NLP_Emotion_Classifier_DistilBERT_KaggleData

January 26, 2025

[ ]: !kaggle datasets download -d nelgiriyewithana/emotions

Dataset URL: https://www.kaggle.com/datasets/nelgiriyewithana/emotions

License(s): other
Downloading emotions.zip to /content
89% 14.0M/15.7M [00:01<00:00, 13.8MB/s]
100% 15.7M/15.7M [00:01<00:00, 8.88MB/s]

[ ]: !unzip emotions.zip -d /content/emotions

Archive: emotions.zip
inflating: /content/emotions/text.csv

1 Importing
[ ]: import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import matplotlib.pyplot as plt
from matplotlib import cm
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')

[ ]: df = pd.read_csv("/content/emotions/text.csv")
df.head()

[ ]: Unnamed: 0 text label

0 0 i just feel really helpless and heavy hearted 4
1 1 ive enjoyed being able to slouch about relax a… 0
2 2 i gave up my internship with the dmrg and am f… 4
3 3 i dont know i feel so lost 0
4 4 i am a kindergarten teacher and i am thoroughl… 4

[ ]: df.drop(columns="Unnamed: 0", inplace=True)

df.head()

[ ]: text label
0 i just feel really helpless and heavy hearted 4

1
 1 ive enjoyed being able to slouch about relax a… 0
2 i gave up my internship with the dmrg and am f… 4
3 i dont know i feel so lost 0
4 i am a kindergarten teacher and i am thoroughl… 4

[ ]: emotion_map = {
0: "sadness",
1: "joy",
2: "love",
3: "anger",
4: "fear",
5: "surprise"
}
df["label"] = df["label"].map(emotion_map)

[ ]: df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 416809 entries, 0 to 416808
Data columns (total 2 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 text 416809 non-null object
1 label 416809 non-null object
dtypes: object(2)
memory usage: 6.4+ MB

2 Pre Processing
[ ]: # Set the size of the plot
plt.figure(figsize=(10, 6))

# Create a heatmap to visualize missing values

sns.heatmap(df.isnull(), cbar=False, cmap='viridis')

# Display the plot

plt.title('Heatmap of Missing Values')
plt.show()

2
[ ]: # Identify duplicate rows
duplicate_rows = df.duplicated()

# Count the number of duplicate rows

num_duplicates = duplicate_rows.sum()
print(f"Number of duplicate rows: {num_duplicates}")

Number of duplicate rows: 686

[ ]: # Remove exact duplicates, keeping the first occurrence

df = df.drop_duplicates()
print(f"DataFrame shape after removing exact duplicates: {df.shape}")

DataFrame shape after removing exact duplicates: (416123, 2)

[ ]: # Calculate the length of each text entry

df['text_length'] = df['text'].apply(len)

# Display basic statistics

print(df['text_length'].describe())

count 416123.000000
mean 97.102662
std 56.176302
min 2.000000
25% 54.000000

3
 50% 86.000000
75% 128.000000
max 830.000000
Name: text_length, dtype: float64

[ ]: # Set plot style

sns.set(style="whitegrid")

# Plot distribution of text lengths

plt.figure(figsize=(10,6))
sns.histplot(df['text_length'], bins=50, kde=True, color='blue')
plt.title('Distribution of Text Lengths')
plt.xlabel('Number of Characters')
plt.ylabel('Frequency')
plt.show()

[ ]: # Calculate Q1 and Q3
Q1 = df['text_length'].quantile(0.25)
Q3 = df['text_length'].quantile(0.75)
IQR = Q3 - Q1

# Define outlier boundaries

lower_bound = Q1 - 1.5 * IQR
upper_bound = Q3 + 1.5 * IQR

4
 # Identify outliers
outliers_length = df[(df['text_length'] < lower_bound) | (df['text_length'] >␣
↪upper_bound)]

print(f"Number of text length outliers: {outliers_length.shape[0]}")

plt.figure(figsize=(10,6))
sns.histplot(df['text_length'], bins=50, kde=True, color='blue')
plt.axvline(lower_bound, color='red', linestyle='--', label='Lower Bound')
plt.axvline(upper_bound, color='red', linestyle='--', label='Upper Bound')
plt.title('Distribution of Text Lengths with Outlier Boundaries')
plt.xlabel('Number of Characters')
plt.ylabel('Frequency')
plt.legend()
plt.show()

Number of text length outliers: 9952

[ ]: def feature_outlier_analysis(df, feature, start=None, end=None, mode=0,␣

↪save_filename=None):

"""
Analyze and identify outliers based on a numerical feature.

