if measure not in ['umass', 'uci', 'npmi', 'c_a', 'c_p', 'c_v']:
raise Exception('unknown coherence measure: %s' % measure)
self.measure = measure
self.index = index
self.topWords = topWords
self.windowSize = windowSize
self.wordTransform = wordTransform
self.standard = standard
IdComposer.__init__(self)
self.coherence = None
def __init__(self,
dimReduce=None,
covariance='diag',
center='mean',
seed=12345):
'''
:param dimReduce: None or number of components to reduce to with PCA
:param covariance: full, diag, spherical
:param scoreMeasure: ll (log likelihood), aic, bic
:param seed: random seed for initialization
'''
self.covariance = covariance
self.dimReduce = dimReduce
self.center = center
IdComposer.__init__(self)
self.seed = seed
def __init__(self,
type,
vectors=None,
threshold=None,
topWords=10,
corpus='us_politics',
text2tokens='RsssuckerTxt2Tokens',
dict='us_politics_dict',
cache=None,
**params):
self.type, self.vectors, self.threshold = type, vectors, threshold
self.__p = params
attrs = ['type', 'vectors', 'threshold']
for k, v in params.iteritems():
setattr(self, k, v)
attrs.append(k)
self.topWords = topWords
IdComposer.__init__(self, attributes=attrs, class_='Coherence')
self.corpus, self.text2tokens, self.dictionary = corpus, text2tokens, dict
self.cache = cache
def __init__(self, corpus, text2tokens, lowercase):
self.corpus, self.text2tokens = resolveIds(corpus, text2tokens)
self.lowercase = lowercase
IdComposer.__init__(self)
def __init__(self, corpus):
'''
:param corpus: pytopia corpus or id
'''
self.corpus = resolve(corpus).id
IdComposer.__init__(self)
def __init__(self, text2tokens, word2vector, tokenMod=None, avg=None):
# todo remove tokenMod, clients should use function composition insted
# todo solve passing params by id
self.text2tokens, self.word2vector = text2tokens, word2vector
self.tokenMod = tokenMod; self.avg = avg
IdComposer.__init__(self)
def __init__(self, corpus, text2tokens, dictionary):
self.corpus, self.dictionary, self.text2tokens = \
resolveIds(corpus, dictionary, text2tokens)
IdComposer.__init__(self, ['corpus', 'dictionary', 'text2tokens'])
def __init__(self, model, topicId, vector=None):
self.model = resolveId(model)
self.topicId = topicId
IdComposer.__init__(self, ['model', 'topicId'])
self._vector = vector
def __init__(self, corpus, model, dictionary=None, txt2tokens=None):
self.corpus, self.model = resolveIds(corpus, model)
self.dictionary, self.text2tokens = resolveIds(dictionary, txt2tokens)
IdComposer.__init__(
self, attributes=['corpus', 'model', 'dictionary', 'text2tokens'])
def __init__(self, text2tokens, dictionary):
self.text2tokens, self.dictionary = resolveIds(text2tokens, dictionary)
IdComposer.__init__(self)
def __init__(self, measure):
'''
:param measure: measure of distance between two distributions
'''
self.measure = measure
IdComposer.__init__(self)