jonas/nlp-lab
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

documentation and python file for simple_twitter_learning

Browse Source
This commit is contained in:
Jonas Weinz 2018-06-20 13:38:18 +02:00
parent a623322ebb
commit 0bedb6060d
2 changed files with 749 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

148
Project/simple_approach/simple_twitter_learning.ipynb
View File

@ -9,6 +9,8 @@
"name": "stderr",
"output_type": "stream",
"text": [
"/home/jonas/.local/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
" from ._conv import register_converters as _register_converters\n",
"Using TensorFlow backend.\n"
]
},
@ -180,7 +182,14 @@
"cell_type": "markdown",
"metadata": {},
"source": [
"* the sample data manager loads and preprocesses data"
"### sample data manager\n",
"the sample data manager loads and preprocesses data\n",
"most common way to use:\n",
"\n",
"\n",
"* `sdm = sample_data_manager.generate_and_read(path:str, only_emoticons=True, apply_stemming=True, n_top_emojis=-1, file_range=None)`\n",
"\n",
" * Generates a sample_data_manager object and preprocess data in one step\n"
]
},
{
@ -192,6 +201,16 @@
"class sample_data_manager(object):\n",
" @staticmethod\n",
" def generate_and_read(path:str, only_emoticons=True, apply_stemming=True, n_top_emojis=-1, file_range=None):\n",
" \"\"\"\n",
" generate, read and process train data in one step.\n",
" \n",
" @param path: folder containing json files to process\n",
" @param only_emoticons: if True, only messages containing emoticons (provided by Tools.Emoji_Distance) are used\n",
" @param apply_stemming: apply stemming and lemmatization on dataset\n",
" @param n_top_emojis: only use messages containing one of <`n_top_emojis`>-top emojis. set to `-1` to prevent top emoji filtering\n",
" @param file_range: range of file's indices to read (eg `range(3)` to read the first three files). If `None`: all files are read\n",
" @return: sample_data_manager object\n",
" \"\"\"\n",
" sdm = sample_data_manager(path)\n",
" sdm.read_files(file_index_range=range(sdm.n_files) if file_range is None else file_range, only_emoticons=only_emoticons)\n",
" if apply_stemming:\n",
@ -206,6 +225,11 @@
" \n",
" \n",
" def __init__(self, data_root_folder:str):\n",
" \"\"\"\n",
" constructor for manual initialization\n",
" \n",
" @param data_root_folder: folder containing json files to process\n",
" \"\"\"\n",
" self.data_root_folder = data_root_folder\n",
" self.json_files = sorted(glob.glob(self.data_root_folder + \"/*.json\"))\n",
" self.n_files = len(self.json_files)\n",
@ -222,6 +246,12 @@
" self.top_emojis = None\n",
" \n",
" def read_files(self, file_index_range:list, only_emoticons=True):\n",
" \"\"\"\n",
" reading (multiple) files to one panda table.\n",
" \n",
" @param file_index_range: range of file's indices to read (eg `range(3)` to read the first three files)\n",
" @param only_emoticons: if True, only messages containing emoticons (aka smileys) are used. This classification is derived from Tools.Emoji_Distance\n",
" \"\"\"\n",
" assert np.min(file_index_range) >= 0 and np.max(file_index_range) < self.n_files\n",
" for i in file_index_range:\n",
" print(\"reading file: \" + self.json_files[i] + \"...\")\n",
@ -249,6 +279,9 @@
" print(\"imported \" + str(len(self.labels)) + \" samples\")\n",
" \n",
" def apply_stemming_and_lemmatization(self):\n",
" \"\"\"\n",
" apply stemming and lemmatization to plain text samples\n",
" \"\"\"\n",
" stemmer = SnowballStemmer(\"english\")\n",
" for key in self.plain_text.keys():\n",
" stemmed_sent = []\n",
@ -270,6 +303,9 @@
" self.plain_text[key] = lemmatized_sent\n",
" \n",
" def generate_emoji_count_and_weights(self):\n",
" \"\"\"\n",
