Review Sentiment Aggregator

1. Overview

The Review Sentiment Aggregator reads a set of customer reviews and produces a concise summary that tells you the overall feeling (sentiment) of the customers and highlights the most common pain‑points they mention.

2. Business Value

• Customer Insight – Quickly see how customers feel about the product.

• Prioritization – Identify the most frequent problems so the product team can focus on fixing what matters most.

• Efficient Reporting – Provides a ready‑to‑share summary for product managers, marketers, and leadership without manual analysis.

3. Operational Context

• When to run – After a product release, a marketing campaign, or any period when a batch of customer feedback has been collected.

• Who uses it – Product managers, product owners, and marketing analysts who need a quick, data‑driven view of customer sentiment.

• How often – As often as new review data is available (e.g., weekly, after each launch, or on a scheduled cadence).

4. Inputs

4.1 List of Customer Reviews

Only the above two fields are required for each review.

Note: If any review is missing its text, that review will be excluded from analysis and flagged for manual review (see Section 8).

5. Outputs

5.1 Sentiment Summary Report

Formatting Rules

• Use plain, concise language.

• Use bullet points for lists.

• Keep each item on its own line.

• Do not generate any new identifiers (e.g., “ID‑001”).

6. Detailed Plan & Execution Steps

1. Verify Input

   • Confirm the list of reviews contains at least one entry.

   • If any entry is missing the Review Text, record the Review ID in a “Missing Text” list and remove that entry from further processing.

2. Assign Sentiment to Each Review

   • Scan the Review Text for words and phrases from the Sentiment Keyword List (Appendix C).

   • If more positive words than negative words are found, label the review Positive.

   • If more negative words than positive words are found, label it Negative.

   • If the count is equal or no clear words are found, label it Neutral.

3. Extract Potential Pain Points

   • Search the Review Text for any phrase that matches a Pain‑Point Keyword (e.g., “slow”, “crash”, “missing”, “confusing”, “hard to use”).

   • Record each distinct phrase as a candidate pain‑point.

4. Aggregate Pain‑Points

   • Normalize similar phrases (e.g., “slow loading” and “slow performance” become “Performance”) using the Pain‑Point Normalization Table (Appendix C).

   • Count how many times each normalized pain‑point appears across all reviews.

5. Rank Pain Points

   • Sort the aggregated pain‑points by frequency, descending.

   • Take the top‑5 entries (or fewer if less than five exist).

6. Calculate Overall Sentiment

   • Count the number of Positive, Negative, and Neutral reviews.

   • Determine the majority label:

     • If Positive > Negative and Positive > Neutral → Overall Sentiment = Positive

     • If Negative > Positive and Negative > Neutral → Overall Sentiment = Negative

     • Otherwise → Overall Sentiment = Mixed.

7. Compose the Summary Report

   • Write Overall Sentiment as a single line.

   • Provide Sentiment Breakdown as a bullet list (e.g., “Positive: 7 (58 %)”).

   • List Top Pain Points in ranked order with count (e.g., “1. Slow performance – 5 mentions”).

   • Add a Notes section with any anomalies (e.g., number of excluded reviews).

8. Produce Output

   • Assemble the report exactly as described in Section 5.1.

   • Return the Sentiment Summary Report as plain text (no files, no IDs).

7. Validation & Quality Checks

• Input Presence – Ensure at least one review with non‑empty text is present; otherwise, abort and flag for manual review.

• Sentiment Consistency – Re‑run the word‑count check for each processed review to verify that the assigned label matches the keyword count.

• Pain‑Point Normalization – Verify that each extracted phrase matches a term in the Pain‑Point Keyword List; if not, flag as “Unrecognized” and list in notes.

• Report Completeness – Confirm that the report includes all four components (Overall Sentiment, Sentiment Breakdown, Top Pain Points, Notes).

• Count Accuracy – Verify that the sum of Positive, Neutral, and Negative counts equals the total number of processed reviews.

8. Special Rules / Edge Cases

• Missing Review Text – Exclude the review, note the Review ID in “Notes”, and continue processing remaining reviews.

• Tie for Top Pain Points – If more than five pain‑points share the same frequency and exceed the limit, include all tied items up to the limit of five; if the limit would be exceeded, list the first five alphabetically.

• All Reviews Neutral – Set Overall Sentiment = Mixed and note “All reviews neutral”.

• No Pain Points Identified – Report “No significant pain‑points detected” in the Top Pain Points section.

• Ambiguous Sentiment – If a review contains equal numbers of positive and negative keywords, label it Neutral.

9. Example

Input (list of 3 reviews)

Output (Sentiment Summary Report)

• Overall Sentiment: Mixed

• Sentiment Breakdown

  • Positive: 2 (66 %)

  • Negative: 1 (33 %)

  • Neutral: 0 (0 %)

• Top Pain Points

  1. Slow loading – 1 mention

  2. Frequent crashes – 1 mention

• Notes

  • No missing review text.

  • All three reviews processed successfully.

Appendix A – FAQ

1. Can this process handle reviews in other languages?

   • The current implementation only looks for English keywords. Non‑English reviews will be treated as Neutral unless they contain English keywords.

2. What if a review contains multiple pain‑points?

   • All distinct pain‑points found in a single review are counted separately.

3. Do I need to pre‑process the reviews (e.g., remove HTML tags)?

   • Yes. Reviews should be plain text without markup.

4. What if the sentiment keywords list is too short?

   • The process will still work but may miss some sentiment cues. You can add more keywords in the Sentiment Keyword List (Appendix C).

5. How are synonyms handled?

   • The Pain‑Point Normalization Table (Appendix C) maps synonyms to a common term (e.g., “slow”, “sluggish”, “slow‑loading” → “Slow performance”).

Appendix B – Glossary

Appendix C – Reference Materials

C.1 Sentiment Keyword List

Positive Keywords (any occurrence counts as a positive cue)

• amazing, awesome, excellent, great, good, fantastic, love, nice, wonderful, perfect, superb, wonderful, smooth, fast, quick, intuitive, easy, helpful, reliable, stable, intuitive, happy, satisfied, impressive, reliable, works, smooth, seamless, fast, responsive, excellent, perfect, love, great, amazing, fantastic.

Negative Keywords (any occurrence counts as a negative cue)

• bad, terrible, awful, poor, slow, lag, lagging, delay, delayed, glitch, bug, error, failure, fail, crash, crashes, broken, confusing, difficult, hard, impossible, messy, problematic, frustrating, unsatisfied, unsatisfied, unsatisfying, terrible, awful, terrible, problematic, bug, issue, problem, flaw, mistake, complaint.

Neutral/None – Any word not listed above is ignored for sentiment scoring.

C.2 Pain‑Point Keyword List

C.3 Pain‑Point Normalization Table

C.4 Example Sentiment & Pain‑Point Mapping

Additional Notes

• Ensure that the Sentiment Keyword List and Pain‑Point Keyword List are kept up‑to‑date as product language and customer terminology evolve.

• For best results, standardize the format of the review text before providing it to the process (e.g., remove HTML tags, convert smart quotes to standard quotes).

**