app.py

# AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/02_app_gradio.ipynb.

# %% auto 0
__all__ = ['categories', 'k', 'min_words', 'max_words', 'ignore_text', 'ignore_common', 'learn', 'text', 'label', 'examples',
           'intf', 'predict']

# %% ../nbs/02_app_gradio.ipynb 4
import warnings
warnings.filterwarnings('ignore')
from fastai.text.all import *
import gradio as gr
import requests
from bs4 import BeautifulSoup
import enchant
import re
import random
from collections import Counter
import hashlib
import pickle
from wordcloud import WordCloud


# %% ../nbs/01_data.ipynb 8
class Webpage:
    def __init__(self, url):
        self.url = url
        self.hash = self.get_hash_str()
        self.requested = False
        self.page_text = ""
        self.html = ""
        self.links = []
        self.text = []
        self.cleaned_text = []
        self.most_common_words = []
    
    def get_page(self, headers, min_size, max_size):
        r = requests.get(self.url, stream=True, headers=headers)
        content_length = int(r.headers.get('Content-Length', 0))
        data = []
        length = 0

        if content_length > max_size:
            return None

        for chunk in r.iter_content(1024):
            data.append(chunk)
            length += len(chunk)
            if length > max_size:
                return None
        r._content = b''.join(data)
        if len(r.text) < min_size: return None
        return r.text

    def get_page_html(self, min_size=1000, max_size=2000000):
        user_agents = [ 
            'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36', 
            'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/92.0.4515.107 Safari/537.36', 
            'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/90.0.4430.212 Safari/537.36', 
            'Mozilla/5.0 (iPhone; CPU iPhone OS 12_2 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Mobile/15E148', 
            'Mozilla/5.0 (Linux; Android 11; SM-G960U) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/89.0.4389.72 Mobile Safari/537.36' 
        ] 
        user_agent = random.choice(user_agents) 
        headers = {'User-Agent': user_agent} 
        self.page_text = self.get_page(headers, min_size, max_size)
        self.html = BeautifulSoup(self.page_text, "html.parser")
        self.requested = True

    def get_hash_str(self, inp=""):
        return hashlib.sha3_256((self.url+inp).encode()).hexdigest()

    def get_html_anchors(self, keyword="http"):
        for anchor in self.html.findAll('a'):
            link = anchor.get('href')
            if link == None or link == "":
                continue
            if keyword in link:
                self.links.append(link)
                
    def get_html_text(self, tags=["p"]):
        for tag in tags:
            for p in self.html.findAll(tag):
                p_text = p.getText().strip()
                if p_text == None or p_text == '':
                    continue
                self.text.append(p_text)

    def clean_html_text(self, max_words, enchant_dict="en_US", ignore=[], rx="[^a-zA-Z ]+", min_word_len=2):
        all_text = ' '.join(self.text).lower()
        regex_text = re.sub(rx,'',all_text).strip()
        split = regex_text.split()
        split = [word for word in split if word not in ignore]
        if enchant_dict != "": d = enchant.Dict(enchant_dict)
        for word in split:
            if len(self.cleaned_text) >= max_words: break
            if len(word) >= min_word_len:
                if enchant_dict == "":
                    self.cleaned_text.append(word)
                elif d.check(word): 
                    self.cleaned_text.append(word)

    def k_common_words(self, k=10, ignore=[]):
        if self.cleaned_text == "":
            text = self.text
        else:
            text = self.cleaned_text
        all_text = ' '.join(text).lower()
        split = all_text.split()
        split_ignore = [word for word in split if word not in ignore]
        counts = Counter(split_ignore)
        k_most_common = counts.most_common(k)
        self.most_common_words = k_most_common

    def save_text(self, path, fname):
        file = open(path+fname, 'wb')
        pickle.dump(self.text, file)
        file.close()

    def load_text(self, path, fname):
        file = open(path+fname, 'rb')
        self.text = pickle.load(file)
        file.close()

    def save_links(self, path, fname):
        file = open(path+fname, 'wb')
        pickle.dump(self.links, file)
        file.close()

    def load_links(self, path, fname):
        file = open(path+fname, 'rb')
        self.links = pickle.load(file)
        file.close()