" counting occurences of emojis\n",
" \"\"\"\n",
" self.emoji_count = {}\n",
" for e_list in self.emojis:\n",
" for e in set(e_list):\n",
@ -294,11 +330,23 @@
" self.emoji_count['X'] = 0\n",
" \n",
" def get_emoji_count(self):\n",
" \"\"\"\n",
" @return: descending list of tuples in form (<emoji as character>, <emoji count>) \n",
" \"\"\"\n",
" assert self.emoji_count is not None\n",
" \n",
" sorted_emoji_count = list(reversed(sorted(self.emoji_count.items(), key=operator.itemgetter(1))))\n",
" #display(sorted_emoji_count)\n",
" return sorted_emoji_count\n",
" \n",
" def filter_by_top_emojis(self,n_top = 20):\n",
" \"\"\"\n",
" filgter out messages not containing one of the `n_top` emojis\n",
" \n",
" @param n_top: number of top emojis used for filtering\n",
" \"\"\"\n",
" assert self.labels is not None # ← messages are already read in\n",
" \n",
" self.top_emojis = [x[0] for x in self.get_emoji_count()[:n_top]]\n",
" in_top = [sentiment_vector_to_emoji(x) in self.top_emojis for x in self.labels]\n",
" self.labels = self.labels[in_top]\n",
@ -307,6 +355,8 @@
" print(\"remaining samples after top emoji filtering: \", len(self.labels))\n",
" \n",
" def create_train_test_split(self, split = 0.1, random_state = 4222):\n",
" if self.X is not None:\n",
" sys.stderr.write(\"WARNING: overwriting existing train/test split \\n\")\n",
" self.X, self.Xt, self.y, self.yt = train_test_split(self.plain_text, self.labels, test_size=split, random_state=random_state)\n",
"\n"
]
@ -327,6 +377,16 @@
"class pipeline_manager(object):\n",
" @staticmethod\n",
" def load_pipeline_from_files(file_prefix:str, keras_models = [], all_models = []):\n",
" \"\"\"\n",
" load a pipeline from files. A pipeline should be represented by multiple model files in the form '<file_prefix>.<model_name>'\n",
" \n",
" @param file_prefix: basename of all files (without extension)\n",
" @param keras_models: list of keras models (keras model files, only extension name). Leave this list empty if this is not a keras pipeline\n",
" @param all_models: list of all models (including keras_models, only extension name).\n",
" \n",
" @return a pipeline manager object\n",
" \"\"\"\n",
" \n",
" pm = pipeline_manager(keras_models=keras_models)\n",
" pm.load(file_prefix, all_models)\n",
" return pm\n",
@ -335,6 +395,13 @@
" def create_keras_pipeline_with_vectorizer(vectorizer, layers, sdm:sample_data_manager):\n",
" '''\n",
" creates pipeline with vectorizer and keras classifier\n",
" \n",
" @param vectorizer: Vectorizer object. will be fitted with data provided by sdm\n",
" @param layers: list of keras layers. One keras layer is a tuple in form: (<#neurons:int>, <activation_func:str>)\n",
" @param sdm: sample data manager to get data for the vectorizer\n",
" \n",
" @return: a pipeline manager object\n",
" \n",
" '''\n",
" from keras.models import Sequential\n",
" from keras.layers import Dense\n",
@ -369,8 +436,13 @@
" @staticmethod\n",
" def create_pipeline_with_classifier_and_vectorizer(vectorizer, classifier, sdm:sample_data_manager = None):\n",
" '''\n",
" creates a pipeline with vectorizer and classifier for non keras classifiers\n",
" if sample data manager is given, the vectorizer will be also fitted!\n",
" creates pipeline with vectorizer and non-keras classifier\n",
" \n",
" @param vectorizer: Vectorizer object. will be fitted with data provided by sdm\n",
" @param classifier: unfitted classifier object (should be compatible with all sklearn classifiers)\n",
" @param sdm: sample data manager to get data for the vectorizer\n",
" \n",
" @return: a pipeline manager object\n",
" '''\n",
" if sdm is not None:\n",
" if sdm.X is None:\n",
@ -387,11 +459,24 @@
" return pipeline_manager(pipeline=pipeline, keras_models=[])\n",
