app.py

import spacy
import streamlit as st
import re 
import logging
from presidio_anonymizer import AnonymizerEngine
from presidio_analyzer import AnalyzerEngine, PatternRecognizer, RecognizerResult, EntityRecognizer

from annotated_text import annotated_text
from flair_recognizer import FlairRecognizer
from detoxify import Detoxify

###############################
#### Render Streamlit page ####
###############################

st.title("Anonymise your text!")
st.markdown(
    "This mini-app anonymises text using Flair and Presidio. You can find the code in the Files and Versions tabs in the [HuggingFace page](https://huggingface.co/spaces/arogeriogel/anonymise_this)"
)

# Configure logger
logging.basicConfig(format="\n%(asctime)s\n%(message)s", level=logging.INFO, force=True)

##############################
###### Define functions ######
##############################

@st.cache_resource(show_spinner="Fetching model from cache...")
def analyzer_engine():
    """Return AnalyzerEngine."""
    analyzer = AnalyzerEngine()
    flair_recognizer = FlairRecognizer()
    analyzer.registry.add_recognizer(flair_recognizer)

    return analyzer

def analyze(**kwargs):
    """Analyze input using Analyzer engine and input arguments (kwargs)."""

    if "entities" not in kwargs or "All" in kwargs["entities"]:
        kwargs["entities"] = None

    results =  analyzer_engine().analyze(**kwargs)
    st.session_state.analyze_results = results

def annotate():
    text = st.session_state.text 
    analyze_results = st.session_state.analyze_results
    tokens = []
    starts=[]
    # sort by start index
    results = sorted(analyze_results, key=lambda x: x.start)
    for i, res in enumerate(results):
        # if we already have an entity for this token don't add another
        if res.start not in starts:
            if i == 0:
                tokens.append(text[:res.start])

            # append entity text and entity type
            tokens.append((text[res.start: res.end], res.entity_type))

            # if another entity coming i.e. we're not at the last results element, add text up to next entity
            if i != len(results) - 1:
                tokens.append(text[res.end:results[i+1].start])
            # if no more entities coming, add all remaining text
            else:
                tokens.append(text[res.end:])

            # append this token to the list so we don't repeat results per token
            starts.append(res.start)
    return tokens

def get_supported_entities():
    """Return supported entities from the Analyzer Engine."""
    return analyzer_engine().get_supported_entities()

def analyze_text():
    if not st.session_state.text:
        st.session_state.text_error = "Please enter your text"
        return
    
    toxicity_results = Detoxify('original').predict(st.session_state.text)
    is_toxic=False
    for k in toxicity_results.keys():
        for k in toxicity_results.keys():
            if k!='toxicity':
                if toxicity_results[k]>0.5:
                    is_toxic=True
            else:
                if toxicity_results[k]>0.65:
                    is_toxic=True
    if is_toxic:
        st.session_state.text_error = "Your text entry was detected as toxic, please re-write it."
        return
    else:
        with text_spinner_placeholder:
            with st.spinner("Please wait while your text is being analysed..."):
                logging.info(f"This is the text being analysed: {st.session_state.text}")
                st.session_state.text_error = ""
                st.session_state.n_requests += 1
                analyze(
                    text=st.session_state.text,
                    entities=st_entities,
                    language="en",
                    return_decision_process=False,
                )

                if st.session_state.excluded_words:
                    exclude_manual_input()

                if st.session_state.allowed_words:
                    allow_manual_input()
            
                logging.info(
                    f"analyse results: {st.session_state.analyze_results}\n"
                )


def exclude_manual_input():
    deny_list = [i.strip() for i in st.session_state.excluded_words.split(',')]
    def _deny_list_to_regex(deny_list):
        """
        Convert a list of words to a matching regex.
        To be analyzed by the analyze method as any other regex patterns.
        :param deny_list: the list of words to detect
        :return:the regex of the words for detection
        """
        # Escape deny list elements as preparation for regex
        escaped_deny_list = [re.escape(element) for element in deny_list]
        regex = r"(?:^|(?<=\W))(" + "|".join(escaped_deny_list) + r")(?:(?=\W)|$)"
        return regex

    deny_list_pattern = _deny_list_to_regex(deny_list)
    matches = re.finditer(deny_list_pattern, st.session_state.text)
    results = []
    for match in matches:
        start, end = match.span()
        current_match = st.session_state.text[start:end]

