classifier demo

.gitignore

@@ -0,0 +1,134 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+data
+artifacts/
+wandb/
+results

README.md

@@ -1,11 +1,11 @@
 ---
-title: UnreliableNews
-emoji: 📉
-colorFrom: yellow
+title: Unreliable News Classifier
+emoji: 📰
+colorFrom: red
 colorTo: pink
 sdk: streamlit
 app_file: app.py
-pinned: false
+pinned: true
 ---
 
 # Configuration

app.py

@@ -0,0 +1,121 @@
+import os
+import json
+import numpy as np
+import pandas as pd
+
+import streamlit as st
+import torch
+import torch.nn.functional as F
+from transformers import DistilBertTokenizer, DistilBertForSequenceClassification
+
+@st.cache(allow_output_mutation=True)
+def init_model():
+    tokenizer = DistilBertTokenizer.from_pretrained('distilbert-base-cased')
+    model = DistilBertForSequenceClassification.from_pretrained('khizon/distilbert-unreliable-news-eng-4L', num_labels = 2)
+
+    return tokenizer, model
+
+def download_dataset():
+    url = 'https://drive.google.com/drive/folders/11mRvsHAkggFEJvG4axH4mmWI6FHMQp7X?usp=sharing'
+    data = 'data/nela_gt_2018_site_split'
+    
+    os.system(f'gdown --folder {url} -O {data}')
+
+@st.cache(allow_output_mutation=True)
+def jsonl_to_df(file_path):
+    with open(file_path) as f:
+        lines = f.read().splitlines()
+
+    df_inter = pd.DataFrame(lines)
+    df_inter.columns = ['json_element']
+
+    df_inter['json_element'].apply(json.loads)
+
+    return pd.json_normalize(df_inter['json_element'].apply(json.loads))
+
+@st.cache
+def load_test_df():
+    file_path = os.path.join('data', 'nela_gt_2018_site_split', 'test.jsonl')
+    test_df = jsonl_to_df(file_path)
+    test_df = pd.get_dummies(test_df, columns = ['label'])
+    return test_df
+
+@st.cache(allow_output_mutation=True)
+def predict(model, tokenizer, data):
+  
+    labels = data[['label_0', 'label_1']]
+    labels = torch.tensor(labels, dtype=torch.float32)
+    encoding = tokenizer.encode_plus(
+                data['title'],
+                ' [SEP] ' + data['content'],
+                add_special_tokens=True,
+                max_length = 512,
+                return_token_type_ids = False,
+                padding = 'max_length',
+                truncation = 'only_second',
+                return_attention_mask = True,
+                return_tensors = 'pt'
+            )
+
+    output = model(**encoding)
+    return correct_preds(output['logits'], labels)
+
+@st.cache(allow_output_mutation=True)
+def predict_new(model, tokenizer, title, content):
+    encoding = tokenizer.encode_plus(
+                title,
+                ' [SEP] ' + content,
+                add_special_tokens=True,
+                max_length = 512,
+                return_token_type_ids = False,
+                padding = 'max_length',
+                truncation = 'only_second',
+                return_attention_mask = True,
+                return_tensors = 'pt'
+            )
+    output = model(**encoding)
+    preds = F.softmax(output['logits'], dim = 1)
+    p_idx = torch.argmax(preds, dim = 1)
+    return 'reliable' if p_idx > 0 else 'unreliable'
+
+def correct_preds(preds, labels):
+    preds = torch.nn.functional.softmax(preds, dim = 1)
+    p_idx = torch.argmax(preds, dim=1)
+    l_idx = torch.argmax(labels, dim=0)
+
+    pred_label = 'reliable' if p_idx > 0 else 'unreliable'
+    correct = True if (p_idx == l_idx).sum().item() > 0 else False
+    return  pred_label, correct
+
+
+if __name__ == '__main__':
+    if not os.path.exists('data/nela_gt_2018_site_split/test.jsonl'):
+        download_dataset()
+    df = load_test_df()
+    tokenizer, model = init_model()
+
+    st.title("Unreliable News classifier")
+    mode = st.radio(
+        '', ('Test article', 'Input own article')
+    )
+    if mode == 'Test article':
+        if st.button('Get random article'):
+            idx = np.random.randint(0, len(df))
+            sample = df.iloc[idx]
+            
+            prediction, correct = predict(model, tokenizer, sample)
+            label = 'reliable' if sample['label_1'] > sample['label_0'] else 'unreliable'
+            st.header(sample['title'])
+            if correct:
+                st.success(f'Prediction: {prediction}')
+            else:
+                st.error(f'Prediction: {prediction}')
+            st.caption(f'Source: {sample["source"]} ({label})')
+            st.markdown(sample['content'])
+    else:
+        title = st.text_input('Article title', 'Test title')
+        content = st.text_area('Article content', 'Lorem ipsum')
+        if st.button('Submit'):
+            pred = predict_new(model, tokenizer, title, content)
+            st.markdown(f'Prediction: {pred}')
+            # st.success('success')

requirements.txt

@@ -0,0 +1,6 @@
+-f https://download.pytorch.org/whl/cu113/torch_stable.html
+gdown==4.2.0
+numpy==1.21.4
+pandas==1.3.4
+torch==1.10.1
+transformers==4.13.0