Parameters:

5
 - df (pd.DataFrame): The DataFrame containing the data.
- feature (str): The numerical feature to analyze.
- start (int, optional): The starting value of the feature to consider.␣
↪Defaults to the minimum value.

- end (int, optional): The ending value of the feature to consider.␣

↪Defaults to the maximum value.

- mode (int, optional): If not zero, saves the subset to CSV. Defaults to 0.
- save_filename (str, optional): The filename to save the subset if mode !=␣
↪0.

Returns:
- None
"""
# Ensure the feature exists in the DataFrame
if feature not in df.columns:
print(f"Feature '{feature}' not found in the DataFrame.")
return

# Set default start and end if not provided

if start is None:
start = int(df[feature].min())
if end is None:
end = int(df[feature].max())

print(f"Analyzing feature '{feature}' from {start} to {end}.")

# Iterate over the specified range

for i in range(start, end + 1):
subset = df[df[feature] == i]
count = len(subset)
print(f"Feature '{feature}' = {i}: {count} entries.")

# Save to CSV if mode is not zero

if mode != 0 and save_filename:
subset.to_csv(save_filename, mode='a', index=False) # Append to␣
↪the file

print("Feature outlier analysis completed.")

[ ]: # Calculate sentence length using vectorized string operations

df['sentence_length'] = df['text'].str.split().str.len()

plt.figure(figsize=(14, 8))
sns.boxplot(x='label', y='sentence_length', data=df, palette="coolwarm")
plt.title('Correlation between Labels and Sentence Length', fontsize=16,␣
↪fontweight='bold')

plt.xlabel("Labels", fontsize=14)

6
 plt.ylabel("Sentence Length", fontsize=14)
plt.grid(True, which='both', linestyle='--', linewidth=0.5)
plt.show()

[ ]: def sentence_length_outlier(start=1, end=2, save_csv=False):

"""
Identifies and optionally saves DataFrame rows with sentence lengths within␣
↪a specified range.

Parameters:
- start (int): Starting sentence length (inclusive).
- end (int): Ending sentence length (exclusive).
- save_csv (bool): If True, saves each group to a separate CSV file.

Returns:
- outliers (dict): A dictionary with sentence lengths as keys and␣
↪corresponding DataFrames as values.

"""
outliers = {}
for length in range(start, end):
filtered_df = df[df['sentence_length'] == length]
count = len(filtered_df)

print(f"Sentence length {length} has {count} occurrences.")

7
 if count > 0:
outliers[length] = filtered_df
if save_csv:
filename = f"df_sentence_length_{length}.csv"
filtered_df.to_csv(filename, index=False)
print(f"Saved {count} records to {filename}")

return outliers

# Example usage:
outlier_data = sentence_length_outlier(start=1, end=5, save_csv=True)

Sentence length 1 has 25 occurrences.

Saved 25 records to df_sentence_length_1.csv
Sentence length 2 has 76 occurrences.
Saved 76 records to df_sentence_length_2.csv
Sentence length 3 has 957 occurrences.
Saved 957 records to df_sentence_length_3.csv
Sentence length 4 has 7539 occurrences.
Saved 7539 records to df_sentence_length_4.csv

[ ]: # Define the sentence lengths to remove

lengths_to_remove = [1, 2, 3]

# Filter the DataFrame to exclude these lengths

df= df[~df['sentence_length'].isin(lengths_to_remove)].copy()

# Optionally, reset the index for cleanliness

df.reset_index(drop=True, inplace=True)

# Optional: Save the filtered DataFrame for future use

df.to_csv("df_filtered_sentence_length.csv", index=False)
print(f"Filtered DataFrame saved with {len(df)} records.")

