adds readme

.vscode/.gitignore

@@ -0,0 +1 @@
+/settings.json

ClassificationFake.py

@@ -0,0 +1,123 @@
+import numpy as np
+import pandas as pd
+from datetime import datetime
+from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
+from datasets import load_dataset
+from transformers.pipelines.pt_utils import KeyDataset
+
+
+#%%
+# prepare & define paths
+# install xformers (pip install xformers) for better performance
+###################
+# Setup directories
+# WD Michael
+wd = "/home/michael/Documents/PS/Data/collectTweets/"
+# WD Server
+# wd = '/home/yunohost.multimedia/polsoc/Politics & Society/TweetCollection/'
+
+# datafile input directory
+di = "data/IN/"
+
+# Tweet-datafile output directory
+ud = "data/OUT/"
+
+# Name of file that all senator data will be written to
+senCSV = "Tweets-Classified-Topic-Results.csv"
+
+# Name of Classify datafile
+senCSVClassifiedPrep = "Tweets-Classified-Fake-Prep.csv"
+senCSVClassifiedResult = "Tweets-Classified-Fake-Results.csv"
+
+# don't change this one
+senCSVPath = wd + ud + senCSV
+senCSVcClassificationPrepPath = wd + ud + senCSVClassifiedPrep
+senCSVcClassificationResultPath = wd + ud + senCSVClassifiedResult
+
+import sys
+funs = wd+"funs"
+sys.path.insert(1, funs)
+import CleanTweets
+
+
+#%%
+# get datafra,e
+dfClassify = pd.read_csv(senCSVPath, dtype=(object))
+def encode_labels(label):
+    if label == 'True':
+        return 'False'
+    elif label == 'False':
+        return 'True'
+    return 0
+dfClassify['output_label_topicCov'] = dfClassify['output_label_topicCov'].apply(encode_labels)
+dfClassify.to_csv("/home/michael/Documents/PS/Data/collectTweets/data/OUT/Tweets-Classified-Topic-Results.csv", encoding='utf-8')
+
+dfClassify = dfClassify[dfClassify['output_label_topicCov']=='True'] 
+
+# dataframe from csv
+dfClassify['fake'] = False
+
+
+#%%
+# https://huggingface.co/bvrau/covid-twitter-bert-v2-struth
+# HowTo:
+# https://huggingface.co/docs/transformers/main/en/model_doc/bert#transformers.BertForSequenceClassification
+# https://stackoverflow.com/questions/75932605/getting-the-input-text-from-transformers-pipeline
+pipe = pipeline("text-classification", model="/home/michael/Documents/PS/Data/collectTweets/models/FakeClass/2023-08-15_14-35-43/")
+model = AutoModelForSequenceClassification.from_pretrained("/home/michael/Documents/PS/Data/collectTweets/models/FakeClass/2023-08-15_14-35-43/")
+tokenizer = AutoTokenizer.from_pretrained("/home/michael/Documents/PS/Data/collectTweets/models/FakeClass/2023-08-15_14-35-43/")
+
+# Source https://www.kaggle.com/code/daotan/tweet-analysis-with-transformers-bert
+
+dfClassify['cleanContent'] = dfClassify['rawContent'].apply(CleanTweets.preprocess_text)
+
+
+#%%
+# remove empty rows
+dfClassify.cleanContent.replace('',np.nan,inplace=True)
+dfClassify.dropna(subset=['cleanContent'], inplace=True)
+
+#%%
+timeStart = datetime.now() # start counting execution time
+
+max_length = 128
+dfClassify['input_ids'] = dfClassify['cleanContent'].apply(lambda x: tokenizer(x, max_length=max_length, padding="max_length",)['input_ids'])
+#train.rename(columns={'target': 'labels'}, inplace=True)
+#train.head()
+
+# %%
+dfClassify.to_csv(senCSVcClassificationPrepPath, encoding='utf-8', columns=['id', 'cleanContent'])
+
+#%%
+dataset = load_dataset("csv", data_files=senCSVcClassificationPrepPath)
+
+# %%from datetime import datetime
+
+#from tqdm.auto import tqdm
+#for out in tqdm(pipe(KeyDataset(dataset['train'], "cleanContent"))):
+#    print(out)
+
+#%% 
+output_labels = []
+output_score = []
+for out in pipe(KeyDataset(dataset['train'], "cleanContent"), batch_size=8, truncation="only_first"):
+    output_labels.append(out['label'])
+    output_score.append(out['score'])
+    # [{'label': 'POSITIVE', 'score': 0.9998743534088135}]
+    # Exactly the same output as before, but the content are passed
+    # as batches to the model
+# %%
+dfClassify['output_label_fake'] = output_labels
+dfClassify['output_score_fake'] = output_score
+
+timeEnd = datetime.now()
+timeTotal = timeEnd - timeStart
+timePerTweet = timeTotal / 96
+
+print(f"Total classification execution time: {timeTotal} seconds")
+print(f"Time per tweet classification: {timePerTweet}")
+
+# %%
+dfClassify.to_csv(senCSVcClassificationResultPath, encoding='utf-8')
+
+# %%

ClassificationTopic.py

@@ -1,12 +1,9 @@
-import re
-import string
 import numpy as np
 import pandas as pd
 from datetime import datetime
 from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
 from datasets import load_dataset
 from transformers.pipelines.pt_utils import KeyDataset
-from funs.CleanTweets import remove_URL, remove_emoji, remove_html, remove_punct
 
 
 #%%
@@ -29,14 +26,20 @@ ud = "data/OUT/"
 senCSV = "SenatorsTweets-OnlyCov.csv"
 
 # Name of Classify datafile
-senCSVClassifiedPrep = "Tweets-Classified-Prep.csv"
+senCSVClassifiedPrep = "Tweets-Classified-Topic-Prep.csv"
-senCSVClassifiedResult = "Tweets-Classified-Results.csv"
+senCSVClassifiedResult = "Tweets-Classified-Topic-Results.csv"
 
 # don't change this one
 senCSVPath = wd + ud + senCSV
 senCSVcClassificationPrepPath = wd + ud + senCSVClassifiedPrep
 senCSVcClassificationResultPath = wd + ud + senCSVClassifiedResult
 
+import sys
+funs = wd+"funs"
+sys.path.insert(1, funs)
+import CleanTweets
+
+
 #%%
 # get datafra,e
 dfClassify = pd.read_csv(senCSVPath, dtype=(object))
@@ -50,17 +53,13 @@ dfClassify['fake'] = False
 # HowTo:
 # https://huggingface.co/docs/transformers/main/en/model_doc/bert#transformers.BertForSequenceClassification
 # https://stackoverflow.com/questions/75932605/getting-the-input-text-from-transformers-pipeline
-pipe = pipeline("text-classification", model="bvrau/covid-twitter-bert-v2-struth")
+pipe = pipeline("text-classification", model="/home/michael/Documents/PS/Data/collectTweets/models/CovClass/2023-08-15_05-56-50/")
-model = AutoModelForSequenceClassification.from_pretrained("bvrau/covid-twitter-bert-v2-struth")
+model = AutoModelForSequenceClassification.from_pretrained("/home/michael/Documents/PS/Data/collectTweets/models/CovClass/2023-08-15_05-56-50/")
-tokenizer = AutoTokenizer.from_pretrained("bvrau/covid-twitter-bert-v2-struth")
+tokenizer = AutoTokenizer.from_pretrained("/home/michael/Documents/PS/Data/collectTweets/models/CovClass/2023-08-15_05-56-50/")
 
 # Source https://www.kaggle.com/code/daotan/tweet-analysis-with-transformers-bert
 
-dfClassify['cleanContent'] = dfClassify['rawContent'].apply(remove_URL)
+dfClassify['cleanContent'] = dfClassify['rawContent'].apply(CleanTweets.preprocess_text)
-dfClassify['cleanContent'] = dfClassify['cleanContent'].apply(remove_emoji)
-dfClassify['cleanContent'] = dfClassify['cleanContent'].apply(remove_html)
-dfClassify['cleanContent'] = dfClassify['cleanContent'].apply(remove_punct)
-dfClassify['cleanContent'] = dfClassify['cleanContent'].apply(lambda x: x.lower())
 
 #%%
 # remove empty rows
@@ -97,8 +96,8 @@ for out in pipe(KeyDataset(dataset['train'], "cleanContent"), batch_size=8, trun
     # Exactly the same output as before, but the content are passed
     # as batches to the model
 # %%
-dfClassify['output_label'] = output_labels
+dfClassify['output_label_topicCov'] = output_labels
-dfClassify['output_score'] = output_score
+dfClassify['output_score_topicCov'] = output_score
 
 timeEnd = datetime.now()
 timeTotal = timeEnd - timeStart
@@ -111,3 +110,14 @@ print(f"Time per tweet classification: {timePerTweet}")
 dfClassify.to_csv(senCSVcClassificationResultPath, encoding='utf-8')
 
 # %%
+## corrections
+def encode_labels(label):
+    if label == 'real':
+        return 'True'
+    elif label == 'fake':
+        return 'False'
+    return 0
+dfClassify['output_label_topicCov'] = dfClassify['output_label_topicCov'].apply(encode_labels)
+dfClassify.to_csv(senCSVcClassificationResultPath, encoding='utf-8')
+#still wrong, will be corrected in ClassificationFake.py
+

README.md

@@ -1,7 +1,127 @@
-# How to use
+# Requirements
 
-Execute collect.py to scrape tweets and generate the ´ALL-SENATORS-TWEETS.csv´.
+- python 3.10+
+- snscrape 0.6.2.20230321+ (see git repo in this folder)
+- transformers 4.31.0
+- numpy 1.23.5
+- pandas 2.0.3
+- scikit-learn 1.3.0
+- torch 2.0.1
 
-Execute collectSenData.py to scrape senator data and generate ´ALL-SENATORS.csv´.
+# About
 
