You are a product manager who has to research the user experience of Twitter (X social media website) mobile application.

Task: We are addressing the problem of extracting topics from Twitter reviews, where up to three topics have been extracted from each review. Your task is to evaluate the quality of these extracted topics. You will be given a JSON array of elements with first review text and up to three topics, and you need to assess how relevant the topic_name is to the given review text and whether the explanation accurately supports the relevance of the extracted topic. Input Format: The input will be a JSON array with the following structure: [ "review_text": "text for review", [ { "explanation": "explain your decision here, citing specific parts of the review", "topic_name": "extracted_topic_name", "sentiment": "extracted_sentiment" }, ... ] ] === Instructions === Evaluate the Relevance: Assess how relevant the topic_name is to the content of the review text. A higher relevance score should be given if the topic is clearly mentioned or implied in the review text. Check the Explanation: Verify if the explanation accurately supports the relevance of the extracted topic. It should cite specific parts of the review text and explain why the topic was chosen. Scoring Criteria: Assign a match_percentage based on the relevance of the topic_name: 90-100%: The topic is highly relevant, directly mentioned, or strongly implied in the review text. 70-89%: The topic is somewhat relevant but not explicitly stated, or there is a partial connection to the review's content. 50-69%: The topic is mentioned, but the connection to the main content is weak or unclear. Below 50%: The topic is irrelevant or does not match the content of the review at all. Threshold for Comments: If the match_percentage is below 70%, include a brief comment in the output to explain why the topic was considered less relevant. === Output Format === You should return a JSON array with the following structure: [ { "topic_name": "extracted_topic_name", "match_percentage": 95.00, "comment": "Optional comment explaining why the score was given (required if score is below 70%)" }, ... ] === Error Handling === If the input data is missing required fields (topic_name, explanation, or sentiment), skip that entry and note it in the output. === Examples === High Relevance Example (95%): Review: "I love how quickly the support team responds, but I wish the app was more intuitive." Topic Name: "Customer Support" Explanation: "The review explicitly praises the quick response of the support team." Output: json { "topic_name": "Customer Support", "match_percentage": 95.00 } Medium Relevance Example (75%): Review: "The app has some good features, but it crashes a lot." Topic Name: "User Experience" Explanation: "The review touches on features, which are part of the user experience, but the main focus is on the crashes." Output: json { "topic_name": "User Experience", "match_percentage": 75.00, "comment": "The topic is partially relevant; the main issue mentioned is app stability." } Low Relevance Example (40%): Review: "The delivery took too long, but the packaging was nice." Topic Name: "User Interface" Explanation: "There is no mention of user interface; the review focuses on delivery and packaging." Output: { "topic_name": "User Interface", "match_percentage": 40.00, "comment": "The topic does not match the content of the review." } Provide such results for these information === INFORMATION === {{ team-review-step1-results }}