Filtered DataFrame saved with 415065 records.

[ ]: df.shape

[ ]: (415065, 4)

[ ]: # Get the 10 longest sentences and their indices

top_sentences = df.sort_values(by='sentence_length', ascending=False).head(20)
top_indices = top_sentences.index

# Display the top sentences

print("Top 20 Longest Sentences:")
print(top_sentences[['text', 'sentence_length']])

Top 20 Longest Sentences:

8
 text sentence_length
347001 a few days back i was waiting for the bus at t… 178
290349 two years back someone invited me to be the tu… 110
97687 i have been thinking of changing my major for … 101
38584 when i got into a bus i found that my wallet h… 100
22750 my living and working conditions at home were … 100
332276 i had a dream i had a very close friend who ha… 94
249491 i worked with several classmates on a project … 80
158527 i was camping in an old broken hut which had n… 79
56688 last semester when i dated a girl whom ive kno… 78
162121 i was a prefect at secondary school on the spo… 77
174240 when i was in lower six class during the summe… 76
387931 a boy phoned me at night and wanted to talk to… 75
212944 a friend female and i were on holiday on great… 74
35550 i was studying in class at night i was in form… 74
114318 i suddenly found that those whom i considerere… 74
263129 my friend often played a joke on me and someti… 73
337213 a few years ago my mother suffered from cancce… 72
343770 i can get a feel cuz ya make me so horny all i… 71
37247 after attending a song contest proposed by a b… 70
68607 a new gas connection was to be installed and t… 70

[ ]: df = df.drop(index=[348392, 291479, 97990, 38708, 22827, 333596])

[ ]: df.shape

[ ]: (415059, 4)

2.1 Data Cleaning

[ ]: import re
import string
import unicodedata

# Pre-compile regex patterns for efficiency

HTML_TAG_RE = re.compile(r'<.*?>', re.DOTALL)
URL_RE = re.compile(r'https?://\S+|www\.\S+')
NUMERIC_RE = re.compile(r'\d+')
NON_ALPHANUMERIC_RE = re.compile(r'[^a-zA-Z\s]')
EXTRA_WHITESPACE_RE = re.compile(r'\s+')
EMOJI_RE = re.compile(
"["
"\U0001F600-\U0001F64F" # Emoticons
"\U0001F300-\U0001F5FF" # Symbols & Pictographs
"\U0001F680-\U0001F6FF" # Transport & Map Symbols
"\U0001F1E0-\U0001F1FF" # Flags
"]+",
flags=re.UNICODE

9
)

def processing_data(text):
"""
Preprocesses the input text by performing the following operations:
1. Removes HTML tags.
2. Removes URLs.
3. Removes emojis and non-standard Unicode characters.
4. Removes numeric values.
5. Removes non-alphanumeric characters (excluding spaces).
6. Converts text to lowercase.
7. Removes extra whitespaces.
8. Strips leading and trailing whitespaces.

Parameters:
- text (str): The input text to preprocess.

Returns:
- str: The cleaned and preprocessed text.
"""
if pd.isnull(text) or not isinstance(text, str):
return ""

# Normalize unicode characters

text = unicodedata.normalize('NFKD', text)

# Remove HTML tags

text = HTML_TAG_RE.sub('', text)

# Remove URLs
text = URL_RE.sub('', text)

# Remove emojis
text = EMOJI_RE.sub('', text)

# Remove numeric values

text = NUMERIC_RE.sub('', text)

# Remove non-alphanumeric characters (excluding spaces)

text = NON_ALPHANUMERIC_RE.sub('', text)

# Convert to lowercase
text = text.lower()

# Remove extra whitespaces

text = EXTRA_WHITESPACE_RE.sub(' ', text)

10
 # Strip leading and trailing whitespaces
text = text.strip()

return text

[ ]: df['text'] = df['text'].apply(processing_data)

[ ]: df.head(20)