-All new files will be written to ´data/OUT/´. Necessary data has to be located in ´data/IN/´
+This collection of scripts scrapes tweets of US-senators in the time from 2020-01-01T00:00:00Z to 2023-01-03T00:00:00Z, scrapes account data of the senators, prepares the tweets for the training of a NLP-model, trains two models to (1) classify the tweets topic as covid or non-covid and (2) the tweets as either "fake news" tweets or "non-fake news" tweets.
+Training only works with a prepared dataset in which the tweets are pre classified.
+More info in the comments of the scripts.
+Due to time constraints, most of the code is procedurally coded and ugly but effective.
+
+# How to
+
+Tested on Ubuntu 22.04. 
+If needed, the virual environment can be exported and send to you.
+
+All files in the folder data/in have to exist in order to execute the scripts.
+Execute in the following order:
+
+01 collect.py (see more for further info on scraping)
+02 collectSenData.py
+03 cleanTweets
+04 preTestClassification.py
+05 trainTopic.py
+06 trainFake.py
+07 ClassificationFake.py
+08 ClassificationTopic.py
+
+# Files & Folders
+
+├── data
+│   ├── IN
+│   │   ├── counterKeywordsFinal.txt
+│   │   ├── counterKeywords.txt
+│   │   ├── keywords-raw.txt
+│   │   ├── keywords.txt
+│   │   ├── own_keywords.txt
+│   │   ├── pretest-tweets_fake.txt				contains tweet ids for pretest
+│   │   ├── pretest-tweets_not_fake.txt			contains tweet ids for pretest
+│   │   └── senators-raw.csv					senator datafile
+│   ├── OUT
+│   │   ├── ALL-SENATORS-TWEETS.csv
+│   │   ├── graphs
+│   │   │   ├── Timeline.png
+│   │   │   ├── Wordcloud-All.png
+│   │   │   └── Wordcloud-Cov.png
+│   │   ├── Pretest-Prep.csv
+│   │   ├── Pretest-Results.csv
+│   │   ├── Pretest-SENATORS-TWEETS.csv
+│   │   ├── profiles							dataset profiles
+│   │   │   ├── AllTweets.html
+│   │   │   └── CovTweets.html
+│   │   ├── SenatorsTweets-Final.csv
+│   │   ├── SenatorsTweets-OnlyCov.csv
+│   │   ├── SenatorsTweets-train-CovClassification.csv
+│   │   ├── SenatorsTweets-train-CovClassificationTRAIN.csv
+│   │   ├── SenatorsTweets-train-CovClassification.tsv
+│   │   ├── SenatorsTweets-train-FakeClassification.csv
+│   │   ├── SenatorsTweets-train-FakeClassificationTRAIN.csv
+│   │   ├── SenatorsTweets-train-FakeClassification.tsv
+│   │   ├── SenatorsTweets-Training.csv
+│   │   ├── SenatorsTweets-Training_WORKING-COPY.csv
+│   │   ├── topClass-PRETEST-Prep.csv
+│   │   ├── topClass-PRETEST-Results.csv
+│   │   ├── Tweets-All-slices.zip
+│   │   ├── Tweets-Classified-Fake-Prep.csv
+│   │   ├── Tweets-Classified-Fake-Results.csv
+│   │   ├── Tweets-Classified-Prep.csv
+│   │   ├── Tweets-Classified-Topic-Prep.csv
+│   │   ├── Tweets-Classified-Topic-Results.csv
+│   │   └── Tweets-Stub.csv
+├── funs
+│   ├── CleanTweets.py					2023-01-03T00:00:00Z		multiple functions to clean tweet contents for NLN-processing
+│   ├── ClearDupes.py							function for deletion of duplicate keywords
+│   ├── __init__.py
+│   ├── Scrape.py								scraper functions to be used for multiprocessing
+│   └── TimeSlice.py							time slice script to slice the time span in 24 slices, speeds up scraping through multiprocessing
+├── log											logs of the scraping process
+│   ├── log_2023-06-23_21-06-10_err.log
+│   ├── log_2023-06-23_21-06-10.log
+│   └── log_2023-06-23_21-06-10_missing.log
+├── models
+│   ├── CovClass								Covid tweet classification model
+│   │   └── 2023-08-15_05-56-50
+│   │       ├── 2023-08-15_05-56-50.csv			training output
+│   │       ├── config.json
+│   │       ├── pytorch_model.bin
+│   │       ├── special_tokens_map.json
+│   │       ├── tokenizer_config.json
+│   │       ├── tokenizer.json
+│   │       └── vocab.txt
+│   └── FakeClass								Fake tweet classification model
+│       └── 2023-08-15_14-35-43
+│           ├── 2023-08-15_14-35-43.csv			training output
+│           ├── config.json
+│           ├── pytorch_model.bin
+│           ├── special_tokens_map.json
+│           ├── tokenizer_config.json
+│           ├── tokenizer.json
+│           └── vocab.txt
+├── snscrape									contains snscrape 0.6.2.20230321+ git repo
+├── ClassificationFake.py						classifies tweets as fake or non-fake, saves:
+│													Tweets-Classified-Fake-Prep.csv		- prepared training dataset
+│													Tweets-Classified-Fake-Results.csv	- Tweets-Classified-Topic-Results.csv with cov classification results
+├── ClassificationTopic.py						classifies tweet topic, saves: 
+│													Tweets-Classified-Topic-Prep.csv 	- prepared training dataset
+│													Tweets-Classified-Topic-Results.csv	- SenatorsTweets-OnlyCov.csv with cov classification results
+├── cleanTweets.py								Curates keywordlists 
+│												Merges senator and tweet datasets
+│												Creates multiple datasets:
+│													SenatorsTweets-Final.csv	- all tweets with keyword columns
+│													SenatorsTweets-OnlyCov.csv	- only covid tweets, filtered by keywordlist
+│													SenatorsTweets-Training.csv	- training dataset, containing ~1800 randomly selected tweets from SenatorsTweets-OnlyCov.csv
+├── collect.py									scrapes tweets, saves to ALL-SENATORS-TWEETS.csv
+├── collectSenData.py							scrapes senator account data, saves to ALL-SENATORS.csv
+├── createGraphs.py								creates wordcloud & timeline graphs
+├── preTestClassification.py					pretest script that uses bvrau/covid-twitter-bert-v2-struth to analyze 100 preclassified tweets
+├── profiler.py									creates dataset profiles
+├── README.md									readme
+├── trainFake.py								training script for the fake tweet classification model
+└── trainTopic.py								training script for the tweet topic classification model

analyze.py

@@ -1,129 +0,0 @@
-import re
-import string
-import numpy as np
-import pandas as pd
-from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
-from datasets import load_dataset
-from transformers.pipelines.pt_utils import KeyDataset
-from funs.CleanTweets import remove_URL, remove_emoji, remove_html, remove_punct
-
-
-#%%
-# prepare
-# install xformers (pip install xformers) for better performance
-###################
-# Setup directories
-# WD Michael
-wd = "/home/michael/Documents/PS/Data/collectTweets/"
-# WD Server
-# wd = '/home/yunohost.multimedia/polsoc/Politics & Society/TweetCollection/'
-
-# datafile input directory
-di = "data/IN/"
-
-# Tweet-datafile output directory
-ud = "data/OUT/"
-
-# Name of file that all senator data will be written to
-senCSV = "ALL-SENATORS-TWEETS.csv"
-
-# Name of new datafile generated
-senCSVc = "Tweets-Stub.csv"
-
-# Name of pretest files
-preTestIDsFake = "pretest-tweets_fake.txt"
-preTestIDsNot = "pretest-tweets_not_fake.txt"
-
-# Name of pretest datafile
-senCSVPretest = "Pretest.csv"
-senCSVPretestPrep = "Pretest-Prep.csv"
-senCSVPretestResult = "Pretest-Results.csv"
-
-
-# don't change this one
-senCSVPath = wd + ud + senCSV
-senCSVcPath = wd + ud + senCSVc
-senCSVcPretestPath = wd + ud + senCSVPretest
-senCSVcPretestPrepPath = wd + ud + senCSVPretestPrep
-senCSVcPretestResultPath = wd + ud + senCSVPretestResult
-preTestIDsFakePath = wd + di + preTestIDsFake
-preTestIDsNotPath = wd + di + preTestIDsNot
-
-# List of IDs to select
-# Read the IDs from a file
-preTestIDsFakeL = []
-preTestIDsNotL  = []
-with open(preTestIDsFakePath, "r") as file:
-    lines = file.readlines()
-    for line in lines:
-        tid = line.strip()  # Remove the newline character
-        preTestIDsFakeL.append(tid)
-with open(preTestIDsNotPath, "r") as file:
-    lines = file.readlines()
-    for line in lines:
-        tid = line.strip()  # Remove the newline character
-        preTestIDsNotL.append(tid)
-
-# Select rows based on the IDs
-df = pd.read_csv(senCSVPath, dtype=(object))
-#%%
-# Create pretest dataframe
-dfPreTest = df[df['id'].isin(preTestIDsFakeL)].copy()
-dfPreTest['fake'] = True
-dfPreTest = pd.concat([dfPreTest, df[df['id'].isin(preTestIDsNotL)]], ignore_index=True)
-dfPreTest['fake'] = dfPreTest['fake'].fillna(False)
-
-#%%
-# https://huggingface.co/bvrau/covid-twitter-bert-v2-struth
-# HowTo:
-# https://huggingface.co/docs/transformers/main/en/model_doc/bert#transformers.BertForSequenceClassification
-# https://stackoverflow.com/questions/75932605/getting-the-input-text-from-transformers-pipeline
-pipe = pipeline("text-classification", model="bvrau/covid-twitter-bert-v2-struth")
-model = AutoModelForSequenceClassification.from_pretrained("bvrau/covid-twitter-bert-v2-struth")
-tokenizer = AutoTokenizer.from_pretrained("bvrau/covid-twitter-bert-v2-struth")
-
-# Source https://www.kaggle.com/code/daotan/tweet-analysis-with-transformers-bert
-
-dfPreTest['cleanContent'] = dfPreTest['rawContent'].apply(remove_URL)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(remove_emoji)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(remove_html)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(remove_punct)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(lambda x: x.lower())
-
-#%%
-max_length = 128
-dfPreTest['input_ids'] = dfPreTest['cleanContent'].apply(lambda x: tokenizer(x, max_length=max_length, padding="max_length",)['input_ids'])
-#train.rename(columns={'target': 'labels'}, inplace=True)
-#train.head()
-
-# %%
-dfPreTest.to_csv(senCSVcPretestPrepPath, encoding='utf-8', columns=['id', 'cleanContent'])
-
-
-#%%
-dataset = load_dataset("csv", data_files=senCSVcPretestPrepPath)
-
-# %%
-results = pipe(KeyDataset(dataset, "text"))
-# %%
-#from tqdm.auto import tqdm
-#for out in tqdm(pipe(KeyDataset(dataset['train'], "cleanContent"))):
-#    print(out)
-
-#%% 
-output_labels = []
-output_score = []
-for out in pipe(KeyDataset(dataset['train'], "cleanContent"), batch_size=8, truncation="only_first"):
-    output_labels.append(out['label'])
-    output_score.append(out['score'])
-    # [{'label': 'POSITIVE', 'score': 0.9998743534088135}]
-    # Exactly the same output as before, but the content are passed
-    # as batches to the model
-# %%
-dfPreTest['output_label'] = output_labels
-dfPreTest['output_score'] = output_score
-
-# %%
-dfPreTest.to_csv(senCSVcPretestResultPath, encoding='utf-8')
-
-# %%

cleanTweets.py

@@ -9,9 +9,12 @@ Created on Mon Jun 26 20:36:43 2023
 import pandas as pd
 # import pyreadstat
 import numpy as np
-from funs.ClearDupes import deDupe
+import sys
 
 
+# Seet for training dataset generation
+seed = 86431891
+
 ###################
 # Setup directories
 # WD Michael
@@ -34,17 +37,24 @@ senDataset = "senators-raw.csv"
 # Name of new datafile generated
 senCSVc = "SenatorsTweets-Final"
 senCSVcCov = "SenatorsTweets-OnlyCov"
+senCSVcTrain = "SenatorsTweets-Training"
 
 # don't change this one
 senCSVPath = wd + ud + senCSV
 senCSVcPath = wd + ud + senCSVc + ".csv"
 senCSVcCovPath = wd + ud + senCSVcCov + ".csv"
+senCSVcTrainPath = wd + ud + senCSVcTrain + ".csv"
 senSAVcPath = wd + ud + senCSV + ".sav"
 senDTAcPath = wd + ud + senCSV + ".dta"
 senDatasetPath = wd + di + senDataset
 
 df = pd.read_csv(senCSVPath, dtype=(object))
 
