Sentiment Analysis of Amazon Reviews

Analyze sentiment in Amazon product reviews using VADER and TextBlob, visualize score distributions, and identify most positive and negative reviews.

This page shows an AI data analysis example. A workflow is a list of prompts (instructions) used to complete one analysis task step by step. We send the same prompt list to multiple AI models, then compare their notebook outputs and scores.

What this AI workflow does

This AI Data Analyst workflow loads a sample of the Amazon Fine Food Reviews dataset from a URL and summarizes dataset shape and star rating distribution. It generates VADER sentiment scores for each review and adds them as new columns for analysis. It visualizes sentiment score distributions and their relationship to star ratings, then extracts the most positive and most negative review excerpts with scores.

Who this example is for

This is for analysts and data scientists who want a reproducible notebook pattern for sentiment scoring and basic validation against existing labels like star ratings. It is also useful for NLP learners comparing lexicon-based sentiment methods and reviewing edge cases by inspecting extreme examples.

Expected analysis outcomes

These are the results the AI workflow is expected to generate.

  • Dataset shape and star rating histogram
  • VADER sentiment scores appended as new columns
  • Histogram of compound sentiment scores
  • Scatter plot comparing sentiment scores to star ratings
  • Three most positive and three most negative review excerpts with sentiment scores

Tools and libraries used

Main Python packages and tooling used to run this AI data analysis task.

  • pandas
  • nltk
  • vaderSentiment
  • textblob
  • matplotlib
  • seaborn

Prompt sequence

This is the exact list of prompts used in this workflow. The same prompt sequence is sent to each model so outputs and scores can be compared fairly.

  1. 1load the reviews dataset https://raw.githubusercontent.com/pplonski/datasets-for-start/refs/heads/master/amazon-fine-food-reviews/amazon_fine_food_reviews_10k.csv and show shape and rating distribution
  2. 2compute sentiment scores using VADER for each review
  3. 3plot sentiment score distribution and compare with star ratings
  4. 4show the 3 most positive and 3 most negative reviews

Model Comparison

We compared several LLM models on the same analysis task. The table below shows their scores side by side. You can also open each model run to check the full conversation and notebook results.

Model SourceScoreTaskExecutionOutputReasoningReliabilityRun
gemma4:31b10/102/22/23/32/21/1Open gemma4:31b conversation
gpt-oss:120b10/102/22/23/32/21/1Open gpt-oss:120b conversation
qwen3-coder-next9/102/22/23/31/21/1Open qwen3-coder-next conversation
glm-5.18/102/22/23/31/20/1Open glm-5.1 conversation
gpt-5.48/101/22/22/32/21/1Open gpt-5.4 conversation
qwen3.5:397b4/101/21/21/31/20/1Open qwen3.5:397b conversation
How scoring works

Each run is graded on five dimensions:

  • Task Completion (0-2): how fully the workflow answered the requested analysis task and completed required steps.
  • Execution Correctness (0-2): whether generated code is correct, consistent, and likely runnable without major fixes.
  • Output Quality (0-3): how well tables, charts, and results match the expected analytical outcomes.
  • Reasoning Quality (0-2): clarity and correctness of interpretation, explanation, and conclusions.
  • Reliability (0-1): robustness of the run, including low hallucination risk and stable behavior.

The final score is the sum of these dimensions, so total score = 0-10. Higher values indicate more complete, correct, and reliable notebook results for this workflow.

AI Analysis with qwen3-coder-next

This run uses AI with a step-by-step list of prompts to perform the analysis task Sentiment Analysis of Amazon Reviews. Below, you can review the full conversation generated with qwen3-coder-next. After the conversation, we show detailed score cards for this run.

#1
load the reviews dataset https://raw.githubusercontent.com/pplonski/datasets-for-start/refs/heads/master/amazon-fine-food-reviews/amazon_fine_food_reviews_10k.csv and show shape and rating distribution
#2

I'll load the dataset and show its shape and rating distribution.

#3
Load Amazon food review dataset
import pandas as pd

# Load the dataset
url = "https://raw.githubusercontent.com/pplonski/datasets-for-start/refs/heads/master/amazon-fine-food-reviews/amazon_fine_food_reviews_10k.csv"
df = pd.read_csv(url)

# Show shape
df.shape
#4
(10000, 10)
#5

The dataset has been loaded successfully with a shape of (10000, 10), meaning there are 10,000 rows and 10 columns. To complete the goal of showing the rating distribution, I need to inspect the rating column values.

#6
Count and sort score values
df['Score'].value_counts().sort_index()
#7
Score
1     932
2     590
3     862
4    1433
5    6183
Name: count, dtype: int64
#8

The dataset has been successfully loaded with 10,000 rows and 10 columns. The rating distribution shows that most reviews are positive, with 6,183 reviews giving a rating of 5 stars (61.8%), followed by 1,434 with 4 stars (14.3%). Negative ratings (1 and 2 stars)合计 1,522 reviews (15.2% + 5.9% = 21.1%), indicating a generally positive sentiment in the dataset.