" \n",
" def __init__(self, pipeline = None, keras_models = []):\n",
" \"\"\"\n",
" constructor\n",
" \n",
" @param pipeline: a sklearn pipeline\n",
" @param keras_models: list of keras steps in pipeline. Neccessary because saving and loading from keras models differs from the scikit ones\n",
" \"\"\"\n",
" \n",
" self.pipeline = pipeline\n",
" self.additional_objects = {}\n",
" self.keras_models = keras_models\n",
" \n",
" def save(self, prefix:str):\n",
" \"\"\"\n",
" saving the pipeline. It generates one file per model in the form: '<prefix>.<model_name>'\n",
" \n",
" @param prefix: file prefix for all models\n",
" \"\"\"\n",
" \n",
" print(self.keras_models)\n",
" # doing this like explained here: https://stackoverflow.com/a/43415459\n",
" for step in self.pipeline.named_steps:\n",
@ -407,13 +492,20 @@
" import __main__ as main\n",
" if not hasattr(main, '__file__'):\n",
" display(\"saved pipeline. It can be loaded the following way:\")\n",
" display(Markdown(\"> ```\\n\"+load_command+\"\\n```\"))\n",
" display(Markdown(\"> ```\\n\"+load_command+\"\\n```\")) # ← if we're in jupyter, print the fancy way :)\n",
" else:\n",
" print(\"saved pipeline. It can be loaded the following way:\")\n",
" print(load_command)\n",
" \n",
" \n",
" def load(self, prefix:str, models = []):\n",
" \"\"\"\n",
" load a pipeline. A pipeline should be represented by multiple model files in the form '<prefix>.<model_name>'\n",
" NOTE: keras model names (if there are some) have to be defined in self.keras_models first!\n",
" \n",
" @param prefix: the prefix for all model files\n",
" @param models: model_names to load\n",
" \"\"\"\n",
" self.pipeline = None\n",
" model_list = []\n",
" for model in models:\n",
@ -424,9 +516,11 @@
" self.pipeline = Pipeline(model_list)\n",
" \n",
" def fit(self,X,y):\n",
" \"\"\"fitting the pipeline\"\"\"\n",
" self.pipeline.fit(X,y)\n",
" \n",
" def predict(self,X):\n",
" \"\"\"predict\"\"\"\n",
" return self.pipeline.predict(X)\n",
" "
]
@ -446,10 +540,17 @@
"source": [
"class trainer(object):\n",
" def __init__(self, sdm:sample_data_manager, pm:pipeline_manager):\n",
" \"\"\"constructor\"\"\"\n",
" self.sdm = sdm\n",
" self.pm = pm\n",
" \n",
" def fit(self, max_size=10000, disabled_fit_steps=['vectorizer']):\n",
" \"\"\"\n",
" fitting data in the pipeline. Because we don't want to refit the vectorizer, the pipeline models containing the vectorizer have to be named explicitly\n",
" \n",
" @param max_size: don't train more examples than that number\n",
" @param disabled_fit_steps: list of pipeline steps that we want to prevent to refit. Normally all vectorizer steps\n",
" \"\"\"\n",
" # TODO: make batch fitting available here (eg: continous waiting for data and fitting them)\n",
" if self.sdm.X is None:\n",
" self.sdm.create_train_test_split()\n",
@ -475,7 +576,7 @@
" \n",
" def test(self):\n",
" '''\n",
" return: prediction:list, teacher:list\n",
" @return: prediction:list, teacher:list\n",
" '''\n",
" if self.sdm.X is None:\n",
" self.sdm.create_train_test_split()\n",
@ -510,7 +611,9 @@
"text": [
"reading file: ./data_en/2017-11-01.json...\n",
"imported 33368 samples\n",
"remaining samples after top emoji filtering: 26197\n"
"remaining samples after top emoji filtering: 26197\n",
"Epoch 1/1\n",
"100/100 [==============================] - 3s 27ms/step - loss: 0.1227\n"
]
}
],
@ -520,31 +623,12 @@
" # we are in an interactive environment (probably in jupyter)\n",
" # load data:\n",
" sdm = sample_data_manager.generate_and_read(path=\"./data_en/\", n_top_emojis=20, file_range=range(1))\n",
" "
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/1\n",
"10000/10000 [==============================] - 109s 11ms/step - loss: 0.0197\n"
]
}
],
"source": [
" #pm = pipeline_manager.create_keras_pipeline_with_vectorizer(vectorizer=TfidfVectorizer(stop_words='english'),\n",
" # layers=[(10000, 'relu'),(5000, 'relu'),(2500, 'relu'),(y1[0].shape[0],None)], sdm=sdm)\n",
" #pm = pipeline_manager.create_keras_pipeline_with_vectorizer(vectorizer=TfidfVectorizer(stop_words='english'),\\n\",\n",
" # layers=[(10000, 'relu'),(5000, 'relu'),(2500, 'relu'),(y1[0].shape[0],None)], sdm=sdm)\\n\",\n",
" pm = pipeline_manager.create_keras_pipeline_with_vectorizer(vectorizer=TfidfVectorizer(stop_words='english'),\n",
" layers=[(2500, 'relu'),(3,None)], sdm=sdm)\n",
" \n",
" tr = trainer(sdm=sdm, pm=pm)\n",
" tr.fit(10000)"
" tr.fit(100)"
]
},
{
@ -610,7 +694,9 @@
"metadata": {},
"source": [
"----\n",
"## Prediction"
"## Prediction\n",
"\n",
"* predict and save to `test.csv`"
]
},
{
@ -803,7 +889,9 @@
"metadata": {},
"source": [
"----\n",
"## Load classifier"
"## Load classifier\n",
"\n",
"* loading classifier and show a test widget"
]
},
{

631
Project/simple_approach/simple_twitter_learning.py Normal file
View File

@ -0,0 +1,631 @@
# coding: utf-8
# In[1]:
import pandas as pd
from IPython.display import clear_output, Markdown, Math
import ipywidgets as widgets
import os
import glob
import json
import numpy as np
import itertools
import sklearn.utils as sku
from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer, HashingVectorizer
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MultiLabelBinarizer
import nltk
from keras.models import load_model
from sklearn.externals import joblib
import pickle
import operator
from sklearn.pipeline import Pipeline
nltk.download('punkt')
nltk.download('averaged_perceptron_tagger')
nltk.download('wordnet')
# In[2]:
import sys
sys.path.append("..")
from Tools.Emoji_Distance import sentiment_vector_to_emoji
from Tools.Emoji_Distance import emoji_to_sentiment_vector
def emoji2sent(emoji_arr, only_emoticons=True):
return np.array([emoji_to_sentiment_vector(e, only_emoticons=only_emoticons) for e in emoji_arr])
def sent2emoji(sent_arr, custom_target_emojis=None, only_emoticons=True):
return [sentiment_vector_to_emoji(s, custom_target_emojis=custom_target_emojis, only_emoticons=only_emoticons) for s in sent_arr]
# In[3]:
SINGLE_LABEL = True
# ----
# ## classes and functions we are using later:
# ----
# * functions for selecting items from a set / list
# In[4]:
def latest(lst):
return lst[-1] if len(lst) > 0 else 'X'
def most_common(lst):
# trying to find the most common used emoji in the given lst
return max(set(lst), key=lst.count) if len(lst) > 0 else "X" # setting label to 'X' if there is an empty emoji list
# * our emoji blacklist (skin and sex modifiers)
# In[5]:
# defining blacklist for modifier emojis:
emoji_blacklist = set([
chr(0x1F3FB),
chr(0x1F3FC),
chr(0x1F3FD),
chr(0x1F3FE),
chr(0x1F3FF),
chr(0x2642),
chr(0x2640)
])
# * lemmatization helper functions
# In[6]:
from nltk.stem.snowball import SnowballStemmer
from nltk.stem import WordNetLemmatizer
from nltk import pos_tag
from nltk import word_tokenize
from nltk.corpus import wordnet
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
# ### sample data manager
# the sample data manager loads and preprocesses data
# most common way to use:
#
#
# * `sdm = sample_data_manager.generate_and_read(path:str, only_emoticons=True, apply_stemming=True, n_top_emojis=-1, file_range=None)`
#
# * Generates a sample_data_manager object and preprocess data in one step
#
# In[7]:
class sample_data_manager(object):
@staticmethod
def generate_and_read(path:str, only_emoticons=True, apply_stemming=True, n_top_emojis=-1, file_range=None):
"""
generate, read and process train data in one step.