+## Import own functions
+funs = wd+"funs"
+sys.path.insert(1, funs)
+from ClearDupes import deDupe
+
 mixed_columns = df.columns[df.nunique() != len(df)]
 print(mixed_columns)
 
@@ -188,7 +198,6 @@ dfCov = dfAll[dfAll['contains_counterKeyword']==False]
 dfCov = dfCov[dfCov['contains_keyword']==True]
 dfCov = dfCov.drop(columns=['contains_counterKeyword', 'counterKeywords'])
 
-
 #%%
 # create column with tweet length
 
@@ -211,3 +220,14 @@ dfCov.to_csv(senCSVcCovPath, encoding='utf-8', index_label = 'id')
 # =========================
 
 # %%
+# Create training dataset
+np.random.seed(seed); 
+dfTrain = pd.dfCov(np.random.rand(1800))
+# %%
+# Create training dataset
+np.random.seed(seed); 
+dfTrain = dfCov.loc[np.random.choice(dfCov.index, 1800, replace=False)]
+dfTrain = dfTrain[['tid', 'date', 'rawContent']]
+dfTrain['topicCovid'] = True
+dfTrain['fake'] = False
+dfTrain.to_csv(senCSVcTrainPath, encoding='utf-8')

collect.py

@@ -66,7 +66,6 @@ which is the final output.
 import os
 import pandas as pd
 import glob
-import time
 import sys
 from datetime import datetime
 import concurrent.futures
@@ -149,10 +148,12 @@ tweetDFColumns = [
 ################## do NOT change anything below this line ###################
 #############################################################################
 
-## Import functions
+## Import own functions
-from funs.TimeSlice import *
+funs = wd+"funs"
-from funs.ClearDupes import deDupe
+sys.path.insert(1, funs)
-from funs.Scrape import scrapeTweets
+from TimeSlice import get_Tslices
+from ClearDupes import deDupe
+from Scrape import scrapeTweets
 
 ################### 
 # Create logfile & log all outputs

data/IN/counterKeywordsFinal.txt

@@ -0,0 +1,23 @@
+meth
+gun violence
+flu season
+vaping
+chd
+addiction
+indigenous women
+separating children
+tobacco
+e-cigarette
+muslim ban
+soleimani
+cocaine
+separating families
+muslim travel ban
+usmca trade deal
+shooting
+overdose
+separated children
+coronary heart disease
+gun-violence
+opioid
+flu-season

data/IN/keywords.txt

@@ -18,44 +18,43 @@ socialdistancing
 wear a mask
 lockdown
 covd
-Coronavirus
+coronavirus
-Koronavirus
+koronavirus
-Corona
+corona
-CDC
+cdc
-Wuhancoronavirus
+wuhancoronavirus
-Wuhanlockdown
+wuhanlockdown
-Ncov
+ncov
-Wuhan
+wuhan
-N95
+n95
-Kungflu
+kungflu
-Epidemic
+epidemic
 outbreak
-Sinophobia
+sinophobia
-China
 covid-19
 corona virus
 covid
 covid19
 sars-cov-2
-COVIDー19
+covidー19
-COVD
+covd
 pandemic
 coronapocalypse
 canceleverything
-Coronials
+coronials
-SocialDistancingNow
+socialdistancingnow
-Social Distancing
+social distancing
-SocialDistancing
+socialdistancing
 panicbuy
 panic buy
 panicbuying
 panic buying
-14DayQuarantine
+14dayquarantine
-DuringMy14DayQuarantine
+duringmy14dayquarantine
 panic shop
 panic shopping
 panicshop
-InMyQuarantineSurvivalKit
+inmyquarantinesurvivalkit
 panic-buy
 panic-shop
 coronakindness
@@ -65,7 +64,7 @@ chinesevirus
 stayhomechallenge
 stay home challenge
 sflockdown
-DontBeASpreader
+dontbeaspreader
 lockdown
 lock down
 shelteringinplace
@@ -79,13 +78,13 @@ flatten the curve
 china virus
 chinavirus
 quarentinelife
-PPEshortage
+ppeshortage
 saferathome
 stayathome
 stay at home
 stay home
 stayhome
-GetMePPE
+getmeppe
 covidiot
 epitwitter
 pandemie
@@ -93,7 +92,7 @@ wear a mask
 wearamask
 kung flu
 covididiot
-COVID__19
+covid__19
 omicron
 variant
 vaccine
@@ -139,9 +138,7 @@ work from home
 workfromhome
 working from home
 workingfromhome
-ppe
 n95
-ppe
 n95
 covidiots
 covidiots

data/OUT/.gitignore

@@ -0,0 +1,8 @@
+/ALL-SENATORS-TWEETS.csv
+/Pretest-Prep.csv
+/Pretest-Results.csv
+/Pretest-SENATORS-TWEETS.csv
+/SenatorsTweets-Final.csv
+/SenatorsTweets-OnlyCov.csv
+/Tweets-Classified-Prep.csv
+/Tweets-Stub.csv

data/OUT/graphs/.gitignore

@@ -0,0 +1,3 @@
+/Timeline.png
+/Wordcloud-All.png
+/Wordcloud-Cov.png

funs/CleanTweets.py

@@ -1,10 +1,20 @@
 import re
 import string
 
-def remove_URL(text):
+def preprocess_roberta(text): # https://huggingface.co/cardiffnlp/twitter-roberta-base-sep2022
-    url = re.compile(r'https?://\S+|www\.\S+')
+    preprocessed_text = []
-    return url.sub(r'', text)
+    for t in text.split():
+        if len(t) > 1:
+            t = '@user' if t[0] == '@' and t.count('@') == 1 else t
+            t = 'http' if t.startswith('http') else t
+        preprocessed_text.append(t)
+    return ' '.join(preprocessed_text)
 
+def remove_URL(text):
+    try: 
+        url = re.compile(r'https?://\S+|www\.\S+')
+    except: print(text)
+    return url.sub(r'', text)
 
 def remove_emoji(text):
     emoji_pattern = re.compile(
@@ -19,21 +29,61 @@ def remove_emoji(text):
         flags=re.UNICODE)
     return emoji_pattern.sub(r'', text)
 
-
 def remove_html(text):
     html = re.compile(r'<.*?>|&([a-z0-9]+|#[0-9]{1,6}|#x[0-9a-f]{1,6});')
     return re.sub(html, '', text)
 
-
 def remove_punct(text):
     table = str.maketrans('', '', string.punctuation)
     return text.translate(table)
 
-def clean_all(text):
+def remove_nonascii(text):
-    if not isinstance(text, str):
+    return re.sub(r'[^\x00-\x7F]+', '', text)
-        text = str(text) # Convert non-string values to string
+
+def remove_spec(text):
+    text = re.sub(r'&amp;?', r'and', text)
+    text = re.sub(r'&lt;', r'<', text)
+    return re.sub(r'&gt;', r'>', text)
+
+def remove_spaces(text): # also new line chars and to lower case
+    text = re.sub(r'&lt;', r'<', text)
+    text = " ".join(text.splitlines()) # remove newline characters
+    text = text.lower()
+    text = text.strip()
+    return re.sub(r'\s{2,}', ' ', text)
+
+def remove_retw(text):
+    text = re.sub(r'(RT|rt)[ ]*@[ ]*[\S]+', '', text)
+    return re.sub(r'@[\S]+', '', text)
+
+def preprocess_text(text):
     text = remove_URL(text)
     text = remove_emoji(text)
     text = remove_html(text)
     text = remove_punct(text)
+    text = remove_nonascii(text)
+    text = remove_spec(text)
+    text = remove_spaces(text)
+    text = remove_retw(text)
     return text
+
+def preprocess_text_series(series):
+    series = series.apply(remove_URL)
+    series = series.apply(remove_emoji)
+    series = series.apply(remove_html)
+    series = series.apply(remove_punct)
+    series = series.apply(remove_nonascii)
+    series = series.apply(remove_spec)
+    series = series.apply(remove_spaces)
+    series = series.apply(remove_retw)
+    return series
+
+# Check all functions:
+input_text = """
+    Check out this amazing website: https://www.example.com! 😃
+    <html>This is an HTML tag.</html>
+    RT @user123: Just received a package from @companyXYZ. It's awesome! 📦
+    This is a test text with lots of punctuations!!! Can't wait to see more...
+"""
+processed_text = preprocess_text(input_text)
+# print(processed_text)

models/CovClass/2023-08-15_01-55-11/statsTopicClassification-2023-08-15_01-53-12.csv

@@ -0,0 +1,7 @@
+epoch,Training Loss,Valid. Loss,Valid. Accur.,Training Time,Validation Time
+1,0.39025546515679493,0.40877932761593355,0.9103260869565217,0:10:21,0:00:40
+2,0.3057803610952067,0.3502063500978377,0.9103260869565217,0:10:53,0:00:43
+3,0.17910970049364833,0.27903796154904464,0.9375,0:10:30,0:00:38
+4,0.09279396105943587,0.41342766528301267,0.904891304347826,0:11:03,0:00:43
+5,0.06132459050129317,0.4468563502887264,0.9239130434782609,0:12:07,0:00:44
+6,0.04195396880810895,0.4350045176675928,0.9266304347826086,0:11:21,0:00:40

models/CovClass/2023-08-15_02-14-21/2023-08-15_02-14-21.csv

@@ -0,0 +1,7 @@
+epoch,Training Loss,Valid. Loss,Valid. Accur.,Training Time,Validation Time
+1,0.6699380816093513,0.6216431430407933,0.6964285714285714,0:01:03,0:00:02
+2,0.6649796058024678,0.621175297669002,0.6964285714285714,0:01:03,0:00:01
+3,0.642247314964022,0.6377243144171578,0.6964285714285714,0:01:05,0:00:02
+4,0.6300328698541436,0.6038827853543418,0.6964285714285714,0:01:04,0:00:02
+5,0.544977219509227,0.6619421115943364,0.625,0:01:02,0:00:02
+6,0.3951783587357828,0.48477122613361906,0.7857142857142857,0:01:05,0:00:01

models/CovClass/2023-08-15_05-56-50/2023-08-15_05-56-50.csv

@@ -0,0 +1,7 @@
+epoch,Training Loss,Valid. Loss,Valid. Accur.,Training Time,Validation Time
+1,0.5610552686641376,0.4569096086310089,0.9116022099447514,0:37:20,0:00:31
+2,0.43647773836513126,0.5441495520680196,0.9005524861878453,0:36:14,0:00:30
+3,0.288773139899344,0.43471020716692715,0.9392265193370166,0:36:10,0:00:29
+4,0.19330878817686287,0.4555162174395349,0.9281767955801105,0:36:17,0:00:30
+5,0.09109889855869348,0.5060150003684702,0.9281767955801105,0:36:13,0:00:30
+6,0.05734757932275739,0.6043995772428771,0.9226519337016574,0:36:11,0:00:31

models/FakeClass/2023-08-15_12-03-05/2023-08-15_12-03-05.csv

@@ -0,0 +1,7 @@
+epoch,Training Loss,Valid. Loss,Valid. Accur.,Training Time,Validation Time
+1,0.21681843259712502,0.0005426188472483773,1.0,0:01:13,0:00:02
+2,0.00016121647037353423,0.0002873415878639207,1.0,0:01:12,0:00:02
+3,6.752021149355535e-05,0.00024319994372490328,1.0,0:01:12,0:00:02
+4,4.7950222591787355e-05,0.00022139604243420763,1.0,0:01:13,0:00:02
+5,3.99839740138679e-05,0.00021302999493855168,1.0,0:01:11,0:00:02
+6,3.5356899656214995e-05,0.00020912183117616223,1.0,0:01:13,0:00:02

preTestClassification.py

@@ -1,13 +1,8 @@
-import re
-import string
-import numpy as np
 import pandas as pd
 from datetime import datetime
 from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
 from datasets import load_dataset
 from transformers.pipelines.pt_utils import KeyDataset
-from funs.CleanTweets import remove_URL, remove_emoji, remove_html, remove_punct
-
 