[ ]: text label text_length \

0 i just feel really helpless and heavy hearted fear 45
1 ive enjoyed being able to slouch about relax a… sadness 234
2 i gave up my internship with the dmrg and am f… fear 63
3 i dont know i feel so lost sadness 26
4 i am a kindergarten teacher and i am thoroughl… fear 222
5 i was beginning to feel quite disheartened sadness 42
6 i would think that whomever would be lucky eno… love 125
7 i fear that they won t ever feel that deliciou… joy 124
8 im forever taking some time out to have a lie … surprise 71
9 i can still lose the weight without feeling de… sadness 52
10 i try to be nice though so if you get a bitchy… joy 145
11 im feeling a little like a damaged tree and th… sadness 82
12 i have officially graduated im not feeling as … joy 75
13 i feel like a jerk because the library student… anger 217
14 i feel my portfolio demonstrates how eager i a… joy 120
15 i may be more biased than the next because i h… joy 156
16 i didn t feel terrific joy 22
17 i miss all the others as well that feel that i… anger 102
18 i feel so stupid that i realise it so late sadness 42
19 i saunter through the airport terminals feelin… anger 161

sentence_length
0 8
1 45
2 12
3 7
4 42
5 7
6 25
7 25
8 15
9 9
10 34
11 18
12 14
13 37
14 22

11
 15 35
16 5
17 21
18 10
19 25

[ ]: # Calculate the counts of each emotion

label_counts = df['label'].value_counts()
label_percentages = df['label'].value_counts(normalize=True) * 100

# Create a DataFrame for better readability

label_distribution = pd.DataFrame({
'Count': label_counts,
'Percentage': label_percentages.round(2)
})

print(label_distribution)

Count Percentage
label
joy 140474 33.84
sadness 120692 29.08
anger 57008 13.73
fear 47510 11.45
love 34441 8.30
surprise 14934 3.60

[ ]: # Define colors for each emotion

colors = ['#FF9999','#66B3FF','#99FF99','#FFCC99','#C2C2F0','#FFB3E6']

plt.figure(figsize=(8, 8))
plt.pie(label_counts, labels=label_counts.index, autopct='%1.1f%%',␣
↪startangle=140, colors=colors, explode=(0.05, 0.05, 0.05, 0.05, 0.05, 0.05))

plt.title('Emotion Distribution in Dataset', fontsize=16, fontweight='bold')

plt.axis('equal') # Ensures that pie is drawn as a circle.
plt.show()

12
[ ]: plt.figure(figsize=(10, 6))
sns.countplot(x='label', data=df, palette="Set2", order=label_counts.index)
plt.title('Emotion Distribution in Dataset', fontsize=16, fontweight='bold')
plt.xlabel('Emotion', fontsize=14)
plt.ylabel('Number of Occurrences', fontsize=14)

# Adding the count labels on top of each bar

for index, value in enumerate(label_counts):
plt.text(index, value + 50, str(value), ha='center', va='bottom',␣
↪fontsize=12)

plt.show()

13
 2.2 Preprocessing Lemmatization and Filtering StopWords
[ ]: import time
import re
import spacy
from tqdm import tqdm
import pandas as pd

# Load SpaCy model with parser and NER disabled for efficiency
nlp = spacy.load("en_core_web_sm", disable=['parser', 'ner'])

# Set maximum text length if needed (adjust as per your dataset)

nlp.max_length = 1_500_000

def process_text(doc):
"""
Processes a SpaCy Doc object by lemmatizing tokens, removing stopwords, and␣
↪filtering non-alphabetic tokens.

Parameters:
- doc (spacy.tokens.Doc): The SpaCy Doc object to process.

Returns:
- str: The processed text.

14
 """
return ' '.join(
token.lemma_ for token in doc
if not token.is_stop and token.is_alpha
)

def process_texts_sequentially(texts, batch_size=500):

"""
Processes a list of texts sequentially using SpaCy's nlp.pipe with a␣
↪progress bar.

Parameters:
- texts (list of str): The list of text strings to process.
- batch_size (int): The number of texts to process in each batch.