@param path: folder containing json files to process
@param only_emoticons: if True, only messages containing emoticons (provided by Tools.Emoji_Distance) are used
@param apply_stemming: apply stemming and lemmatization on dataset
@param n_top_emojis: only use messages containing one of <`n_top_emojis`>-top emojis. set to `-1` to prevent top emoji filtering
@param file_range: range of file's indices to read (eg `range(3)` to read the first three files). If `None`: all files are read
@return: sample_data_manager object
"""
sdm = sample_data_manager(path)
sdm.read_files(file_index_range=range(sdm.n_files) if file_range is None else file_range, only_emoticons=only_emoticons)
if apply_stemming:
sdm.apply_stemming_and_lemmatization()
sdm.generate_emoji_count_and_weights()
if n_top_emojis > 0:
sdm.filter_by_top_emojis(n_top=n_top_emojis)
return sdm
def __init__(self, data_root_folder:str):
"""
constructor for manual initialization
@param data_root_folder: folder containing json files to process
"""
self.data_root_folder = data_root_folder
self.json_files = sorted(glob.glob(self.data_root_folder + "/*.json"))
self.n_files = len(self.json_files)
self.raw_data = None
self.emojis = None
self.plain_text = None
self.labels = None
self.emoji_count = None
self.emoji_weights = None
self.X = None
self.y = None
self.Xt = None
self.yt = None
self.top_emojis = None
def read_files(self, file_index_range:list, only_emoticons=True):
"""
reading (multiple) files to one panda table.
@param file_index_range: range of file's indices to read (eg `range(3)` to read the first three files)
@param only_emoticons: if True, only messages containing emoticons (aka smileys) are used. This classification is derived from Tools.Emoji_Distance
"""
assert np.min(file_index_range) >= 0 and np.max(file_index_range) < self.n_files
for i in file_index_range:
print("reading file: " + self.json_files[i] + "...")
if self.raw_data is None:
self.raw_data = pd.read_json(self.json_files[i], encoding="utf-8")
else:
self.raw_data = self.raw_data.append(pd.read_json(self.json_files[i], encoding="utf-8"))
self.emojis = self.raw_data['EMOJI']
self.plain_text = self.raw_data['text']
# replacing keywords. TODO: maybe these information can be extracted and used
self.plain_text = self.plain_text.str.replace("(<EMOJI>|<USER>|<HASHTAG>)","").str.replace("[" + "".join(list(emoji_blacklist)) + "]","")
# so far filtering for the latest emoji. TODO: maybe there are also better approaches
self.labels = emoji2sent([latest(e) for e in self.emojis], only_emoticons=only_emoticons )
# and filter out all samples we have no label for:
wrong_labels = np.isnan(np.linalg.norm(self.labels, axis=1))
self.labels = self.labels[np.invert(wrong_labels)]
self.plain_text = self.plain_text[np.invert(wrong_labels)]
self.emojis = self.emojis[np.invert(wrong_labels)]
print("imported " + str(len(self.labels)) + " samples")
def apply_stemming_and_lemmatization(self):
"""
apply stemming and lemmatization to plain text samples
"""
stemmer = SnowballStemmer("english")
for key in self.plain_text.keys():
stemmed_sent = []
for word in self.plain_text[key].split(" "):
word_stemmed = stemmer.stem(word)
stemmed_sent.append(word_stemmed)
stemmed_sent = (" ").join(stemmed_sent)
self.plain_text[key] = stemmed_sent
lemmatizer = WordNetLemmatizer()
for key in self.plain_text.keys():
lemmatized_sent = []
sent_pos = pos_tag(word_tokenize(self.plain_text[key]))
for word in sent_pos:
wordnet_pos = get_wordnet_pos(word[1].lower())
word_lemmatized = lemmatizer.lemmatize(word[0], pos=wordnet_pos)
lemmatized_sent.append(word_lemmatized)
lemmatized_sent = (" ").join(lemmatized_sent)
self.plain_text[key] = lemmatized_sent
def generate_emoji_count_and_weights(self):
"""