 #%%
 # prepare
@@ -40,7 +35,6 @@ senCSVPretest = "Pretest.csv"
 senCSVPretestPrep = "Pretest-Prep.csv"
 senCSVPretestResult = "Pretest-Results.csv"
 
-
 # don't change this one
 senCSVPath = wd + ud + senCSV
 senCSVcPath = wd + ud + senCSVc
@@ -50,6 +44,11 @@ senCSVcPretestResultPath = wd + ud + senCSVPretestResult
 preTestIDsFakePath = wd + di + preTestIDsFake
 preTestIDsNotPath = wd + di + preTestIDsNot
 
+import sys
+funs = wd+"funs"
+sys.path.insert(1, funs)
+import CleanTweets
+
 # List of IDs to select
 # Read the IDs from a file
 preTestIDsFakeL = []
@@ -85,11 +84,7 @@ tokenizer = AutoTokenizer.from_pretrained("bvrau/covid-twitter-bert-v2-struth")
 
 # Source https://www.kaggle.com/code/daotan/tweet-analysis-with-transformers-bert
 
-dfPreTest['cleanContent'] = dfPreTest['rawContent'].apply(remove_URL)
+dfPreTest['cleanContent'] = dfPreTest['rawContent'].apply(CleanTweets.preprocess_text)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(remove_emoji)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(remove_html)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(remove_punct)
-dfPreTest['cleanContent'] = dfPreTest['cleanContent'].apply(lambda x: x.lower())
 
 #%%
 timeStart = datetime.now() # start counting execution time

profiler.py

@@ -0,0 +1,55 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Tue Aug  8 14:49:02 2023
+
+@author: michael
+"""
+
+import pandas as pd
+import pandas_profiling as pp
+import numpy
+  
+###################
+# Setup directories
+# WD Michael
+wd = "/home/michael/Documents/PS/Data/collectTweets/"
+# WD Server
+# wd = '/home/yunohost.multimedia/polsoc/Politics & Society/TweetCollection/'
+
+# datafile input directory
+di = "data/IN/"
+
+# Tweet-datafile output directory
+ud = "data/OUT/"
+
+# Name of file that all senator data will be written to
+senCSV = "ALL-SENATORS-TWEETS.csv"
+
+# Name of file that all senator data will be written to
+senDataset = "senators-raw.csv"
+
+# Name of new datafile generated
+senCSVc = "SenatorsTweets-Final"
+senCSVcCov = "SenatorsTweets-OnlyCov"
+
+# don't change this one
+senCSVPath = wd + ud + senCSV
+senCSVcPath = wd + ud + senCSVc + ".csv"
+senCSVcCovPath = wd + ud + senCSVcCov + ".csv"
+senSAVcPath = wd + ud + senCSV + ".sav"
+senDTAcPath = wd + ud + senCSV + ".dta"
+senDatasetPath = wd + di + senDataset
+  
+# forming dataframe and printing
+df = pd.read_csv(senCSVPath, dtype=(object))
+  
+# forming ProfileReport and save
+# as output.html file
+profileAll = pp.ProfileReport(df, minimal=True)
+profileAll.to_file("data/OUT/profiles/AllTweets.html")
+
+df = pd.read_csv(senCSVcCovPath, dtype=(object))
+
+profileAll = pp.ProfileReport(df, minimal=True)
+profileAll.to_file("data/OUT/profiles/CovTweets.html")

repairmystupidity.py

@@ -0,0 +1,35 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Mon Aug 14 20:47:22 2023
+
+@author: michael
+"""
+import pandas as pd
+
+wd = "/home/michael/Documents/PS/Data/collectTweets/"
+
+# datafile input directory
+di = "data/IN/"
+
+# Tweet-datafile output directory
+ud = "data/OUT/"
+
+falsch = wd + ud + "SenatorsTweets-Training_WORKING-COPY-correct.csv"
+richtig = wd + ud + "SenatorsTweets-Training.csv"
+correct = wd + ud + "SenatorsTweets-Training_WORKING-COPY-correct2.csv"
+
+# Name of new datafile generated
+senCSVprep = "SenatorsTweets-Training_WORKING-COPY-prepared"
+
+# don't change this one
+falsch = pd.read_csv(falsch, dtype=(object), sep=";")
+richtig = pd.read_csv(richtig, dtype=(object))
+
+df = pd.merge(falsch,richtig[['tid','rawContent', 'date']],on='tid', how='left')
+df.drop(columns=['rawContent_x', 'date_x'], inplace=True)
+df.rename(columns={'tid_y':'tid', 'rawContent_y':'rawContent', 'date_y':'date'}, inplace=True)
+df = df[['tid','date','topicCovid','fake','rawContent','Unnamed: 6']]
+df.rename(columns={'Unnamed: 6':'comment'}, inplace=True)
+
+df.to_csv(correct, encoding='utf-8', sep=";")

