NLP Program to Analyze Sentiment with Example

Use the Hugging Face transformers library with pipeline('sentiment-analysis') to create an NLP program that analyzes sentiment, like from transformers import pipeline; sentiment = pipeline('sentiment-analysis'); print(sentiment('I love this!')).

Examples

InputI love this product!

Output[{'label': 'POSITIVE', 'score': 0.9998}]

InputThis is the worst movie I have ever seen.

Output[{'label': 'NEGATIVE', 'score': 0.9997}]

Input

Output[]

How to Think About It

To analyze sentiment, the program takes a text input and uses a pre-trained model to classify the text as positive or negative. The model looks at the words and their context to decide the overall feeling. We use a ready-made tool that handles this complex task so we only need to give it the text and get the sentiment result.

Algorithm

1

Import the sentiment analysis tool from the NLP library.

2

Create a sentiment analysis pipeline or model instance.

3

Input the text to analyze.

4

Run the text through the model to get sentiment prediction.

5

Output the sentiment label and confidence score.

Code

python

from transformers import pipeline

sentiment = pipeline('sentiment-analysis')

texts = [
    'I love this product!',
    'This is the worst movie I have ever seen.',
    ''
]

for text in texts:
    result = sentiment(text) if text else []
    print(f'Text: "{text}" -> Sentiment: {result}')

Output

Text: "I love this product!" -> Sentiment: [{'label': 'POSITIVE', 'score': 0.9998}]
Text: "This is the worst movie I have ever seen." -> Sentiment: [{'label': 'NEGATIVE', 'score': 0.9997}]
Text: "" -> Sentiment: []

Dry Run

Let's trace the input 'I love this product!' through the code.

1

Import pipeline

Load the sentiment-analysis pipeline from transformers.

2

Create sentiment pipeline

sentiment = pipeline('sentiment-analysis') creates the model ready to analyze text.

3

Input text

Text to analyze: 'I love this product!'

4

Run analysis

Call sentiment('I love this product!') which returns [{'label': 'POSITIVE', 'score': 0.9998}]

5

Print output

Prints: Text: "I love this product!" -> Sentiment: [{'label': 'POSITIVE', 'score': 0.9998}]

Step	Input Text	Output Sentiment
1	I love this product!	[{'label': 'POSITIVE', 'score': 0.9998}]

Why This Works

Step 1: Use pre-trained model

The pipeline('sentiment-analysis') loads a model trained on many texts to understand positive or negative feelings.

Step 2: Input text processing

The model reads the input text and converts it into numbers it can understand.

Step 3: Prediction output

The model outputs a label like POSITIVE or NEGATIVE with a confidence score showing how sure it is.

Alternative Approaches

TextBlob library

python

from textblob import TextBlob

def analyze_sentiment(text):
    blob = TextBlob(text)
    polarity = blob.sentiment.polarity
    label = 'POSITIVE' if polarity > 0 else 'NEGATIVE' if polarity < 0 else 'NEUTRAL'
    return {'label': label, 'score': abs(polarity)}

print(analyze_sentiment('I love this product!'))

TextBlob is simpler and faster but less accurate than transformer models.

VADER Sentiment Analyzer

python

from nltk.sentiment import SentimentIntensityAnalyzer

sia = SentimentIntensityAnalyzer()

text = 'I love this product!'
scores = sia.polarity_scores(text)
label = 'POSITIVE' if scores['compound'] > 0 else 'NEGATIVE' if scores['compound'] < 0 else 'NEUTRAL'
print({'label': label, 'score': scores['compound']})

VADER works well for social media text and is lightweight but less nuanced.

Complexity: O(n) time, O(n) space

Time Complexity

The time depends linearly on the length of the input text n because the model processes each token once.

Space Complexity

Space grows with input size n to store tokenized text and intermediate model states.

Which Approach is Fastest?

TextBlob and VADER are faster but less accurate; transformer pipelines are slower but provide better results.

Approach	Time	Space	Best For
Hugging Face pipeline	O(n)	O(n)	High accuracy, general use
TextBlob	O(n)	O(n)	Simple, fast, less accurate
VADER	O(n)	O(n)	Social media, lightweight