counting occurences of emojis
"""
self.emoji_count = {}
for e_list in self.emojis:
for e in set(e_list):
if e not in self.emoji_count:
self.emoji_count[e] = 0
self.emoji_count[e] += 1
emoji_sum = sum([self.emoji_count[e] for e in self.emoji_count])
self.emoji_weights = {}
for e in self.emoji_count:
# tfidf for emojis
self.emoji_weights[e] = np.log((emoji_sum / self.emoji_count[e]))
weights_sum= sum([self.emoji_weights[x] for x in self.emoji_weights])
# normalize:
for e in self.emoji_weights:
self.emoji_weights[e] = self.emoji_weights[e] / weights_sum
self.emoji_weights['X'] = 0 # dummy values
self.emoji_count['X'] = 0
def get_emoji_count(self):
"""
@return: descending list of tuples in form (<emoji as character>, <emoji count>)
"""
assert self.emoji_count is not None
sorted_emoji_count = list(reversed(sorted(self.emoji_count.items(), key=operator.itemgetter(1))))
#display(sorted_emoji_count)
return sorted_emoji_count
def filter_by_top_emojis(self,n_top = 20):
"""
filgter out messages not containing one of the `n_top` emojis
@param n_top: number of top emojis used for filtering
"""
assert self.labels is not None # ← messages are already read in
self.top_emojis = [x[0] for x in self.get_emoji_count()[:n_top]]
in_top = [sentiment_vector_to_emoji(x) in self.top_emojis for x in self.labels]
self.labels = self.labels[in_top]
self.plain_text = self.plain_text[in_top]
self.emojis = self.emojis[in_top]
print("remaining samples after top emoji filtering: ", len(self.labels))
def create_train_test_split(self, split = 0.1, random_state = 4222):
if self.X is not None:
sys.stderr.write("WARNING: overwriting existing train/test split \n")
self.X, self.Xt, self.y, self.yt = train_test_split(self.plain_text, self.labels, test_size=split, random_state=random_state)
# * the pipeline manager saves and stores sklearn pipelines. Keras models are handled differently, so the have to be named explicitly during save and load operations
# In[8]:
class pipeline_manager(object):
@staticmethod
def load_pipeline_from_files(file_prefix:str, keras_models = [], all_models = []):
"""
load a pipeline from files. A pipeline should be represented by multiple model files in the form '<file_prefix>.<model_name>'
@param file_prefix: basename of all files (without extension)
@param keras_models: list of keras models (keras model files, only extension name). Leave this list empty if this is not a keras pipeline
@param all_models: list of all models (including keras_models, only extension name).
@return a pipeline manager object
"""
pm = pipeline_manager(keras_models=keras_models)
pm.load(file_prefix, all_models)
return pm
@staticmethod
def create_keras_pipeline_with_vectorizer(vectorizer, layers, sdm:sample_data_manager):
'''
creates pipeline with vectorizer and keras classifier
@param vectorizer: Vectorizer object. will be fitted with data provided by sdm
@param layers: list of keras layers. One keras layer is a tuple in form: (<#neurons:int>, <activation_func:str>)
@param sdm: sample data manager to get data for the vectorizer
@return: a pipeline manager object
'''
from keras.models import Sequential
from keras.layers import Dense
if sdm.X is None:
sdm.create_train_test_split()
vec_train = vectorizer.fit_transform(sdm.X)
vec_test = vectorizer.transform(sdm.Xt)
# creating keras model:
model=Sequential()
keras_layers = []
first_layer = True
for layer in layers:
if first_layer:
model.add(Dense(units=layer[0], activation=layer[1], input_dim=vectorizer.transform([" "])[0]._shape[1]))
first_layer = False
else:
model.add(Dense(units=layer[0], activation=layer[1]))
model.compile(loss='mean_squared_error',
optimizer='adam')
pipeline = Pipeline([
('vectorizer',vectorizer),
('keras_model', model)
])