trainFake.py

@@ -0,0 +1,613 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Sat Aug 12 12:25:18 2023
+
+@author: michael
+"""
+#from datasets import load_dataset
+#from transformers import Trainer
+#from transformers import AutoModelForSequenceClassification
+from transformers import AutoTokenizer
+import torch 
+import numpy as np
+from sklearn.model_selection import train_test_split # pip install scikit-learn
+
+import pandas as pd
+
+## Uses snippets from this guide:
+# https://mccormickml.com/2019/07/22/BERT-fine-tuning/
+
+###################
+# Setup directories
+# WD Michael
+wd = "/home/michael/Documents/PS/Data/collectTweets/"
+# WD Server
+# wd = '/home/yunohost.multimedia/polsoc/Politics & Society/TweetCollection/'
+
+import sys
+funs = wd+"funs"
+sys.path.insert(1, funs)
+import CleanTweets
+
+# datafile input directory
+di = "data/IN/"
+
+# Tweet-datafile output directory
+ud = "data/OUT/"
+
+# Training CSV dataset
+twtCSV = "SenatorsTweets-Training_WORKING-COPY-correct2"
+twtCSVtrainCovClass = "SenatorsTweets-train-CovClassification"
+twtCSVtrainFakeClass = "SenatorsTweets-train-FakeClassification"
+statsTrainingTopicClass = "statsTopicClassification-"
+
+# don't change this one
+twtCSVPath = wd + ud + twtCSV + ".csv"
+twtCSVtrainCovClassPath = wd + ud + twtCSVtrainCovClass + ".csv"
+twtCSVtrainFakeClassPath = wd + ud + twtCSVtrainFakeClass + ".csv"
+
+statsTrainingTopicClassPath = wd + ud + statsTrainingTopicClass
+
+twtCSVtrainCovClassPathTrain = wd + ud + twtCSVtrainCovClass + "TRAIN.csv"
+twtCSVtrainFakeClassPathTrain = wd + ud + twtCSVtrainFakeClass + "TRAIN.csv"
+twtTSVtrainCovClassPathTrain = wd + ud + "cov-train.tsv"
+twtTSVtrainFakeClassPathTrain = wd + ud + "fake-train.tsv"
+
+twtTSVtrainCovClassPathEval = wd + ud + "cov-eval.tsv" 
+twtTSVtrainFakeClassPathEval = wd + ud + "fake-eval.tsv"
+
+seed = 12355
+
+# Model paths
+modCovClassPath = wd + "models/CovClass/"
+modFakeClassPath = wd + "models/FakeClass/"
+
+model_name = 'digitalepidemiologylab/covid-twitter-bert-v2' # accuracy 69
+#model_name = 'justinqbui/bertweet-covid19-base-uncased-pretraining-covid-vaccine-tweets' #48
+#model_name = "cardiffnlp/tweet-topic-latest-multi"
+model_name = "bvrau/covid-twitter-bert-v2-struth"
+#model_name = "cardiffnlp/roberta-base-tweet-topic-single-all"
+model_fake_name = 'bvrau/covid-twitter-bert-v2-struth' 
+
+# More models for fake detection:
+# https://huggingface.co/justinqbui/bertweet-covid-vaccine-tweets-finetuned
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+max_length = 64 # max token sentence length
+
+#%%
+# Create training and testing dataset
+dfTest = pd.read_csv(twtCSVPath, dtype=(object), delimiter=";")
+
+#dfTest = dfTest[:-900] # remove last 800 rows
+#dfTest = dfTest.iloc[:,:-3] # remove last 800 rows
+
+dfTest['text'] = dfTest['rawContent'].apply(CleanTweets.preprocess_roberta)
+
+dfTest.drop(columns=['rawContent'], inplace=True)
+
+# Only keep tweets that are longer than 3 words
+dfTest['tweet_proc_length'] = [len(text.split(' ')) for text in dfTest['text']]
+dfTest['tweet_proc_length'].value_counts()
+dfTest = dfTest[dfTest['tweet_proc_length']>3]
+dfTest = dfTest.drop_duplicates(subset=['text'])
+dfTest = dfTest.drop(columns=['date', 'Unnamed: 0']) 
+
+# Create datasets for each classification
+dfCovClass = dfTest
+dfFakeClass = dfTest
+dfCovClass = dfCovClass.drop(columns=['fake']) # fake column not neeeded in covid topic classification data
+dfFakeClass = dfFakeClass[dfFakeClass['topicCovid']=='True'].drop(columns=['topicCovid']) # topicCovid column not neeeded in covid topic classification data
+
+#type_map = {'Covid tweet': 'covid tweets', 'Noncovid tweet': 'noncovid tweet'}
+dfCovClass.rename(index = str, columns={'topicCovid': 'labels', 'tid': 'id'}, inplace = True)
+dfCovClass.labels = dfCovClass.labels.replace({"True": 'Covid', "False": 'NonCovid'})
+
+#type_map = {'fake news tweet': 'fake news tweet', 'non-fake-news-tweet': 'non-fake-news-tweet'}
+dfFakeClass.rename(index = str, columns={'fake': 'labels', 'tid': 'id'}, inplace = True)
+
+#%%
+# Tokenize tweets
+dfCovClass = dfCovClass[dfCovClass['labels'].notna()]
+dfFakeClass['labels'].replace({'Check': '','check': '', 'FALSE':''}, inplace=True)
+dfFakeClass = dfFakeClass[dfFakeClass['labels'].notna()]
+dfCovClass['input_ids'] = dfCovClass['text'].apply(lambda x: tokenizer(x, max_length=max_length, padding="max_length",)['input_ids'])
+dfFakeClass['input_ids'] = dfFakeClass['text'].apply(lambda x: tokenizer(x, max_length=max_length, padding="max_length",)['input_ids'])
+
+def encode_labels(label):
+    if label == 'Covid':
+        return 1
+    elif label == 'NonCovid':
+        return 0
+    elif label == 'False':
+        return 1
+    elif label == 'True':
+        return 0
+    return 0
+dfCovClass['labels_encoded'] = dfCovClass['labels'].apply(encode_labels)
+dfFakeClass['labels_encoded'] = dfFakeClass['labels'].apply(encode_labels)
+dfFakeClass = dfFakeClass[dfFakeClass['labels']!=""]
+#dfFakeClass = dfFakeClass[(dfFakeClass['labels']=="Fake") | (dfFakeClass['labels']=="True")]
+
+# get n of classes
+print("# of Non-Covid tweets (coded 0):")
+print(dfCovClass.groupby('labels_encoded', group_keys=False)['id'].nunique())
+# 62 non-covid tweets, disproportionate sample for training has to be 124 tweets
+
+print("# of Fake-news tweets (coded 1):")
+print(dfFakeClass.groupby('labels_encoded', group_keys=False)['id'].nunique())
+
+# create disproportionate sample - 50/50 of both
+#dfCovClass.groupby('labels_encoded', group_keys=False)['id'].nunique()
+#dfCovClass = dfCovClass.groupby('labels_encoded', group_keys=False).apply(lambda x: x.sample(164, random_state=seed))
+# after a lot of tests, it seems that a sample in which non-fake news tweets are overrepresented leads to better results.
+# because of this, performance limitations and time constraints, group 1 (covid topic) will be overrepresented (twice as many), which still doesn't reflect the real preoportions ~10/1
+
+'''dfCovClassa = dfCovClass.groupby('labels_encoded', group_keys=False).get_group(1).sample(frac=1, replace=True).reset_index()
+dfCovClassb = dfCovClass.groupby('labels_encoded', group_keys=False).get_group(0).sample(frac=1, replace=True).reset_index()
+dfCovClassab= pd.concat([dfCovClassa,dfCovClassb]) 
+dfCovClassab.reset_index(inplace=True)
+dfCovClass_train, dfCovClass_test = train_test_split(dfCovClassab, test_size=0.1, random_state=seed, stratify=dfCovClassab['labels_encoded'])
+'''
+
+# create training and validation samples
+dfFakeClass_train, dfFakeClass_test = train_test_split(dfFakeClass, test_size=0.1, random_state=seed, stratify=dfFakeClass['labels_encoded'])
+
+# reset index and drop unnecessary columns
+dfFakeClass_train.reset_index(drop=True, inplace=True)
+dfFakeClass_train.drop(inplace=True, columns=['tweet_proc_length'])
+dfFakeClass_train.groupby('labels_encoded', group_keys=False)['id'].nunique()
+
+dfFakeClass_test.reset_index(drop=True, inplace=True)
+dfFakeClass_test.drop(inplace=True, columns=['tweet_proc_length'])
+dfFakeClass_test.groupby('labels_encoded', group_keys=False)['id'].nunique()
+
+# save dfs as csvs and tsvs, for training and validation
+# covid classification datafiles
+# rows 0-41 = noncovid, 42-81 covid, therfore:
+#dfCovClass = dfCovClass.drop(columns=['tweet_proc_length'])
+#dfCovClass.reset_index(inplace=True, drop=True)
+#dfCovClass.loc[np.r_[0:31, 42:71], :].reset_index(drop=True).to_csv(twtCSVtrainCovClassPathTrain, encoding='utf-8', sep=";") 
+#dfCovClass.loc[np.r_[0:31, 42:72], :].reset_index(drop=True).to_csv(twtTSVtrainCovClassPathTrain, encoding='utf-8', sep="\t")
+#dfCovClass.loc[np.r_[31:41, 72:81], :].reset_index(drop=True).to_csv(twtCSVtrainCovClassPath, encoding='utf-8', sep=";")
+#dfCovClass.loc[np.r_[31:41, 72:81], :].reset_index(drop=True).to_csv(twtTSVtrainCovClassPathEval, encoding='utf-8', sep="\t")
+
+# fake news classification datafiles
+#dfFakeClass = dfFakeClass.drop(columns=['tweet_proc_length'])
+#dfFakeClass[200:1000].reset_index(drop=True).to_csv(twtCSVtrainFakeClassPathTrain, encoding='utf-8', sep=";")
+#dfFakeClass[200:1000].reset_index(drop=True).to_csv(twtTSVtrainFakeClassPathTrain, encoding='utf-8', sep="\t")
+#dfFakeClass[0:199].reset_index(drop=True).to_csv(twtCSVtrainFakeClassPath, encoding='utf-8', sep=";")
+#dfFakeClass[0:199].reset_index(drop=True).to_csv(twtTSVtrainFakeClassPathEval, encoding='utf-8', sep="\t")
+
+#%%
+# Prepare trainer
+#from transformers import TrainingArguments
+
+#training_args = TrainingArguments(
+#     report_to = 'wandb',
+#    output_dir=wd+'results',          # output directory/
+#   overwrite_output_dir = True,
+#    num_train_epochs=6,              # total number of training epochs
+#    per_device_train_batch_size=8,  # batch size per device during training
+#    per_device_eval_batch_size=16,   # batch size for evaluation
+#    learning_rate=2e-5,
+#    warmup_steps=1000,                # number of warmup steps for learning rate scheduler
+#    weight_decay=0.01,               # strength of weight decay
+#    logging_dir='./logs3',            # directory for storing logs
+#    logging_steps=1000,
+#    evaluation_strategy="epoch",
+#    save_strategy="epoch",
+#    load_best_model_at_end=True
+#)
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+from transformers import BertForSequenceClassification, AdamW#, BertConfig
+#from torch.utils.data import TensorDataset, random_split
+from torch.utils.data import DataLoader, RandomSampler, SequentialSampler
+
+"""
+train_dataset = load_dataset('csv', data_files={'train': twtCSVtrainCovClassPathTrain}, encoding = "utf-8")
+train_dataset = train_dataset['train'] 
+eval_dataset = load_dataset('csv', data_files={'test': twtCSVtrainCovClassPath}, encoding = "utf-8")
+eval_dataset = eval_dataset['test'] 
+"""
+batch_size = 1
+
+from torch.utils.data import Dataset
+
+class PandasDataset(Dataset):
+    def __init__(self, dataframe, tokenizer, max_length):
+        self.dataframe = dataframe
+        self.tokenizer = tokenizer
+        self.max_length = max_length
+
+    def __len__(self):
+        return len(self.dataframe)
+
+    def __getitem__(self, index):
+        row = self.dataframe.iloc[index]
+        text = row['text']
+        labels = row['labels_encoded']
+        
+        encoded = self.tokenizer(text, max_length=self.max_length, padding="max_length", truncation=True)
+        input_ids = torch.tensor(encoded['input_ids'])
+        attention_mask = torch.tensor(encoded['attention_mask'])
+        
+        return {
+            'input_ids': input_ids,
+            'attention_mask': attention_mask,
+            'labels': torch.tensor(labels)  # Assuming labels are already encoded
+        }
+
+
+train_dataset = PandasDataset(dfFakeClass_train, tokenizer, max_length)
+train_dataloader = DataLoader(
+    train_dataset,
+    sampler=RandomSampler(train_dataset),
+    batch_size=batch_size
+)
+
+eval_dataset = PandasDataset(dfFakeClass_test, tokenizer, max_length)
+validation_dataloader = DataLoader(
+    eval_dataset,
+    sampler=SequentialSampler(eval_dataset),
+    batch_size=batch_size
+)
+
+for idx, batch in enumerate(train_dataloader):
+    print('Batch index: ', idx)
+    print('Batch size: ', batch['input_ids'].size())  # Access 'input_ids' field
+    print('Batch label: ', batch['labels'])           # Access 'labels' field
+    break
+
+model = BertForSequenceClassification.from_pretrained(
+    model_name,
+    num_labels = 2, # The number of output labels--2 for binary classification.
+                    # You can increase this for multi-class tasks.   
+    output_attentions = False, # Whether the model returns attentions weights.
+    output_hidden_states = False, # Whether the model returns all hidden-states.
+)
+
+#trainer = Trainer(
+#    model=model,                         # the instantiated 🤗 Transformers model to be trained
+#    args=training_args,                  # training arguments, defined above
+#    train_dataset=train_dataset,         # training dataset
+#    eval_dataset=eval_dataset             # evaluation dataset
+#)
+
+
+# Note: AdamW is a class from the huggingface library (as opposed to pytorch) 
+# I believe the 'W' stands for 'Weight Decay fix"
+optimizer = AdamW(model.parameters(),