Returns:
- list of str: The list of processed text strings.
"""
processed_texts = []

# Initialize tqdm progress bar

for doc in tqdm(nlp.pipe(texts, batch_size=batch_size, disable=["parser",␣
↪"ner"]), total=len(texts), desc="Processing Texts"):

processed_text = process_text(doc)
processed_texts.append(processed_text)

return processed_texts

if __name__ == "__main__":
# Load your dataset here
# Example:
# df = pd.read_csv("your_data.csv")

# Ensure that the DataFrame 'df' exists and has a 'text' column
try:
df
except NameError:
print("Error: DataFrame 'df' is not defined. Please load your data␣
↪before processing.")

exit()

if 'text' not in df.columns:

print("Error: The DataFrame does not contain a 'text' column.")
exit()

# Handle missing texts

missing_texts = df['text'].isnull().sum()

15
 if missing_texts > 0:
print(f"Warning: Found {missing_texts} missing texts. Filling them with␣
↪empty strings.")

df['text'] = df['text'].fillna("")

# Convert the 'text' column to a list

texts = df['text'].tolist()

# Start the processing timer

start_time = time.time()
print("� Starting text processing...")

# Process the texts sequentially with a progress bar

df['processed_text'] = process_texts_sequentially(texts, batch_size=500)

# End the processing timer

end_time = time.time()
processing_time = end_time - start_time
print(f"� Processing completed in {processing_time:.2f} seconds")

� Starting text processing…

Processing Texts: 100%|����������| 415059/415059 [08:42<00:00, 794.71it/s]
� Processing completed in 522.34 seconds

2.3 Data Preparation for Pre Trained Model

[ ]: !pip install datasets

Collecting datasets
Downloading datasets-3.2.0-py3-none-any.whl.metadata (20 kB)
Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-
packages (from datasets) (3.16.1)
Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-
packages (from datasets) (1.26.4)
Requirement already satisfied: pyarrow>=15.0.0 in
/usr/local/lib/python3.10/dist-packages (from datasets) (17.0.0)
Collecting dill<0.3.9,>=0.3.0 (from datasets)
Downloading dill-0.3.8-py3-none-any.whl.metadata (10 kB)
Requirement already satisfied: pandas in /usr/local/lib/python3.10/dist-packages
(from datasets) (2.2.2)
Requirement already satisfied: requests>=2.32.2 in
/usr/local/lib/python3.10/dist-packages (from datasets) (2.32.3)
Requirement already satisfied: tqdm>=4.66.3 in /usr/local/lib/python3.10/dist-
packages (from datasets) (4.67.1)
Collecting xxhash (from datasets)
Downloading
xxhash-3.5.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata

