Sentiment Analysis on Response PDF file

Response PDF

• The Response PDF file comprises 106 student feedback responses, all centered around the fascinating subject of Mathematics. These responses are neatly organized in a table format in PDF, featuring 15 distinct questions and corresponding answers.
• I have delve into techniques for PDF data extraction, gaining practical skills. Sentiment analysis will allow to gauge student sentiments, providing valuable context.

Project Objective

• The mission is to uncover insights from PDF files and perform sentiment analysis on the student feedback.
• Extracting data from the PDF and transforming it into a structured dataframe will be our key challenge.
• Through this project, I will bridge the gap between raw PDF content and meaningful data insights.

Deployment

Reading PDF and Data Cleaning

• Utilized the PdfReader from the PyPDF2 library to read the Response.pdf file.
• Extracted the data as text into a NumPy array.
• Split the text based on question marks (‘?’) to isolate individual questions.
• Segregated the 15 questions into a list called “ques” and collected student answers into a separate list called “feedbacks”.
• Removed timestamps and dates from the feedbacks.
• Developed a function to format feedback elements into new list after every 14 entries.
• Identified unique student responses and organized them into a nested list representing all student feedback.

Data Frame Conversion

• Rename nested list as “df_row” for feedbacks and renamed the ques list to “col_df”.
• Converted both lists into Pandas DataFrames.
• Concatenated the df_row and col_df DataFrames to create a single table resembling the PDF structure.
• Shifted the first row to serve as the header, resulting in 15 distinct responses across 106 student rows.

Column-wise Sentiment Analysis

• Counted the student responses for each distinctive option within a question column.
• Assigned sentiment scores (positive, zero, negative) to these options.
• Calculated the total sentiment score by multiplying the count with the corresponding sentiment score and dividing by the total count (106 students).
• Computed the average sentiment score for each question.
• Plotted pie graphs to visualize student responses for each question.
• This provided an overview of sentiment distribution across options.

Row-wise Sentiment Analysis

• Developed a function for row-wise sentiment analysis using TextBlob.
• The result was divided by 15 to obtain the final sentiment value for each student.
• If the value was ≤ 0.12, the student was considered confident with a strong foundation in mathematics; otherwise, they were deemed under-confident with a weaker base.

Sentiment Analysis on Response PDF Presentation