+                  lr = 2e-5, # args.learning_rate - default is 5e-5, our notebook had 2e-5
+                  eps = 1e-8 # args.adam_epsilon  - default is 1e-8.
+                )
+
+from transformers import get_linear_schedule_with_warmup
+
+# Number of training epochs. The BERT authors recommend between 2 and 4. 
+# We chose to run for 6
+epochs = 6
+
+# Total number of training steps is [number of batches] x [number of epochs]. 
+# (Note that this is not the same as the number of training samples).
+total_steps = len(train_dataloader) * epochs
+
+# Create the learning rate scheduler.
+scheduler = get_linear_schedule_with_warmup(optimizer, 
+                                            num_warmup_steps = 0, # Default value in run_glue.py
+                                            num_training_steps = total_steps)
+
+# Function to calculate the accuracy of our predictions vs labels
+def flat_accuracy(preds, labels):
+    pred_flat = np.argmax(preds, axis=1).flatten()
+    labels_flat = labels.flatten()
+    return np.sum(pred_flat == labels_flat) / len(labels_flat)
+
+import time
+import datetime
+
+def format_time(elapsed):
+    '''
+    Takes a time in seconds and returns a string hh:mm:ss
+    '''
+    # Round to the nearest second.
+    elapsed_rounded = int(round((elapsed)))
+    
+    # Format as hh:mm:ss
+    return str(datetime.timedelta(seconds=elapsed_rounded))
+
+import random
+
+# This training code is based on the `run_glue.py` script here:
+# https://github.com/huggingface/transformers/blob/5bfcd0485ece086ebcbed2d008813037968a9e58/examples/run_glue.py#L128
+
+# Set the seed value all over the place to make this reproducible.
+seed_val = 12355
+
+# If there's a GPU available...
+if torch.cuda.is_available():    
+
+    # Tell PyTorch to use the GPU.    
+    device = torch.device("cuda")
+
+    print('There are %d GPU(s) available.' % torch.cuda.device_count())
+
+    print('We will use the GPU:', torch.cuda.get_device_name(0))
+    #model.cuda()
+# If not...
+else:
+    print('No GPU available, using the CPU instead.')
+    device = torch.device("cpu")
+
+device = torch.device("cpu")
+
+random.seed(seed_val)
+np.random.seed(seed_val)
+torch.manual_seed(seed_val)
+torch.cuda.manual_seed_all(seed_val)
+
+#%%
+# Start training
+# We'll store a number of quantities such as training and validation loss, 
+# validation accuracy, and timings.
+training_stats = []
+
+# Measure the total training time for the whole run.
+total_t0 = time.time()
+
+# For each epoch...
+for epoch_i in range(0, epochs):
+    # ========================================
+    #               Training
+    # ========================================
+    
+    # Perform one full pass over the training set.
+
+    print("")
+    print('======== Epoch {:} / {:} ========'.format(epoch_i + 1, epochs))
+    print('{:>5,} steps per batch will be calculated.'.format(len(train_dataloader)))
+    print('Training...')
+    
+    # Measure how long the training epoch takes.
+    t0 = time.time()
+    model.to(device)
+    # Reset the total loss for this epoch.
+    total_train_loss = 0
+    # Put the model into training mode. Don't be mislead--the call to 
+    # `train` just changes the *mode*, it doesn't *perform* the training.
+    # `dropout` and `batchnorm` layers behave differently during training
+    # vs. test (source: https://stackoverflow.com/questions/51433378/what-does-model-train-do-in-pytorch)
+    model.train()
+
+    # For each batch of training data...
+    for step, batch in enumerate(train_dataloader):
+
+        # Progress update every 10 batches.
+        if step % 10 == 0 and not step == 0:
+            # Calculate elapsed time in minutes.
+            elapsed = format_time(time.time() - t0)
+            
+            # Report progress.
+            print('  Batch {:>5,}  of  {:>5,}.    Elapsed: {:}.'.format(step, len(train_dataloader), elapsed))
+
+        # Unpack this training batch from our dataloader. 
+        #
+        # As we unpack the batch, we'll also copy each tensor to the GPU using the 
+        # `to` method.
+        #
+        # `batch` contains three pytorch tensors:
+        #   [0]: input ids 
+        #   [1]: attention masks
+        #   [2]: labels 
+        print("Batch keys:", batch.keys())
+        b_input_ids = batch['input_ids'].to(device)
+        b_input_mask = batch['attention_mask'].to(device)
+        b_labels = batch['labels'].to(device)
+
+        # Always clear any previously calculated gradients before performing a
+        # backward pass. PyTorch doesn't do this automatically because 
+        # accumulating the gradients is "convenient while training RNNs". 
+        # (source: https://stackoverflow.com/questions/48001598/why-do-we-need-to-call-zero-grad-in-pytorch)
+        model.zero_grad()        
+
+        # Perform a forward pass (evaluate the model on this training batch).
+        # The documentation for this `model` function is here: 
+        # https://huggingface.co/transformers/v2.2.0/model_doc/bert.html#transformers.BertForSequenceClassification
+        # It returns different numbers of parameters depending on what arguments
+        # arge given and what flags are set. For our useage here, it returns
+        # the loss (because we provided labels) and the "logits"--the model
+        # outputs prior to activation.
+        output = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)
+        loss = output[0]
+        logits = output[1]
+
+        # Accumulate the training loss over all of the batches so that we can
+        # calculate the average loss at the end. `loss` is a Tensor containing a
+        # single value; the `.item()` function just returns the Python value 
+        # from the tensor.
+        total_train_loss += loss.item()
+
+        # Perform a backward pass to calculate the gradients.
+        loss.backward()
+
+        # Clip the norm of the gradients to 1.0.
+        # This is to help prevent the "exploding gradients" problem.
+        torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
+
+        # Update parameters and take a step using the computed gradient.
+        # The optimizer dictates the "update rule"--how the parameters are
+        # modified based on their gradients, the learning rate, etc.
+        optimizer.step()
+
+        # Update the learning rate.
+        scheduler.step()
+
+    # Calculate the average loss over all of the batches.
+    avg_train_loss = total_train_loss / len(train_dataloader)            
+    
+    # Measure how long this epoch took.
+    training_time = format_time(time.time() - t0)
+
+    print("")
+    print("  Average training loss: {0:.2f}".format(avg_train_loss))
+    print("  Training epcoh took: {:}".format(training_time))
+        
+    # ========================================
+    #               Validation
+    # ========================================
+    # After the completion of each training epoch, measure our performance on
+    # our validation set.
+
+    print("")
+    print("Running Validation...")
+
+    t0 = time.time()
+
+    # Put the model in evaluation mode--the dropout layers behave differently
+    # during evaluation.
+    model.eval()
+
+    # Tracking variables 
+    total_eval_accuracy = 0
+    total_eval_loss = 0
+    nb_eval_steps = 0
+
+    # Evaluate data for one epoch
+    for batch in validation_dataloader:
+        
+        # Unpack this training batch from our dataloader. 
+        #
+        # As we unpack the batch, we'll also copy each tensor to the GPU using 
+        # the `to` method.
+        #
+        # `batch` contains three pytorch tensors:
+        #   [0]: input ids 
+        #   [1]: attention masks
+        #   [2]: labels 
+        b_input_ids = batch['input_ids'].to(device)
+        b_input_mask = batch['attention_mask'].to(device)
+        b_labels = batch['labels'].to(device)
+        
+        # Tell pytorch not to bother with constructing the compute graph during
+        # the forward pass, since this is only needed for backprop (training).
+        with torch.no_grad():        
+
+            # Forward pass, calculate logit predictions.
+            # token_type_ids is the same as the "segment ids", which 
+            # differentiates sentence 1 and 2 in 2-sentence tasks.
+            # The documentation for this `model` function is here: 
+            # https://huggingface.co/transformers/v2.2.0/model_doc/bert.html#transformers.BertForSequenceClassification
+            # Get the "logits" output by the model. The "logits" are the output
+            # values prior to applying an activation function like the softmax.
+            output = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)
+            loss = output[0]
+            logits = output[1]
+            
+        # Accumulate the validation loss.
+        total_eval_loss += loss.item()
+
+        # Move logits and labels to CPU
+        logits = logits.detach().cpu().numpy()
+        label_ids = b_labels.to('cpu').numpy()
+
+        # Calculate the accuracy for this batch of test sentences, and
+        # accumulate it over all batches.
+        total_eval_accuracy += flat_accuracy(logits, label_ids)
+        
+
+    # Report the final accuracy for this validation run.
+    avg_val_accuracy = total_eval_accuracy / len(validation_dataloader)
+    print("  Accuracy: {0:.2f}".format(avg_val_accuracy))
+
+    # Calculate the average loss over all of the batches.
+    avg_val_loss = total_eval_loss / len(validation_dataloader)
+    
+    # Measure how long the validation run took.
+    validation_time = format_time(time.time() - t0)
+    
+    print("  Validation Loss: {0:.2f}".format(avg_val_loss))
+    print("  Validation took: {:}".format(validation_time))
+
+    # Record all statistics from this epoch.
+    training_stats.append(
+        {
+            'epoch': epoch_i + 1,
+            'Training Loss': avg_train_loss,
+            'Valid. Loss': avg_val_loss,
+            'Valid. Accur.': avg_val_accuracy,
+            'Training Time': training_time,
+            'Validation Time': validation_time
+        }
+    )
+
+print("")
+print("Training complete!")
+
+print("Total training took {:} (h:mm:ss)".format(format_time(time.time()-total_t0)))
+
+params = list(model.named_parameters())
+
+print('The BERT model has {:} different named parameters.\n'.format(len(params)))
+
+print('==== Embedding Layer ====\n')
+
+for p in params[0:5]:
+    print("{:<55} {:>12}".format(p[0], str(tuple(p[1].size()))))
+
+print('\n==== First Transformer ====\n')
+
+for p in params[5:21]:
+    print("{:<55} {:>12}".format(p[0], str(tuple(p[1].size()))))
+
+print('\n==== Output Layer ====\n')
+
+for p in params[-4:]:
+    print("{:<55} {:>12}".format(p[0], str(tuple(p[1].size()))))
+
+
+import os
+
+# Saving best-practices: if you use defaults names for the model, you can reload it using from_pretrained()
+from datetime import datetime as dt
+
+fTimeFormat = "%Y-%m-%d_%H-%M-%S"
+now = dt.now().strftime(fTimeFormat)
+
+output_dir = modFakeClassPath + now + "/"
+
+# Create output directory if needed
+if not os.path.exists(output_dir):
+    os.makedirs(output_dir)
+
+print("Saving model to %s" % output_dir)
+
+# Save a trained model, configuration and tokenizer using `save_pretrained()`.
+# They can then be reloaded using `from_pretrained()`
+model_to_save = model.module if hasattr(model, 'module') else model  # Take care of distributed/parallel training
+model_to_save.save_pretrained(output_dir)
+tokenizer.save_pretrained(output_dir)
+
+# Good practice: save your training arguments together with the trained model
+# torch.save(args, os.path.join(output_dir, 'training_args.bin'))
+
+import pandas as pd
+
+# Display floats with two decimal places.
+pd.set_option('display.precision', 2)
+
+# Create a DataFrame from our training statistics.
+df_stats = pd.DataFrame(data=training_stats)
+
+# Use the 'epoch' as the row index.# Good practice: save your training arguments together with the trained model
+df_stats = df_stats.set_index('epoch')
+
+# A hack to force the column headers to wrap.
+#df = df.style.set_table_styles([dict(selector="th",props=[('max-width', '70px')])])
+
+
+# Display the table.
+df_stats
+df_stats.to_csv(output_dir + now + ".csv")