16
(12 kB)
Collecting multiprocess<0.70.17 (from datasets)
Downloading multiprocess-0.70.16-py310-none-any.whl.metadata (7.2 kB)
Collecting fsspec<=2024.9.0,>=2023.1.0 (from
fsspec[http]<=2024.9.0,>=2023.1.0->datasets)
Downloading fsspec-2024.9.0-py3-none-any.whl.metadata (11 kB)
Requirement already satisfied: aiohttp in /usr/local/lib/python3.10/dist-
packages (from datasets) (3.11.10)
Requirement already satisfied: huggingface-hub>=0.23.0 in
/usr/local/lib/python3.10/dist-packages (from datasets) (0.27.0)
Requirement already satisfied: packaging in /usr/local/lib/python3.10/dist-
packages (from datasets) (24.2)
Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-
packages (from datasets) (6.0.2)
Requirement already satisfied: aiohappyeyeballs>=2.3.0 in
/usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (2.4.4)
Requirement already satisfied: aiosignal>=1.1.2 in
/usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (1.3.2)
Requirement already satisfied: async-timeout<6.0,>=4.0 in
/usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (4.0.3)
Requirement already satisfied: attrs>=17.3.0 in /usr/local/lib/python3.10/dist-
packages (from aiohttp->datasets) (24.3.0)
Requirement already satisfied: frozenlist>=1.1.1 in
/usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (1.5.0)
Requirement already satisfied: multidict<7.0,>=4.5 in
/usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (6.1.0)
Requirement already satisfied: propcache>=0.2.0 in
/usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (0.2.1)
Requirement already satisfied: yarl<2.0,>=1.17.0 in
/usr/local/lib/python3.10/dist-packages (from aiohttp->datasets) (1.18.3)
Requirement already satisfied: typing-extensions>=3.7.4.3 in
/usr/local/lib/python3.10/dist-packages (from huggingface-hub>=0.23.0->datasets)
(4.12.2)
Requirement already satisfied: charset-normalizer<4,>=2 in
/usr/local/lib/python3.10/dist-packages (from requests>=2.32.2->datasets)
(3.4.0)
Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-
packages (from requests>=2.32.2->datasets) (3.10)
Requirement already satisfied: urllib3<3,>=1.21.1 in
/usr/local/lib/python3.10/dist-packages (from requests>=2.32.2->datasets)
(2.2.3)
Requirement already satisfied: certifi>=2017.4.17 in
/usr/local/lib/python3.10/dist-packages (from requests>=2.32.2->datasets)
(2024.12.14)
Requirement already satisfied: python-dateutil>=2.8.2 in
/usr/local/lib/python3.10/dist-packages (from pandas->datasets) (2.8.2)
Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.10/dist-
packages (from pandas->datasets) (2024.2)

17
 Requirement already satisfied: tzdata>=2022.7 in /usr/local/lib/python3.10/dist-
packages (from pandas->datasets) (2024.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.10/dist-
packages (from python-dateutil>=2.8.2->pandas->datasets) (1.17.0)
Downloading datasets-3.2.0-py3-none-any.whl (480 kB)
���������������������������������������� 480.6/480.6 kB
18.0 MB/s eta 0:00:00
Downloading dill-0.3.8-py3-none-any.whl (116 kB)
���������������������������������������� 116.3/116.3 kB
6.7 MB/s eta 0:00:00
Downloading fsspec-2024.9.0-py3-none-any.whl (179 kB)
���������������������������������������� 179.3/179.3 kB
14.4 MB/s eta 0:00:00
Downloading multiprocess-0.70.16-py310-none-any.whl (134 kB)
���������������������������������������� 134.8/134.8 kB
10.8 MB/s eta 0:00:00
Downloading
xxhash-3.5.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (194 kB)
���������������������������������������� 194.1/194.1 kB
9.6 MB/s eta 0:00:00
Installing collected packages: xxhash, fsspec, dill, multiprocess,
datasets
Attempting uninstall: fsspec
Found existing installation: fsspec 2024.10.0
Uninstalling fsspec-2024.10.0:
Successfully uninstalled fsspec-2024.10.0
ERROR: pip's dependency resolver does not currently take into account all
the packages that are installed. This behaviour is the source of the following
dependency conflicts.
gcsfs 2024.10.0 requires fsspec==2024.10.0, but you have fsspec 2024.9.0 which
is incompatible.
Successfully installed datasets-3.2.0 dill-0.3.8 fsspec-2024.9.0
multiprocess-0.70.16 xxhash-3.5.0

[ ]: from transformers import AutoTokenizer, AutoModelForSequenceClassification,␣

↪Trainer, TrainingArguments

from sklearn.preprocessing import LabelEncoder

from sklearn.metrics import classification_report, accuracy_score,␣
↪precision_recall_fscore_support

from datasets import Dataset

# Initialize the label encoder

label_encoder = LabelEncoder()

# Fit and transform the labels

18
 df['label_encoded'] = label_encoder.fit_transform(df['label'])

# Display the mapping

label_mapping = dict(zip(label_encoder.classes_, label_encoder.
↪transform(label_encoder.classes_)))

print("Label Mapping:", label_mapping)

