Mythbusters

Mythbusters is a webapp built on Anvil for detecting real or fake news from a given text input.

Visit the live app here: MythBusters

Installation/Dependencies

Dependencies:

pandas sklearn anvil-uplink

Installation

pip install pandas

pip install sklearn

We also used kaggle to find datasets used for training the ML model. You can find the link to the libraries here.

Step 1: Setting up and training the model

df = pd.read_csv("Tests/train.csv")
conversion_dict = {0: 'Real', 1: 'Fake'}
df['label']=df['label'].replace(conversion_dict)
df.label.value_counts()

x_train,x_test,y_train,y_test=train_test_split(df['text'], df['label'], test_size=0.25, random_state=7, shuffle=True)
tfidf_vectorizer=TfidfVectorizer(stop_words='english',max_df=0.75)

vec_train=tfidf_vectorizer.fit_transform(x_train.values.astype('U'))
vec_test=tfidf_vectorizer.transform(x_test.values.astype('U'))

pac=PassiveAggressiveClassifier(max_iter=100)
pac.fit(vec_train,y_train)

y_pred=pac.predict(vec_test)
score=accuracy_score(y_test,y_pred)
print(f'PAC Accuracy: {round(score*100,2)}%')

We used a Passive Aggressive Classifier to train the model. PAC Accuracy: 96.25%

Step 2: linking to Anvil.app

Install anvil-uplink and import and link using the uplink key obtained in the anvil.app website

pip install anvil-uplink

Now we can link it to the app using our key and use a function to implement our model.

import anvil.server
anvil.server.connect("Insert Key Here")

@anvil.server.callable
def findlabel(newtext):
    vec_newtest=tfidf_vectorizer.transform([newtext])
    y_predl=pac.predict(vec_newtest)
    return y_predl[0]