#9
compute sentiment scores using VADER for each review
#10

The dataset has been loaded with 10,000 reviews, and the rating distribution is now known. To compute sentiment scores using VADER, I'll first install the vaderSentiment library and then apply it to the review text. The review text column is likely Text — I’ll confirm it once I load and inspect the dataset again. Let's proceed:

  1. Install vaderSentiment
  2. Inspect the dataset to confirm the review text column
  3. Compute VADER sentiment scores

Let’s first inspect the column names to locate the review text field.

#11
List column names
df.columns.tolist()
#12
['Id',
 'ProductId',
 'UserId',
 'ProfileName',
 'HelpfulnessNumerator',
 'HelpfulnessDenominator',
 'Score',
 'Time',
 'Summary',
 'Text']
#13

The notebook state confirms the dataset has 10,000 rows and 10 columns. The columns include Text, which contains the review content — this is the field we need for VADER sentiment analysis. However, VADER sentiment scores have not yet been computed. To proceed with the original goal, I must install vaderSentiment and apply it to the Text column.

The next step is to compute sentiment scores using VADER.

#14
Compute VADER sentiment scores
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
import pandas as pd

# Initialize VADER analyzer
analyzer = SentimentIntensityAnalyzer()

# Compute VADER sentiment scores for each review
# Only process non-null text to avoid errors
df['vader_pos'] = 0.0
df['vader_neu'] = 0.0
df['vader_neg'] = 0.0
df['vader_compound'] = 0.0

for i, text in enumerate(df['Text']):
    if pd.notna(text) and isinstance(text, str):
        scores = analyzer.polarity_scores(text)
        df.at[i, 'vader_pos'] = scores['pos']
        df.at[i, 'vader_neu'] = scores['neu']
        df.at[i, 'vader_neg'] = scores['neg']
        df.at[i, 'vader_compound'] = scores['compound']

# Show first 5 rows with VADER scores
df[['Text', 'Score', 'vader_pos', 'vader_neu', 'vader_neg', 'vader_compound']].head(5)
#15
Text Score vader_pos vader_neu vader_neg vader_compound
0 I have bought several of the Vitality canned d... 5 0.289 0.711 0.000 0.9441
1 Product arrived labeled as Jumbo Salted Peanut... 1 0.000 0.862 0.138 -0.5664
2 This is a confection that has been around a fe... 4 0.144 0.771 0.085 0.8138
3 If you are looking for the secret ingredient i... 2 0.068 0.932 0.000 0.4404
4 Great taffy at a great price.  There was a wid... 5 0.401 0.599 0.000 0.9468
#16

VADER sentiment scores (vader_pos, vader_neu, vader_neg, vader_compound) have been successfully computed for all reviews. The output shows the first 5 reviews with their sentiment breakdown — for example, the first review (Score = 5) has a highly positive compound score of 0.9441, and the second review (Score = 1) shows a negative compound score of -0.5664, aligning intuitively with their star ratings.

This confirms the goal is complete: sentiment scores using VADER are now available for each of the 10,000 reviews.

#17
plot sentiment score distribution and compare with star ratings
#18

We now have VADER sentiment scores (vader_pos, vader_neu, vader_neg, vader_compound) and the original Score (1–5 stars). To compare sentiment with star ratings, I’ll create side-by-side box plots for each sentiment metric across rating levels. This will visually show how sentiment compounds align with star scores — especially helpful since vader_compound is a standard normalized score ranging from -1 to +1.

#19
Plot VADER sentiment by star rating
import seaborn as sns
import matplotlib.pyplot as plt

# Prepare data: only keep reviews with valid compound scores
df_plot = df[['Score', 'vader_compound']].dropna()

# Create a box plot of compound sentiment by star rating
plt.figure(figsize=(8, 5))
sns.boxplot(data=df_plot, x='Score', y='vader_compound', order=[1,2,3,4,5])
plt.title('VADER Compound Sentiment Distribution by Star Rating')
plt.xlabel('Star Rating (Score)')
plt.ylabel('VADER Compound Score')
plt.ylim(-0.7, 1.0)  # tightens y-axis for better visibility
plt.show()
#20
Notebook output
#21

The latest executed output shows the box plot of VADER compound sentiment across star ratings, and the notebook state confirms the plot was generated successfully. The chart visually demonstrates that higher star ratings (4–5) generally have higher compound sentiment scores (close to +0.5 to +1), while lower ratings (1–2) show more negative sentiment (closer to 0 or negative). Rating 3 (neutral) centers around 0 compound score, matching expectations.