Label Mapping: {'anger': 0, 'fear': 1, 'joy': 2, 'love': 3, 'sadness': 4,

'surprise': 5}

[ ]: from sklearn.model_selection import train_test_split

# Define features and target

X = df['processed_text'].values
y = df['label_encoded'].values

# Split the data

X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.2, random_state=42, stratify=y
)

print(f"Training samples: {len(X_train)}")

print(f"Testing samples: {len(X_test)}")

Training samples: 332047

Testing samples: 83012

[ ]: # Create Pandas DataFrames for training and testing

train_df = pd.DataFrame({'text': X_train, 'label': y_train})
test_df = pd.DataFrame({'text': X_test, 'label': y_test})

# Convert to Hugging Face Dataset

train_dataset = Dataset.from_pandas(train_df)
test_dataset = Dataset.from_pandas(test_df)

print("Training Dataset:", train_dataset)

print("Testing Dataset:", test_dataset)

Training Dataset: Dataset({

features: ['text', 'label'],
num_rows: 332047
})
Testing Dataset: Dataset({
features: ['text', 'label'],
num_rows: 83012
})

19
 2.4 Tokenization
[ ]: model_name = "distilbert-base-uncased"
# Initialize the tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)

def tokenize_function(examples):
return tokenizer(
examples['text'],
padding="max_length",
truncation=True,
max_length=128 # Adjust based on your data
)

# Apply tokenization to the datasets

train_dataset = train_dataset.map(tokenize_function, batched=True,␣
↪batch_size=1000)

test_dataset = test_dataset.map(tokenize_function, batched=True,␣

↪batch_size=1000)

# Rename the 'label' column to 'labels' as expected by Hugging Face

train_dataset = train_dataset.rename_column("label", "labels")
test_dataset = test_dataset.rename_column("label", "labels")

# Set the format for PyTorch

train_dataset.set_format(type='torch', columns=['input_ids', 'attention_mask',␣
↪'labels'])

test_dataset.set_format(type='torch', columns=['input_ids', 'attention_mask',␣

↪'labels'])

tokenizer_config.json: 0%| | 0.00/48.0 [00:00<?, ?B/s]

config.json: 0%| | 0.00/483 [00:00<?, ?B/s]
vocab.txt: 0%| | 0.00/232k [00:00<?, ?B/s]
tokenizer.json: 0%| | 0.00/466k [00:00<?, ?B/s]
Map: 0%| | 0/332047 [00:00<?, ? examples/s]
Map: 0%| | 0/83012 [00:00<?, ? examples/s]

2.5 Model Initialization

[ ]: import torch

# Check if CUDA is available

if torch.cuda.is_available():
print("CUDA is available. Training on GPU...")

20
 else:
print("CUDA is not available. Training on CPU...")

CUDA is available. Training on GPU…

[ ]: num_labels = len(label_mapping)

# Load the pre-trained model

model = AutoModelForSequenceClassification.from_pretrained(
model_name,
num_labels=num_labels
)

def compute_metrics(pred):
labels = pred.label_ids
preds = pred.predictions.argmax(-1)
precision, recall, f1, _ = precision_recall_fscore_support(labels, preds,␣
↪average='weighted')

acc = accuracy_score(labels, preds)

return {
'accuracy': acc,
'f1': f1,
'precision': precision,
'recall': recall
}

training_args = TrainingArguments(
output_dir='./results', # Output directory
evaluation_strategy="epoch", # Evaluation strategy to adopt␣
↪during training

save_strategy="epoch", # Save strategy to adopt during␣

↪training

learning_rate=2e-5, # Learning rate

per_device_train_batch_size=16, # Batch size per device during␣
↪training

per_device_eval_batch_size=16, # Batch size for evaluation

num_train_epochs=3, # Total number of training epochs
weight_decay=0.01, # Strength of weight decay
logging_dir='./logs', # Directory for storing logs
logging_steps=100, # Log every X updates steps
load_best_model_at_end=True, # Load the best model when finished␣
↪training

metric_for_best_model='f1', # Use F1 score to evaluate the best␣

↪model

greater_is_better=True # Whether the metric should be␣

↪maximized

21
 Some weights of DistilBertForSequenceClassification were not initialized from
the model checkpoint at distilbert-base-uncased and are newly initialized:
['classifier.bias', 'classifier.weight', 'pre_classifier.bias',
'pre_classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it
for predictions and inference.