# %% ../nbs/01_data.ipynb 14
def get_page_all(url, k, max_words, ignore_text, ignore_common, path = None):
    page = Webpage(url)
    fname_text = page.hash+'.text'
    fname_links = page.hash+'.links'
    if path == None:
        page.get_page_html()
        page.get_html_text(tags=["p","h1","h2","h3","span"])
        page.get_html_anchors()
    else:
        if os.path.isfile(path+fname_text): 
            page.load_text(path, fname_text)
        else:
            page.get_page_html()
            page.get_html_text(tags=["p","h1","h2","h3","span"])
            page.save_text(path, fname_text)

        if os.path.isfile(path+fname_links): 
            page.load_links(path, fname_links)
        else:
            if page.html == "": page.get_page_html()
            page.get_html_anchors()
            page.save_links(path, fname_links)

    if page.text is not None:
        page.clean_html_text(max_words, ignore=ignore_text, rx="[^a-zA-Z ]+")
        page.k_common_words(k=k, ignore=ignore_common)
    return page

# %% ../nbs/02_app_gradio.ipynb 6
categories = ('pseudoscience','science')
k = 30
min_words = 20
max_words = 450
ignore_text = ['the', 'of', 'to', 'and', 'a', 'in', 'it', 'that', 'for', 'on'] 
ignore_common = ignore_text
learn = load_learner('model.pkl', cpu=True)

def predict(url):
    page = get_page_all(url, k, max_words, ignore_text, ignore_common)
    length = len(page.cleaned_text)
    if  length < min_words:
        return "ERROR: Returned "+str(length)+" words"
    else:
        text = ' '.join(page.cleaned_text)
        with learn.no_bar(), learn.no_logging():
            pred,idx,probs = learn.predict(text)
        wordcloud = WordCloud(width = 800, height = 800,
                background_color ='white',
                min_font_size = 10).generate(text)
                
        # plot the WordCloud image                      
        fig = plt.figure(figsize = (8, 8), facecolor = None)
        plt.imshow(wordcloud)
        plt.axis("off")
        plt.tight_layout(pad = 0)
        return (dict(zip(categories, map(float,probs))), fig)

# %% ../nbs/02_app_gradio.ipynb 8
examples = ['https://www.theskepticsguide.org/about','https://www.foxnews.com/opinion']

pseudo_sources = ["http://www.ageofautism.com/",
 "http://www.naturalnews.com", 
 "https://foodbabe.com/starthere/",
 "http://www.chopra.com",
 "https://www.mercola.com/",
 "https://www.history.com/",
 "https://doctoroz.com/",
 "https://www.disclose.tv/",
 "https://nationalreport.net/",
 "https://heartland.org/",
 "https://www.dailymail.co.uk/",
 "https://www.motherjones.com/"]

science_sources = ["https://sciencebasedmedicine.org/",
 "https://www.hopkinsmedicine.org/gim/research/method/ebm.html",
 "https://www.bbc.com/news/science_and_environment",
 "https://www.nature.com/",
 "https://www.science.org/",
 "https://www.snopes.com/top/",
 "https://quackwatch.org/",
 "https://www.skepdic.com/",
 "http://scibabe.com/",
 "http://pandasthumb.org/",
 "https://skepticalscience.com/",
 "https://www.cdc.gov/",
 "https://apnews.com/"]

with gr.Blocks() as blocks:
    gr.Markdown("# Pseudometer")
    gr.Markdown("Prototype machine learning pseudoscience detector for websites!")
    text = gr.Textbox(label="Input URL (http format):")
    label = gr.outputs.Label()
    btn = gr.Button("Analyze!")
    with gr.Accordion("Pseudoscience Primary Training Sources"):
        gr.Markdown(', '.join(pseudo_sources))
    with gr.Accordion("Science Primary Training Sources"):
        gr.Markdown(', '.join(science_sources))
    example = gr.Examples(examples=examples, inputs=text)

    btn.click(fn=predict, inputs=text, outputs=[label, gr.Plot(label="Wordcloud")])
    
blocks.launch()