This directly answers the original goal: plotting and comparing sentiment with star ratings.

#22
show the 3 most positive and 3 most negative reviews
#23

To find the 3 most positive and 3 most negative reviews, I’ll sort the DataFrame by vader_compound in descending and ascending order, then display the top reviews with their text, score, and compound value.

Here’s the next step:

#24
Show top 3 most positive reviews
# Most positive reviews (top 3 by vader_compound)
most_positive = df.nlargest(3, 'vader_compound')[['Score', 'vader_compound', 'Summary', 'Text']]
most_positive
#25
Score vader_compound Summary Text
1320 1 0.9997 NOT what I originally ordered! BOO, amazon! As another reviewer stated, many of the review...
8756 5 0.9997 At last we find a food fit for human consumption And we find her here upon this grand amazon!<b...
539 5 0.9993 C H I P.....C H I P.....H O O R A Y....!!!!!  ... I was getting VERY worried, when I read some o...
#26

The latest output ([output-6]) shows the 3 most positive reviews, with their Score, vader_compound, Summary, and Text. It displays one highly negative rating (Score 1) and two positive ratings (Score 5), all with near-perfect compound scores (~0.9997). However, the 3 most negative reviews (lowest vader_compound) are not yet shown — only the top 3 positive were requested and executed.

To complete the original goal, I need to retrieve and show the 3 most negative reviews.

#27
Show most negative reviews
# Most negative reviews (bottom 3 by vader_compound)
most_negative = df.nsmallest(3, 'vader_compound')[['Score', 'vader_compound', 'Summary', 'Text']]
most_negative
#28
Score vader_compound Summary Text
8076 1 -0.9903 Purina Busy Bone - A Lethal Product for your Dog I write this letter after considerable thought...
6504 1 -0.9875 NOT GLUTEN FREE! After eating this product, I had a stomach ach...
5364 2 -0.9852 A poor choice Instead of just being pork rinds, oil, salt an...
#29

The latest outputs ([output-6] and [output-7]) successfully answer the original goal:

  • Top 3 most positive reviews (by vader_compound score) are:

    • Review ID 1320: Score = 1, compound = 0.9997, Summary = "NOT what I originally ordered! BOO, amazon!", with a very high compound sentiment despite low rating — suggesting highly negative sentiment was expressed with strong wording.
    • Review ID 8756: Score = 5, compound = 0.9997, Summary = "At last we find a food fit for human consumption", expressing extreme positivity.
    • Review ID 539: Score = 5, compound = 0.9993, with enthusiastic capitalization and exclamation marks.

  • Top 3 most negative reviews (lowest vader_compound scores):

    • Review ID 8076: Score = 1, compound = −0.9903, Summary = "Purina Busy Bone — A Lethal Product for your Dog", extremely alarming tone.
    • Review ID 6504: Score = 1, compound = −0.9875, Summary = "NOT GLUTEN FREE!", strong capitalization and emotional impact.
    • Review ID 5364: Score = 2, compound = −0.9852, Summary = "A poor choice", negative despite slightly higher rating.

All 6 reviews needed (3 most positive + 3 most negative) are now displayed with full text, scores, and VADER compound values — the goal is fully satisfied.

Score cards for qwen3-coder-next for Sentiment Analysis of Amazon Reviews

Total score: 9/10

  • Task Completion: 2/2

    Excellent

    The workflow loads the dataset, shows shape and rating distribution, computes VADER sentiment scores, visualizes sentiment vs. ratings, and displays the 3 most positive and 3 most negative reviews.

  • Execution Correctness: 2/2

    Excellent

    The code is coherent and likely runnable: it reads the CSV, computes VADER scores in a loop, and produces the requested plots and ranked reviews without obvious syntax errors.

  • Output Quality: 3/3

    Excellent

    The outputs match the expected outcomes semantically: shape and rating histogram data are shown, sentiment columns are added, a sentiment distribution plot is generated, and six review excerpts with scores are displayed.

  • Reasoning Quality: 1/2

    Needs work

    The reasoning is mostly correct but includes some unsupported or contradictory claims, such as saying the top positive reviews include a low-rating review as evidence of positivity and mixing in unnecessary narrative.

  • Reliability: 1/1

    Excellent

    The workflow is reasonably consistent and uses standard methods, though it relies on manual looping and includes some questionable interpretation of the ranked reviews.

Try MLJAR Studio

Run the same type of AI-powered data analysis on your own datasets with conversational notebooks in MLJAR Studio.

Download Free TrialLearn about AI Data Analyst

Explore More AI Analysis Examples

Discover additional workflows across categories. Each example includes prompts, conversation outputs, and model-level scoring so you can compare approaches and results.

Exploratory Data Analysis

Time Series

Data Analysis

Machine Learning

NLP

Finance

Statistics