Nozzle BigQuery ML Sentiment

WITH

current_paas AS (
    SELECT
    DISTINCT(paa_text)
    FROM `project.dataset.paa_text_table`  -- edit this

),

nozzle_paa_results AS (
    SELECT
        DISTINCT(related_phrase) AS paa_text
        FROM `nozzle.client_ranking_data.locomotive_locomotive`  -- edit this
        WHERE result_type = 'PAA'
        AND related_phrase IS NOT NULL
        AND keyword_group = "brand" -- edit this
        AND related_phrase NOT IN (SELECT paa_text FROM current_paas)
        AND date(requested) > DATE_SUB(current_date(), INTERVAL 150 DAY)
        ORDER BY related_phrase
        LIMIT 50

),

predictions AS (
    SELECT
    paa_text,
    ROUND(classifier[OFFSET(1)], 3) as score
    FROM ML.PREDICT(MODEL `project.bigquery_ml.sentiment`,(SELECT REGEXP_REPLACE(paa_text, '[^\\w\\s]+', '') AS text FROM nozzle_paa_results))
    JOIN (SELECT paa_text, REGEXP_REPLACE(paa_text, '[^\\w\\s]+', '') AS text FROM nozzle_paa_results) USING (text)
    ORDER BY score ASC


)

SELECT * FROM predictions;