return pipeline_manager(pipeline=pipeline, keras_models=['keras_model'])
@staticmethod
def create_pipeline_with_classifier_and_vectorizer(vectorizer, classifier, sdm:sample_data_manager = None):
'''
creates pipeline with vectorizer and non-keras classifier
@param vectorizer: Vectorizer object. will be fitted with data provided by sdm
@param classifier: unfitted classifier object (should be compatible with all sklearn classifiers)
@param sdm: sample data manager to get data for the vectorizer
@return: a pipeline manager object
'''
if sdm is not None:
if sdm.X is None:
sdm.create_train_test_split()
vec_train = vectorizer.fit_transform(sdm.X)
vec_test = vectorizer.transform(sdm.Xt)
pipeline = Pipeline([
('vectorizer',vectorizer),
('classifier', classifier)
])
return pipeline_manager(pipeline=pipeline, keras_models=[])
def __init__(self, pipeline = None, keras_models = []):
"""
constructor
@param pipeline: a sklearn pipeline
@param keras_models: list of keras steps in pipeline. Neccessary because saving and loading from keras models differs from the scikit ones
"""
self.pipeline = pipeline
self.additional_objects = {}
self.keras_models = keras_models
def save(self, prefix:str):
"""
saving the pipeline. It generates one file per model in the form: '<prefix>.<model_name>'
@param prefix: file prefix for all models
"""
print(self.keras_models)
# doing this like explained here: https://stackoverflow.com/a/43415459
for step in self.pipeline.named_steps:
if step in self.keras_models:
self.pipeline.named_steps[step].model.save(prefix + "." + step)
else:
joblib.dump(self.pipeline.named_steps[step], prefix + "." + str(step))
load_command = "pipeline_manager.load_pipeline_from_files( '"
load_command += prefix + "', " + str(self.keras_models) + ", "
load_command += str(list(self.pipeline.named_steps.keys())) + ")"
import __main__ as main
if not hasattr(main, '__file__'):
display("saved pipeline. It can be loaded the following way:")
display(Markdown("> ```\n"+load_command+"\n```")) # ← if we're in jupyter, print the fancy way :)
else:
print("saved pipeline. It can be loaded the following way:")
print(load_command)
def load(self, prefix:str, models = []):
"""
load a pipeline. A pipeline should be represented by multiple model files in the form '<prefix>.<model_name>'
NOTE: keras model names (if there are some) have to be defined in self.keras_models first!
@param prefix: the prefix for all model files
@param models: model_names to load
"""
self.pipeline = None
model_list = []
for model in models:
if model in self.keras_models:
model_list.append((model, load_model(prefix + "." + model)))
else:
model_list.append((model, joblib.load(prefix+"." + model)))
self.pipeline = Pipeline(model_list)
def fit(self,X,y):
"""fitting the pipeline"""
self.pipeline.fit(X,y)
def predict(self,X):
"""predict"""
return self.pipeline.predict(X)
# * the trainer class passes Data from the sample manager to the pipeline manager
# In[9]:
class trainer(object):
def __init__(self, sdm:sample_data_manager, pm:pipeline_manager):
"""constructor"""
self.sdm = sdm
self.pm = pm
def fit(self, max_size=10000, disabled_fit_steps=['vectorizer']):
"""
fitting data in the pipeline. Because we don't want to refit the vectorizer, the pipeline models containing the vectorizer have to be named explicitly
@param max_size: don't train more examples than that number
@param disabled_fit_steps: list of pipeline steps that we want to prevent to refit. Normally all vectorizer steps
"""
# TODO: make batch fitting available here (eg: continous waiting for data and fitting them)
if self.sdm.X is None:
self.sdm.create_train_test_split()
disabled_fits = {}
disabled_fit_transforms = {}
named_steps = self.pm.pipeline.named_steps
for s in disabled_fit_steps:
