Botator/code/toxicity.py

from googleapiclient import discovery
from config import perspective_api_key
import re

toxicity_names = [
    "toxicity",
    "severe_toxicity",
    "identity_attack",
    "insult",
    "profanity",
    "threat",
    "sexually_explicit",
    "flirtation",
    "obscene",
    "spam",
]
toxicity_definitions = [
    "A rude, disrespectful, or unreasonable message that is likely to make people leave a discussion.",
    "A very hateful, aggressive, disrespectful message or otherwise very likely to make a user leave a discussion or give up on sharing their perspective. This attribute is much less sensitive to more mild forms of toxicity, such as messages that include positive uses of curse words.",
    "Negative or hateful messages targeting someone because of their identity.",
    "Insulting, inflammatory, or negative messages towards a person or a group of people.",
    "Swear words, curse words, or other obscene or profane language.",
    "Describes an intention to inflict pain, injury, or violence against an individual or group.",
    "Contains references to sexual acts, body parts, or other lewd content. \n **English only**",
    "Pickup lines, complimenting appearance, subtle sexual innuendos, etc. \n **English only**",
    "Obscene or vulgar language such as cursing. \n **English only**",
    "Irrelevant and unsolicited commercial content. \n **English only**",
]


client = discovery.build(
    "commentanalyzer",
    "v1alpha1",
    developerKey=perspective_api_key,
    discoveryServiceUrl="https://commentanalyzer.googleapis.com/$discovery/rest?version=v1alpha1",
    static_discovery=False,
)

analyze_request = {
    "comment": {"text": ""},  # The text to analyze
    # we will ask the following attributes to google: TOXICITY, SEVERE_TOXICITY, IDENTITY_ATTACK, INSULT, PRPFANITY, THREAT, SEXUALLY_EXPLICIT, FLIRTATION, OBSCENE, SPAM
    "requestedAttributes": {
        "TOXICITY": {},
        "SEVERE_TOXICITY": {},
        "IDENTITY_ATTACK": {},
        "INSULT": {},
        "PROFANITY": {},
        "THREAT": {},
        "SEXUALLY_EXPLICIT": {},
        "FLIRTATION": {},
        "OBSCENE": {},
        "SPAM": {},
    },
    # we will analyze the text in any language automatically detected by google
    "languages": [],
    "doNotStore": "true",  # We don't want google to store the data because of privacy reasons & the GDPR (General Data Protection Regulation, an EU law that protects the privacy of EU citizens and residents for data privacy and security purposes https://gdpr-info.eu/)
}
analyze_request_not_en = {
    "comment": {"text": ""},  # The text to analyze
    # we will ask the following attributes to google: TOXICITY, SEVERE_TOXICITY, IDENTITY_ATTACK, INSULT, PRPFANITY, THREAT, SEXUALLY_EXPLICIT, FLIRTATION, OBSCENE, SPAM
    "requestedAttributes": {
        "TOXICITY": {},
        "SEVERE_TOXICITY": {},
        "IDENTITY_ATTACK": {},
        "INSULT": {},
        "PROFANITY": {},
        "THREAT": {},
    },
    # we will analyze the text in any language automatically detected by google
    "languages": [],
    "doNotStore": "true",  # We don't want google to store the data because of privacy reasons & the GDPR (General Data Protection Regulation, an EU law that protects the privacy of EU citizens and residents for data privacy and security purposes https://gdpr-info.eu/)
}


def get_toxicity(message: str):
    # we first remove all kind of markdown from the message to avoid exploits
    message = re.sub(r"\*([^*]+)\*", r"\1", message)
    message = re.sub(r"\_([^_]+)\_", r"\1", message)
    message = re.sub(r"\*\*([^*]+)\*\*", r"\1", message)
    message = re.sub(r"\_\_([^_]+)\_\_", r"\1", message)
    message = re.sub(r"\|\|([^|]+)\|\|", r"\1", message)
    message = re.sub(r"\~([^~]+)\~", r"\1", message)
    message = re.sub(r"\~\~([^~]+)\~\~", r"\1", message)
    message = re.sub(r"\`([^`]+)\`", r"\1", message)
    message = re.sub(r"\`\`\`([^`]+)\`\`\`", r"\1", message)

    # we try doing the request in english, but if we get 'errorType': 'LANGUAGE_NOT_SUPPORTED_BY_ATTRIBUTE' we try again with the analyze_request_not_en
    try:
        analyze_request["comment"]["text"] = message
        response = client.comments().analyze(body=analyze_request).execute()
    except:
        analyze_request_not_en["comment"]["text"] = message
        response = client.comments().analyze(body=analyze_request_not_en).execute()
    try:
        return [
            float(response["attributeScores"]["TOXICITY"]["summaryScore"]["value"]),
            float(
                response["attributeScores"]["SEVERE_TOXICITY"]["summaryScore"]["value"]
            ),
            float(
                response["attributeScores"]["IDENTITY_ATTACK"]["summaryScore"]["value"]
            ),
            float(response["attributeScores"]["INSULT"]["summaryScore"]["value"]),
            float(response["attributeScores"]["PROFANITY"]["summaryScore"]["value"]),
            float(response["attributeScores"]["THREAT"]["summaryScore"]["value"]),
            float(
                response["attributeScores"]["SEXUALLY_EXPLICIT"]["summaryScore"][
                    "value"
                ]
            ),
            float(response["attributeScores"]["FLIRTATION"]["summaryScore"]["value"]),
            float(response["attributeScores"]["OBSCENE"]["summaryScore"]["value"]),
            float(response["attributeScores"]["SPAM"]["summaryScore"]["value"]),
        ]
    except:
        return [
            float(response["attributeScores"]["TOXICITY"]["summaryScore"]["value"]),
            float(
                response["attributeScores"]["SEVERE_TOXICITY"]["summaryScore"]["value"]
            ),
            float(
                response["attributeScores"]["IDENTITY_ATTACK"]["summaryScore"]["value"]
            ),
            float(response["attributeScores"]["INSULT"]["summaryScore"]["value"]),
            float(response["attributeScores"]["PROFANITY"]["summaryScore"]["value"]),
            float(response["attributeScores"]["THREAT"]["summaryScore"]["value"]),
        ]