trainTopic.py

@@ -0,0 +1,607 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Sat Aug 12 12:25:18 2023
+
+@author: michael
+"""
+#from datasets import load_dataset
+#from transformers import Trainer
+#from transformers import AutoModelForSequenceClassification
+from transformers import AutoTokenizer
+import torch 
+import numpy as np
+from sklearn.model_selection import train_test_split # pip install scikit-learn
+
+import pandas as pd
+
+## Uses snippets from this guide:
+# https://mccormickml.com/2019/07/22/BERT-fine-tuning/
+
+###################
+# Setup directories
+# WD Michael
+wd = "/home/michael/Documents/PS/Data/collectTweets/"
+# WD Server
+# wd = '/home/yunohost.multimedia/polsoc/Politics & Society/TweetCollection/'
+
+import sys
+funs = wd+"funs"
+sys.path.insert(1, funs)
+import CleanTweets
+
+# datafile input directory
+di = "data/IN/"
+
+# Tweet-datafile output directory
+ud = "data/OUT/"
+
+# Training CSV dataset
+twtCSV = "SenatorsTweets-Training_WORKING-COPY-correct2"
+twtCSVtrainCovClass = "SenatorsTweets-train-CovClassification"
+twtCSVtrainFakeClass = "SenatorsTweets-train-FakeClassification"
+statsTrainingTopicClass = "statsTopicClassification-"
+
+# don't change this one
+twtCSVPath = wd + ud + twtCSV + ".csv"
+twtCSVtrainCovClassPath = wd + ud + twtCSVtrainCovClass + ".csv"
+twtCSVtrainFakeClassPath = wd + ud + twtCSVtrainFakeClass + ".csv"
+
+statsTrainingTopicClassPath = wd + ud + statsTrainingTopicClass
+
+twtCSVtrainCovClassPathTrain = wd + ud + twtCSVtrainCovClass + "TRAIN.csv"
+twtCSVtrainFakeClassPathTrain = wd + ud + twtCSVtrainFakeClass + "TRAIN.csv"
+twtTSVtrainCovClassPathTrain = wd + ud + "cov-train.tsv"
+twtTSVtrainFakeClassPathTrain = wd + ud + "fake-train.tsv"
+
+twtTSVtrainCovClassPathEval = wd + ud + "cov-eval.tsv" 
+twtTSVtrainFakeClassPathEval = wd + ud + "fake-eval.tsv"
+
+seed = 12355
+
+# Model paths
+modCovClassPath = wd + "models/CovClass/"
+modFakeClassPath = wd + "models/FakeClass/"
+
+model_name = "bvrau/covid-twitter-bert-v2-struth"
+model_fake_name = 'bvrau/covid-twitter-bert-v2-struth' 
+
+# More models for fake detection:
+# https://huggingface.co/justinqbui/bertweet-covid-vaccine-tweets-finetuned
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+max_length = 64 # max token sentence length
+
+#%%
+# Create training and testing dataset
+dfTest = pd.read_csv(twtCSVPath, dtype=(object), delimiter=";")
+
+#dfTest = dfTest[:-900] # remove last 800 rows
+#dfTest = dfTest.iloc[:,:-3] # remove last 800 rows
+
+dfTest['text'] = dfTest['rawContent'].apply(CleanTweets.preprocess_roberta)
+
+dfTest.drop(columns=['rawContent'], inplace=True)
+
+# Only keep tweets that are longer than 3 words
+dfTest['tweet_proc_length'] = [len(text.split(' ')) for text in dfTest['text']]
+dfTest['tweet_proc_length'].value_counts()
+dfTest = dfTest[dfTest['tweet_proc_length']>3]
+dfTest = dfTest.drop_duplicates(subset=['text'])
+dfTest = dfTest.drop(columns=['date', 'Unnamed: 0']) 
+
+# Create datasets for each classification
+dfCovClass = dfTest
+dfFakeClass = dfTest
+dfCovClass = dfCovClass.drop(columns=['fake']) # fake column not neeeded in covid topic classification data
+dfFakeClass = dfFakeClass[dfFakeClass['topicCovid']=='True'].drop(columns=['topicCovid']) # topicCovid column not neeeded in covid topic classification data
+
+#type_map = {'Covid tweet': 'covid tweets', 'Noncovid tweet': 'noncovid tweet'}
+dfCovClass.rename(index = str, columns={'topicCovid': 'labels', 'tid': 'id'}, inplace = True)
+dfCovClass.labels = dfCovClass.labels.replace({"True": 'Covid', "False": 'NonCovid'})
+
+#type_map = {'fake news tweet': 'fake news tweet', 'non-fake-news-tweet': 'non-fake-news-tweet'}
+dfFakeClass.rename(index = str, columns={'fake': 'labels', 'tid': 'id'}, inplace = True)
+dfFakeClass.labels = dfFakeClass.labels.replace({"True": 'Fake', "False": 'True'})
+
+#%%
+# Tokenize tweets
+dfCovClass = dfCovClass[dfCovClass['labels'].notna()]
+dfFakeClass = dfFakeClass[dfFakeClass['labels'].notna()]
+dfCovClass['input_ids'] = dfCovClass['text'].apply(lambda x: tokenizer(x, max_length=max_length, padding="max_length",)['input_ids'])
+dfFakeClass['input_ids'] = dfFakeClass['text'].apply(lambda x: tokenizer(x, max_length=max_length, padding="max_length",)['input_ids'])
+
+def encode_labels(label):
+    if label == 'Covid':
+        return 1
+    elif label == 'NonCovid':
+        return 0
+    elif label == 'Fake':
+        return 1
+    elif label == 'True':
+        return 0
+    return 0
+dfCovClass['labels_encoded'] = dfCovClass['labels'].apply(encode_labels)
+dfFakeClass['labels_encoded'] = dfFakeClass['labels'].apply(encode_labels)
+
+# get n of classes
+print("# of Non-Covid tweets (coded 0):")
+print(dfCovClass.groupby('labels_encoded', group_keys=False)['id'].nunique())
+# 62 non-covid tweets, disproportionate sample for training has to be 124 tweets
+
+print("# of Fake-news tweets (coded 1):")
+print(dfFakeClass.groupby('labels_encoded', group_keys=False)['id'].nunique())
+
+# create disproportionate sample - 50/50 of both
+#dfCovClass.groupby('labels_encoded', group_keys=False)['id'].nunique()
+#dfCovClass = dfCovClass.groupby('labels_encoded', group_keys=False).apply(lambda x: x.sample(164, random_state=seed))
+# after a lot of tests, it seems that a sample in which non-fake news tweets are overrepresented leads to better results.
+# because of this, performance limitations and time constraints, group 1 (covid topic) will be overrepresented (twice as many), which still doesn't reflect the real preoportions ~10/1
+
+'''dfCovClassa = dfCovClass.groupby('labels_encoded', group_keys=False).get_group(1).sample(frac=1, replace=True).reset_index()
+dfCovClassb = dfCovClass.groupby('labels_encoded', group_keys=False).get_group(0).sample(frac=1, replace=True).reset_index()
+dfCovClassab= pd.concat([dfCovClassa,dfCovClassb]) 
+dfCovClassab.reset_index(inplace=True)
+dfCovClass_train, dfCovClass_test = train_test_split(dfCovClassab, test_size=0.1, random_state=seed, stratify=dfCovClassab['labels_encoded'])
+'''
+
+# create training and validation samples
+dfCovClass_train, dfCovClass_test = train_test_split(dfCovClass, test_size=0.1, random_state=seed, stratify=dfCovClass['labels_encoded'])
+
+# reset index and drop unnecessary columns
+dfCovClass_train.reset_index(drop=True, inplace=True)
+dfCovClass_train.drop(inplace=True, columns=['tweet_proc_length'])
+dfCovClass_train.groupby('labels_encoded', group_keys=False)['id'].nunique()
+
+dfCovClass_test.reset_index(drop=True, inplace=True)
+dfCovClass_test.drop(inplace=True, columns=['tweet_proc_length'])
+dfCovClass_test.groupby('labels_encoded', group_keys=False)['id'].nunique()
+
+# save dfs as csvs and tsvs, for training and validation
+# covid classification datafiles
+# rows 0-41 = noncovid, 42-81 covid, therfore:
+#dfCovClass = dfCovClass.drop(columns=['tweet_proc_length'])
+#dfCovClass.reset_index(inplace=True, drop=True)
+#dfCovClass.loc[np.r_[0:31, 42:71], :].reset_index(drop=True).to_csv(twtCSVtrainCovClassPathTrain, encoding='utf-8', sep=";") 
+#dfCovClass.loc[np.r_[0:31, 42:72], :].reset_index(drop=True).to_csv(twtTSVtrainCovClassPathTrain, encoding='utf-8', sep="\t")
+#dfCovClass.loc[np.r_[31:41, 72:81], :].reset_index(drop=True).to_csv(twtCSVtrainCovClassPath, encoding='utf-8', sep=";")
+#dfCovClass.loc[np.r_[31:41, 72:81], :].reset_index(drop=True).to_csv(twtTSVtrainCovClassPathEval, encoding='utf-8', sep="\t")
+
+# fake news classification datafiles
+#dfFakeClass = dfFakeClass.drop(columns=['tweet_proc_length'])
+#dfFakeClass[200:1000].reset_index(drop=True).to_csv(twtCSVtrainFakeClassPathTrain, encoding='utf-8', sep=";")
+#dfFakeClass[200:1000].reset_index(drop=True).to_csv(twtTSVtrainFakeClassPathTrain, encoding='utf-8', sep="\t")
+#dfFakeClass[0:199].reset_index(drop=True).to_csv(twtCSVtrainFakeClassPath, encoding='utf-8', sep=";")
+#dfFakeClass[0:199].reset_index(drop=True).to_csv(twtTSVtrainFakeClassPathEval, encoding='utf-8', sep="\t")
+
+#%%
+# Prepare trainer
+#from transformers import TrainingArguments
+
+#training_args = TrainingArguments(
+#     report_to = 'wandb',
+#    output_dir=wd+'results',          # output directory/
+#   overwrite_output_dir = True,
+#    num_train_epochs=6,              # total number of training epochs
+#    per_device_train_batch_size=8,  # batch size per device during training
+#    per_device_eval_batch_size=16,   # batch size for evaluation
+#    learning_rate=2e-5,
+#    warmup_steps=1000,                # number of warmup steps for learning rate scheduler
+#    weight_decay=0.01,               # strength of weight decay
+#    logging_dir='./logs3',            # directory for storing logs
+#    logging_steps=1000,
+#    evaluation_strategy="epoch",
+#    save_strategy="epoch",
+#    load_best_model_at_end=True
+#)
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+from transformers import BertForSequenceClassification, AdamW#, BertConfig
+#from torch.utils.data import TensorDataset, random_split
+from torch.utils.data import DataLoader, RandomSampler, SequentialSampler
+
+"""
+train_dataset = load_dataset('csv', data_files={'train': twtCSVtrainCovClassPathTrain}, encoding = "utf-8")
+train_dataset = train_dataset['train'] 
+eval_dataset = load_dataset('csv', data_files={'test': twtCSVtrainCovClassPath}, encoding = "utf-8")
+eval_dataset = eval_dataset['test'] 
+"""
+batch_size = 1
+
+from torch.utils.data import Dataset
+
+class PandasDataset(Dataset):
+    def __init__(self, dataframe, tokenizer, max_length):
+        self.dataframe = dataframe
+        self.tokenizer = tokenizer
+        self.max_length = max_length
+
+    def __len__(self):
+        return len(self.dataframe)
+
+    def __getitem__(self, index):
+        row = self.dataframe.iloc[index]
+        text = row['text']
+        labels = row['labels_encoded']
+        
+        encoded = self.tokenizer(text, max_length=self.max_length, padding="max_length", truncation=True)
+        input_ids = torch.tensor(encoded['input_ids'])
+        attention_mask = torch.tensor(encoded['attention_mask'])
+        
+        return {
+            'input_ids': input_ids,
+            'attention_mask': attention_mask,
+            'labels': torch.tensor(labels)  # Assuming labels are already encoded
+        }
+
+
+train_dataset = PandasDataset(dfCovClass_train, tokenizer, max_length)
+train_dataloader = DataLoader(
+    train_dataset,
+    sampler=RandomSampler(train_dataset),
+    batch_size=batch_size
+)
+
+eval_dataset = PandasDataset(dfCovClass_test, tokenizer, max_length)
+validation_dataloader = DataLoader(
+    eval_dataset,
+    sampler=SequentialSampler(eval_dataset),
+    batch_size=batch_size
+)
+
+for idx, batch in enumerate(train_dataloader):
+    print('Batch index: ', idx)
+    print('Batch size: ', batch['input_ids'].size())  # Access 'input_ids' field
+    print('Batch label: ', batch['labels'])           # Access 'labels' field
+    break
+
+model = BertForSequenceClassification.from_pretrained(
+    model_name,
+    num_labels = 2, # The number of output labels--2 for binary classification.
+                    # You can increase this for multi-class tasks.   
+    output_attentions = False, # Whether the model returns attentions weights.
+    output_hidden_states = False, # Whether the model returns all hidden-states.
+)
+
+#trainer = Trainer(
+#    model=model,                         # the instantiated 🤗 Transformers model to be trained
+#    args=training_args,                  # training arguments, defined above
+#    train_dataset=train_dataset,         # training dataset
+#    eval_dataset=eval_dataset             # evaluation dataset
+#)
+
+
+# Note: AdamW is a class from the huggingface library (as opposed to pytorch) 
+# I believe the 'W' stands for 'Weight Decay fix"
+optimizer = AdamW(model.parameters(),
+                  lr = 2e-5, # args.learning_rate - default is 5e-5, our notebook had 2e-5
+                  eps = 1e-8 # args.adam_epsilon  - default is 1e-8.
+                )
+
+from transformers import get_linear_schedule_with_warmup