# now it gets a little bit dirty:
# replace fit functions we don't want to call again (e.g. for vectorizers)
disabled_fits[s] = named_steps[s].fit
disabled_fit_transforms[s] = named_steps[s].fit_transform
named_steps[s].fit = lambda self, X, y=None: self
named_steps[s].fit_transform = named_steps[s].transform
self.pm.fit(X = self.sdm.X[:max_size], y = self.sdm.y[:max_size])
# restore replaced fit functions:
for s in disabled_fit_steps:
named_steps[s].fit = disabled_fits[s]
named_steps[s].fit_transform = disabled_fit_transforms[s]
def test(self):
'''
@return: prediction:list, teacher:list
'''
if self.sdm.X is None:
self.sdm.create_train_test_split()
return self.pm.predict(self.sdm.Xt), self.sdm.yt
# ----
# ## Train
# * when in notebook environment: run the stuff below:
# In[10]:
import __main__ as main
if not hasattr(main, '__file__'):
# we are in an interactive environment (probably in jupyter)
# load data:
sdm = sample_data_manager.generate_and_read(path="./data_en/", n_top_emojis=20, file_range=range(1))
#pm = pipeline_manager.create_keras_pipeline_with_vectorizer(vectorizer=TfidfVectorizer(stop_words='english'),\n",
# layers=[(10000, 'relu'),(5000, 'relu'),(2500, 'relu'),(y1[0].shape[0],None)], sdm=sdm)\n",
pm = pipeline_manager.create_keras_pipeline_with_vectorizer(vectorizer=TfidfVectorizer(stop_words='english'),
layers=[(2500, 'relu'),(3,None)], sdm=sdm)
tr = trainer(sdm=sdm, pm=pm)
tr.fit(100)
# ----
# ## save classifier
# In[13]:
import __main__ as main
if not hasattr(main, '__file__'):
pm.save('custom_classifier')
# ----
# ## Prediction
#
# * predict and save to `test.csv`
# In[14]:
import __main__ as main
if not hasattr(main, '__file__'):
pred, teacher = tr.test()
display(pred)
display(teacher)
print('prediction variance: ', np.linalg.norm(np.var(pred, axis=0)))
print('teacher variance: ', np.linalg.norm(np.var(teacher, axis=0)))
# build a dataframe to visualize test results:
testlist = pd.DataFrame({'text': sdm.Xt,
'teacher': sent2emoji(sdm.yt),
'teacher_sentiment': sdm.yt.tolist(),
'predict': sent2emoji(pred, custom_target_emojis=sdm.top_emojis),
'predicted_sentiment': pred.tolist()})
# display:
display(testlist.head())
# mean squared error:
teacher_sentiments = np.array([sample[1]['teacher_sentiment'] for sample in testlist.iterrows()])
predicted_sentiments = np.array([sample[1]['predicted_sentiment'] for sample in testlist.iterrows()])
mean_squared_error = ((teacher_sentiments - predicted_sentiments)**2).mean(axis=0)
print("Mean Squared Error: ", mean_squared_error)
print("Variance teacher: ", np.var(teacher_sentiments, axis=0))
print("Variance prediction: ", np.var(predicted_sentiments, axis=0))
# save to csv:
testlist.to_csv('test.csv')
# ----
# ## Load classifier
#
# * loading classifier and show a test widget
# In[15]:
import __main__ as main
if not hasattr(main, '__file__'):
try:
pm
except NameError:
pass
else:
del pm # delete existing pipeline manager if ther is one
pm = pipeline_manager.load_pipeline_from_files( 'custom_classifier', ['keras_model'], ['vectorizer', 'keras_model'])
lookup_emojis = [#'😂',
'😭',
'😍',
'😩',
'😊',
'😘',
'🙏',
'🙌',
'😉',
'😁',
'😅',
'😎',
'😢',
'😒',
'😏',
'😌',
'😔',
'😋',
'😀',
'😤']
out = widgets.Output()
t = widgets.Text()
b = widgets.Button(
description='get emoji',
disabled=False,
button_style='', # 'success', 'info', 'warning', 'danger' or ''
tooltip='Click me',
icon='check'
)
def handle_submit(sender):
with out:
clear_output()
with out:
pred = pm.predict([t.value])
display(Markdown("# Predicted Emoji " + str(sent2emoji(pred, lookup_emojis)[0])))
display(Markdown("# Sentiment Vector: $$ \pmatrix{" + str(pred[0,0]) +
"\\\\" + str(pred[0,1]) + "\\\\" + str(pred[0,2]) + "}$$"))
b.on_click(handle_submit)
display(t)
display(widgets.VBox([b, out]))