Using Gemini in BigQuery for sentiment analysis

CREATE OR REPLACE TABLE `[PROJECT_ID.DATASET_ID].review_sentiment_analysis` AS <br /> WITH PROMPT AS ( <br /> SELECT CONCAT (‘For the given review classify the sentiment as Positive, Neutral or Negative.’, <br /> <br /> <br /> ‘\n input: The driver was able to make some small talk, but he didn\’t go overboard. I liked that he was friendly and chatty, but he also knew when to leave me alone. The trunk fit my belongings, and the car was clean and comfortable. Overall, it was a good ride.’, <br /> <br /> <br /> ‘\n output: \n Positive – Trunk fit my belongings, friendly, chatty’, <br /> <br /> <br /> ‘\n input: I took a rideshare last night and it was an okay experience. The car was adequately clean, but it was a bit warm for my liking. The driver was able to make some small talk, but I wasn\’t really in the mood to talk. Overall, it was a fine ride.’, <br /> <br /> <br /> ‘\n output: Neutral – Clean, A bit warm, fun ride’, <br /> ‘\n input: ‘, customer_review_text, <br /> ‘\n output: ‘ <br /> ) AS prompt, customer_id, <br /> customer_review_text <br /> FROM `[PROJECT_ID.DATASET_ID].customer_review` <br /> LIMIT 20 <br /> ), <br /> REVIEW_RESPONSE_GENERATION AS ( <br /> SELECT * <br /> FROM <br /> ML.GENERATE_TEXT( <br /> MODEL `[PROJECT_ID.DATASET_ID.MODEL_NAME]`, <br /> (SELECT * FROM PROMPT), <br /> STRUCT( <br /> 200 AS max_output_tokens, <br /> 0.5 AS temperature, <br /> 40 AS top_k, <br /> 1.0 AS top_p, <br /> TRUE AS flatten_json_output)) <br /> ) <br /> SELECT ml_generate_text_llm_result, customer_id, customer_review_text, prompt, ml_generate_text_status FROM REVIEW_RESPONSE_GENERATION; <br /> <br /> <br /> SELECT ml_generate_text_llm_result, customer_id, customer_review_text, prompt, ml_generate_text_status FROM `[PROJECT_ID.DATASET_ID].review_sentiment_analysis`;