[ ]: from transformers import Trainer

# Initialize the Trainer

trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset,
eval_dataset=test_dataset,
compute_metrics=compute_metrics
)

# Check the device being used by the model

print("Training on:", trainer.args.device)

# Start training
trainer.train()

Training on: cuda:0

<IPython.core.display.HTML object>

[ ]: TrainOutput(global_step=62259, training_loss=0.1351904600557597,
metrics={'train_runtime': 11252.9092, 'train_samples_per_second': 88.523,
'train_steps_per_second': 5.533, 'total_flos': 3.299140499276851e+16,
'train_loss': 0.1351904600557597, 'epoch': 3.0})

2.6 Evaluation
[ ]: # Evaluate the model on the test set
results = trainer.evaluate()
print("Evaluation Results:", results)

<IPython.core.display.HTML object>
Evaluation Results: {'eval_loss': 0.13016673922538757, 'eval_accuracy':
0.9320821086108032, 'eval_f1': 0.9328861620931987, 'eval_precision':
0.9387182163079509, 'eval_recall': 0.9320821086108032, 'eval_runtime': 271.8072,
'eval_samples_per_second': 305.408, 'eval_steps_per_second': 19.091, 'epoch':
3.0}

[ ]: # Get predictions
predictions = trainer.predict(test_dataset)
pred_labels = np.argmax(predictions.predictions, axis=1)

22
 true_labels = predictions.label_ids

# Classification Report
print("=== Classification Report ===\n")
print(classification_report(true_labels, pred_labels,␣
↪target_names=label_mapping.keys()))

<IPython.core.display.HTML object>
=== Classification Report ===

precision recall f1-score support

anger 0.92 0.96 0.94 11402

fear 0.98 0.81 0.89 9502
joy 0.97 0.92 0.95 28095
love 0.76 0.97 0.85 6888
sadness 0.96 0.97 0.97 24138
surprise 0.76 0.89 0.82 2987

accuracy 0.93 83012

macro avg 0.89 0.92 0.90 83012
weighted avg 0.94 0.93 0.93 83012

[ ]: from sklearn.metrics import confusion_matrix

cm = confusion_matrix(true_labels, pred_labels)
plt.figure(figsize=(10,8))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
xticklabels=label_mapping.keys(),
yticklabels=label_mapping.keys())
plt.ylabel('Actual')
plt.xlabel('Predicted')
plt.title('Confusion Matrix')
plt.show()

23
 2.7 Prediction and Inference
[ ]: def predict_emotion(text):
# Tokenize the input text
inputs = tokenizer(
text,
return_tensors="pt",
padding=True,
truncation=True,
max_length=128
)

# Move inputs to the model's device

device = torch.device('cuda') if torch.cuda.is_available() else torch.
↪device('cpu')

model.to(device)
inputs = {k: v.to(device) for k, v in inputs.items()}

24
 # Get model predictions
with torch.no_grad():
outputs = model(**inputs)
logits = outputs.logits
predicted_class = torch.argmax(logits, dim=1).cpu().numpy()[0]

# Map back to label

emotion = label_encoder.inverse_transform([predicted_class])[0]
return emotion

[ ]: sample_text = "I am so happy and excited about the new opportunities!"

predicted_emotion = predict_emotion(sample_text)
print(f"Text: {sample_text}\nPredicted Emotion: {predicted_emotion}")

Text: I am so happy and excited about the new opportunities!

Predicted Emotion: joy

2.8 Save for future use

[ ]: # Save the model and tokenizer to a directory
output_dir = "/content/saved_model"

# Ensure the directory exists

import os
os.makedirs(output_dir, exist_ok=True)

# Save the model

trainer.save_model(output_dir)

# Save the tokenizer

tokenizer.save_pretrained(output_dir)

print(f"Model and tokenizer saved to {output_dir}")

Model and tokenizer saved to /content/saved_model

25