        # Skip empty results
        if current_match == "":
            continue
            
        pattern_result = RecognizerResult(
                    entity_type='MANUALLY ADDED',
                    start=start,
                    end=end,
                    score=1.0,
                )
        
        # check if already in detected strings
        found=False
        for token in st.session_state.analyze_results:
            if token.start==start and token.end==end:
                found=True
        if found==False:
            results.append(pattern_result)

    results = EntityRecognizer.remove_duplicates(results)    
    
    st.session_state.analyze_results.extend(results)

    logging.info(
                f"analyse results after adding excluded words: {st.session_state.analyze_results}\n"
            )
   
def allow_manual_input():
    analyze_results_fltered=[]

    for token in st.session_state.analyze_results:
        if st.session_state.text[token.start:token.end] not in st.session_state.allowed_words:
            analyze_results_fltered.append(token)
    logging.info(
                f"analyse results after removing allowed words: {analyze_results_fltered}\n"
            )
    st.session_state.analyze_results = analyze_results_fltered

@st.cache_resource(show_spinner="Fetching model from cache...")
def anonymizer_engine():
    """Return AnonymizerEngine."""
    return AnonymizerEngine()

def anonymise_text():
    if st.session_state.n_requests >= 50:
        st.session_state.text_error = "Too many requests. Please wait a few seconds before anonymising more text."
        logging.info(f"Session request limit reached: {st.session_state.n_requests}")
        st.session_state.n_requests = 1

    st.session_state.text_error = ""

    if not st.session_state.text:
        st.session_state.text_error = "Please enter your text"
        return

    if not st.session_state.analyze_results:
        analyze_text()

    with text_spinner_placeholder:
        with st.spinner("Please wait while your text is being anonymised..."):
            anon_results = anonymizer_engine().anonymize(st.session_state.text, st.session_state.analyze_results)
            st.session_state.text_error = ""
            st.session_state.n_requests += 1
            st.session_state.anon_results = anon_results
            logging.info(
                f"text anonymised: {st.session_state.anon_results}"
            )

def clear_results():
    st.session_state.anon_results=""
    st.session_state.analyze_results=""

#######################################
#### Initialize "global" variables ####
#######################################

if "text_error" not in st.session_state:
    st.session_state.text_error = ""
if "analyze_results" not in st.session_state:
    st.session_state.analyze_results = ""
if "anon_results" not in st.session_state:
    st.session_state.anon_results = ""
if "n_requests" not in st.session_state:
    st.session_state.n_requests = 0

##############################
####### Page arguments #######
##############################

# Every widget with a key is automatically added to Session State as a global variable.

# In Streamlit, interacting with a widget triggers a rerun and variables defined
# in the code get reinitialized after each rerun.

# If a callback function is associated with a widget then a change in the widget 
# triggers the following sequence: First the callback function is executed and then
# the app executes from top to bottom.

st.text_input(
    label="Text", 
    placeholder="Write your text here", 
    key='text',
    on_change=clear_results
)
st.text_input(
    label="Data to be redacted (optional)",
    placeholder="John, Mary, London",
    key='excluded_words',
    on_change=clear_results
)
st.text_input(
    label="Data to be ignored (optional)",
    placeholder="NHS, GEL, Lab",
    key='allowed_words',
    on_change=clear_results
)

st_entities = st.sidebar.multiselect(
    label="Which entities to look for?",
    options=get_supported_entities(),
    default=list(get_supported_entities()),
)

##############################
######## Page buttons ########
##############################

# button return true when clicked

col1, col2 = st.columns(2)

analyze_now=False
with col1:
    analyze_now = st.button(
        label="Analyse text",
        type="primary",
        on_click=analyze_text,
    )
anonymise_now=False
with col2:
    anonymise_now = st.button(
        label="Anonymise text",
        type="primary",
        on_click=anonymise_text,
    )

##############################
######## Page actions ########
##############################

text_spinner_placeholder = st.empty()
if st.session_state.text_error:
    st.error(st.session_state.text_error)

with col1:
    if st.session_state.analyze_results:
        annotated_tokens=annotate()
        annotated_text(*annotated_tokens)
        st.write(st.session_state.analyze_results)
    if not st.session_state.analyze_results and analyze_now and not st.session_state.text_error:
        st.write("### No PII was found. ###")

with col2:
    if st.session_state.anon_results:
        st.write(st.session_state.anon_results.text)
    if not st.session_state.analyze_results and anonymise_now and not st.session_state.text_error:
        st.write("### No PII was found. ###")