+
+# Number of training epochs. The BERT authors recommend between 2 and 4. 
+# We chose to run for 6
+epochs = 6
+
+# Total number of training steps is [number of batches] x [number of epochs]. 
+# (Note that this is not the same as the number of training samples).
+total_steps = len(train_dataloader) * epochs
+
+# Create the learning rate scheduler.
+scheduler = get_linear_schedule_with_warmup(optimizer, 
+                                            num_warmup_steps = 0, # Default value in run_glue.py
+                                            num_training_steps = total_steps)
+
+# Function to calculate the accuracy of our predictions vs labels
+def flat_accuracy(preds, labels):
+    pred_flat = np.argmax(preds, axis=1).flatten()
+    labels_flat = labels.flatten()
+    return np.sum(pred_flat == labels_flat) / len(labels_flat)
+
+import time
+import datetime
+
+def format_time(elapsed):
+    '''
+    Takes a time in seconds and returns a string hh:mm:ss
+    '''
+    # Round to the nearest second.
+    elapsed_rounded = int(round((elapsed)))
+    
+    # Format as hh:mm:ss
+    return str(datetime.timedelta(seconds=elapsed_rounded))
+
+import random
+
+# This training code is based on the `run_glue.py` script here:
+# https://github.com/huggingface/transformers/blob/5bfcd0485ece086ebcbed2d008813037968a9e58/examples/run_glue.py#L128
+
+# Set the seed value all over the place to make this reproducible.
+seed_val = 12355
+
+# If there's a GPU available...
+if torch.cuda.is_available():    
+
+    # Tell PyTorch to use the GPU.    
+    device = torch.device("cuda")
+
+    print('There are %d GPU(s) available.' % torch.cuda.device_count())
+
+    print('We will use the GPU:', torch.cuda.get_device_name(0))
+    #model.cuda()
+# If not...
+else:
+    print('No GPU available, using the CPU instead.')
+    device = torch.device("cpu")
+
+device = torch.device("cpu")
+
+random.seed(seed_val)
+np.random.seed(seed_val)
+torch.manual_seed(seed_val)
+torch.cuda.manual_seed_all(seed_val)
+
+#%%
+# Start training
+# We'll store a number of quantities such as training and validation loss, 
+# validation accuracy, and timings.
+training_stats = []
+
+# Measure the total training time for the whole run.
+total_t0 = time.time()
+
+# For each epoch...
+for epoch_i in range(0, epochs):
+    # ========================================
+    #               Training
+    # ========================================
+    
+    # Perform one full pass over the training set.
+
+    print("")
+    print('======== Epoch {:} / {:} ========'.format(epoch_i + 1, epochs))
+    print('{:>5,} steps per batch will be calculated.'.format(len(train_dataloader)))
+    print('Training...')
+    
+    # Measure how long the training epoch takes.
+    t0 = time.time()
+    model.to(device)
+    # Reset the total loss for this epoch.
+    total_train_loss = 0
+    # Put the model into training mode. Don't be mislead--the call to 
+    # `train` just changes the *mode*, it doesn't *perform* the training.
+    # `dropout` and `batchnorm` layers behave differently during training
+    # vs. test (source: https://stackoverflow.com/questions/51433378/what-does-model-train-do-in-pytorch)
+    model.train()
+
+    # For each batch of training data...
+    for step, batch in enumerate(train_dataloader):
+
+        # Progress update every 10 batches.
+        if step % 10 == 0 and not step == 0:
+            # Calculate elapsed time in minutes.
+            elapsed = format_time(time.time() - t0)
+            
+            # Report progress.
+            print('  Batch {:>5,}  of  {:>5,}.    Elapsed: {:}.'.format(step, len(train_dataloader), elapsed))
+
+        # Unpack this training batch from our dataloader. 
+        #
+        # As we unpack the batch, we'll also copy each tensor to the GPU using the 
+        # `to` method.
+        #
+        # `batch` contains three pytorch tensors:
+        #   [0]: input ids 
+        #   [1]: attention masks
+        #   [2]: labels 
+        print("Batch keys:", batch.keys())
+        b_input_ids = batch['input_ids'].to(device)
+        b_input_mask = batch['attention_mask'].to(device)
+        b_labels = batch['labels'].to(device)
+
+        # Always clear any previously calculated gradients before performing a
+        # backward pass. PyTorch doesn't do this automatically because 
+        # accumulating the gradients is "convenient while training RNNs". 
+        # (source: https://stackoverflow.com/questions/48001598/why-do-we-need-to-call-zero-grad-in-pytorch)
+        model.zero_grad()        
+
+        # Perform a forward pass (evaluate the model on this training batch).
+        # The documentation for this `model` function is here: 
+        # https://huggingface.co/transformers/v2.2.0/model_doc/bert.html#transformers.BertForSequenceClassification
+        # It returns different numbers of parameters depending on what arguments
+        # arge given and what flags are set. For our useage here, it returns
+        # the loss (because we provided labels) and the "logits"--the model
+        # outputs prior to activation.
+        output = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)
+        loss = output[0]
+        logits = output[1]
+
+        # Accumulate the training loss over all of the batches so that we can
+        # calculate the average loss at the end. `loss` is a Tensor containing a
+        # single value; the `.item()` function just returns the Python value 
+        # from the tensor.
+        total_train_loss += loss.item()
+
+        # Perform a backward pass to calculate the gradients.
+        loss.backward()
+
+        # Clip the norm of the gradients to 1.0.
+        # This is to help prevent the "exploding gradients" problem.
+        torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
+
+        # Update parameters and take a step using the computed gradient.
+        # The optimizer dictates the "update rule"--how the parameters are
+        # modified based on their gradients, the learning rate, etc.
+        optimizer.step()
+
+        # Update the learning rate.
+        scheduler.step()
+
+    # Calculate the average loss over all of the batches.
+    avg_train_loss = total_train_loss / len(train_dataloader)            
+    
+    # Measure how long this epoch took.
+    training_time = format_time(time.time() - t0)
+
+    print("")
+    print("  Average training loss: {0:.2f}".format(avg_train_loss))
+    print("  Training epcoh took: {:}".format(training_time))
+        
+    # ========================================
+    #               Validation
+    # ========================================
+    # After the completion of each training epoch, measure our performance on
+    # our validation set.
+
+    print("")
+    print("Running Validation...")
+
+    t0 = time.time()
+
+    # Put the model in evaluation mode--the dropout layers behave differently
+    # during evaluation.
+    model.eval()
+
+    # Tracking variables 
+    total_eval_accuracy = 0
+    total_eval_loss = 0
+    nb_eval_steps = 0
+
+    # Evaluate data for one epoch
+    for batch in validation_dataloader:
+        
+        # Unpack this training batch from our dataloader. 
+        #
+        # As we unpack the batch, we'll also copy each tensor to the GPU using 
+        # the `to` method.
+        #
+        # `batch` contains three pytorch tensors:
+        #   [0]: input ids 
+        #   [1]: attention masks
+        #   [2]: labels 
+        b_input_ids = batch['input_ids'].to(device)
+        b_input_mask = batch['attention_mask'].to(device)
+        b_labels = batch['labels'].to(device)
+        
+        # Tell pytorch not to bother with constructing the compute graph during
+        # the forward pass, since this is only needed for backprop (training).
+        with torch.no_grad():        
+
+            # Forward pass, calculate logit predictions.
+            # token_type_ids is the same as the "segment ids", which 
+            # differentiates sentence 1 and 2 in 2-sentence tasks.
+            # The documentation for this `model` function is here: 
+            # https://huggingface.co/transformers/v2.2.0/model_doc/bert.html#transformers.BertForSequenceClassification
+            # Get the "logits" output by the model. The "logits" are the output
+            # values prior to applying an activation function like the softmax.
+            output = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)
+            loss = output[0]
+            logits = output[1]
+            
+        # Accumulate the validation loss.
+        total_eval_loss += loss.item()
+
+        # Move logits and labels to CPU
+        logits = logits.detach().cpu().numpy()
+        label_ids = b_labels.to('cpu').numpy()
+
+        # Calculate the accuracy for this batch of test sentences, and
+        # accumulate it over all batches.
+        total_eval_accuracy += flat_accuracy(logits, label_ids)
+        
+
+    # Report the final accuracy for this validation run.
+    avg_val_accuracy = total_eval_accuracy / len(validation_dataloader)
+    print("  Accuracy: {0:.2f}".format(avg_val_accuracy))
+
+    # Calculate the average loss over all of the batches.
+    avg_val_loss = total_eval_loss / len(validation_dataloader)
+    
+    # Measure how long the validation run took.
+    validation_time = format_time(time.time() - t0)
+    
+    print("  Validation Loss: {0:.2f}".format(avg_val_loss))
+    print("  Validation took: {:}".format(validation_time))
+
+    # Record all statistics from this epoch.
+    training_stats.append(
+        {
+            'epoch': epoch_i + 1,
+            'Training Loss': avg_train_loss,
+            'Valid. Loss': avg_val_loss,
+            'Valid. Accur.': avg_val_accuracy,
+            'Training Time': training_time,
+            'Validation Time': validation_time
+        }
+    )
+
+print("")
+print("Training complete!")
+
+print("Total training took {:} (h:mm:ss)".format(format_time(time.time()-total_t0)))
+
+params = list(model.named_parameters())
+
+print('The BERT model has {:} different named parameters.\n'.format(len(params)))
+
+print('==== Embedding Layer ====\n')
+
+for p in params[0:5]:
+    print("{:<55} {:>12}".format(p[0], str(tuple(p[1].size()))))
+
+print('\n==== First Transformer ====\n')
+
+for p in params[5:21]:
+    print("{:<55} {:>12}".format(p[0], str(tuple(p[1].size()))))
+
+print('\n==== Output Layer ====\n')
+
+for p in params[-4:]:
+    print("{:<55} {:>12}".format(p[0], str(tuple(p[1].size()))))
+
+
+import os
+
+# Saving best-practices: if you use defaults names for the model, you can reload it using from_pretrained()
+from datetime import datetime as dt
+
+fTimeFormat = "%Y-%m-%d_%H-%M-%S"
+now = dt.now().strftime(fTimeFormat)
+
+output_dir = modCovClassPath + now + "/"
+
+# Create output directory if needed
+if not os.path.exists(output_dir):
+    os.makedirs(output_dir)
+
+print("Saving model to %s" % output_dir)
+
+# Save a trained model, configuration and tokenizer using `save_pretrained()`.
+# They can then be reloaded using `from_pretrained()`
+model_to_save = model.module if hasattr(model, 'module') else model  # Take care of distributed/parallel training
+model_to_save.save_pretrained(output_dir)
+tokenizer.save_pretrained(output_dir)
+
+# Good practice: save your training arguments together with the trained model
+# torch.save(args, os.path.join(output_dir, 'training_args.bin'))
+
+import pandas as pd
+
+# Display floats with two decimal places.
+pd.set_option('display.precision', 2)
+
+# Create a DataFrame from our training statistics.
+df_stats = pd.DataFrame(data=training_stats)
+
+# Use the 'epoch' as the row index.# Good practice: save your training arguments together with the trained model
+df_stats = df_stats.set_index('epoch')
+
+# A hack to force the column headers to wrap.
+#df = df.style.set_table_styles([dict(selector="th",props=[('max-width', '70px')])])
+
+
+# Display the table.
+df_stats
+df_stats.to_csv(output_dir + now + ".csv")