Hate Speech API

by Pau Escofet

Overview

Hate speech has long been a problem, but it has grown with the emergence of digital platforms. Many companies and social networks struggle to detect possible cases of hate speech in different texts, which is the objective of this API. I have designed and trained a Machine Learning model capable of detecting whether a sentence contains hate speech.

The dataset with which the predictor has been tested has been successful in 93% of the cases.

This is the first version of the API, and doesn't need authentication, it's free to use

`BASE URL: https://hatespeech.pauescofet.com`

Endpoints

Info

Languages

`GET /languages`

Info about the languages that are currently available

Body

# no body required

Response

[

                   {

                      "code": "en",

                      "language": "english",

                      "status": "active"

                   },

                   {

                      "code": "es",

                      "language": "español",

                      "status": "disabled"

                   }

              ]

Predict

Predict English

`POST /hatespeech/en`

Endpoint to predict a maximum of 10 english quotes at a time

Body

{

                   data: [

                    "quote to be predicted 1",

                    "quote to be predicted 2",

                    " ... ",

                   ] 

                }

Response

# array with the predicted probability of each quote

prob: []

# array with the predicted category in text

text: []

# array with the predicted category in numbers

values: []

}

Predict Spanish

Currently no available

Examples in Python


              import requests 

              import json

# language endpoint


              response = requests.get('https://hatespeech.pauescofet.com/languages') 

              print(response.status_code)

# out: 200


              res_json = json.loads(response.text) 

              print(res_json)

# out:

[
  {
      "code": "en",
      "language": "english",
      "status": "active"
  },
  {
      "code": "es",
      "language": "español",
      "status": "disabled"
  }
]

# predict endpoint


              data = { 
                
 'data': [

                  "I think you are cool",

                  "I am going to kill you",

                  "You are a bitch",

                  "I hate you",

                  "Maybe we could go for a drink some day" 

                 ] 

              } 

              response = requests.post(url = 'https://hatespeech.pauescofet.com/hatespeech/en', json = data)
              print(response.status_code)

# out: 200


              res_json = json.loads(response.text) 

              print(res_json)

# out: