app.py

import gradio as gr
import re
import json
import numpy as np
import nltk
import stanza
from stanza.models.constituency.parse_tree import Tree
from transformers import AutoTokenizer, AutoModelForTokenClassification, TokenClassificationPipeline
from sentence_transformers import CrossEncoder
from autocorrect import Speller
from transformers import BertTokenizer, BertForSequenceClassification
import torch
from torch.nn.utils.rnn import pad_sequence
from openai import OpenAI
from tenacity import (
    retry,
    stop_after_attempt,
    wait_random_exponential,
)  # for exponential backoff
import os

# ***************************** Load needed models *****************************
nlp = stanza.Pipeline(lang='en', processors='tokenize,pos,constituency')
pos_tokenizer = AutoTokenizer.from_pretrained("QCRI/bert-base-multilingual-cased-pos-english")
pos_model = AutoModelForTokenClassification.from_pretrained("QCRI/bert-base-multilingual-cased-pos-english")
#sentences_similarity_model = CrossEncoder('cross-encoder/stsb-roberta-base')
sentences_similarity_model = CrossEncoder('WillHeld/roberta-base-stsb')
nli_model = BertForSequenceClassification.from_pretrained("nouf-sst/bert-base-MultiNLI", use_auth_token="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")
nli_tokenizer = BertTokenizer.from_pretrained("nouf-sst/bert-base-MultiNLI", use_auth_token="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", do_lower_case=True)


# ***************************** GPT API *****************************
client = OpenAI(
    api_key=os.getenv("OpenAI"),
)


@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
def completion_with_backoff(**kwargs):
    return client.chat.completions.create(**kwargs)

def prompt(prompt_message, bad_smell):
  message = [
        {
          "role": "system",
          "content": prompt_message
          },
        {
          "role": "user",
          "content": bad_smell
        }
  ]
  completion = completion_with_backoff(
        model="gpt-3.5-turbo",
        messages=message,
        temperature= 0.2,
      )
  return completion.choices[0].message.content
    
# ***************************** TGRL Parsing *****************************

def parse_tgrl(file_obj):

  with open(file_obj.name, 'r') as f:
    tgrl_text = f.read()
    tgrl_text = tgrl_text.replace('\t', '')
    tgrl_text = tgrl_text.replace('\n', '')
    
  return tgrl_text

def get_elements_per_actor(all_elements, tgrl_text):

  # Output format : elements_per_actor =
  #  {"actor_1": ["element_1", "element_2"],
  #   "actor_2": ["element_1", "element_2", "element_3"]}

  elements_per_actor = {}

  for goal in all_elements["goals"]:
    corresponding_actor = tgrl_text.rfind('actor', 0, tgrl_text.index(goal))
    corresponding_actor = re.split(' |{', tgrl_text[corresponding_actor:])[1]
    if corresponding_actor not in elements_per_actor:
        elements_per_actor[corresponding_actor] = []
    elements_per_actor[corresponding_actor].append(goal)

  for softGoal in all_elements["softGoals"]:
    corresponding_actor = tgrl_text.rfind('actor', 0, tgrl_text.index(softGoal))
    corresponding_actor = re.split(' |{', tgrl_text[corresponding_actor:])[1]
    if corresponding_actor not in elements_per_actor:
        elements_per_actor[corresponding_actor] = []
    elements_per_actor[corresponding_actor].append(softGoal)

  for task in all_elements["tasks"]:
    corresponding_actor = tgrl_text.rfind('actor', 0, tgrl_text.index(task))
    corresponding_actor = re.split(' |{', tgrl_text[corresponding_actor:])[1]
    if corresponding_actor not in elements_per_actor:
        elements_per_actor[corresponding_actor] = []
    elements_per_actor[corresponding_actor].append(task)

  return elements_per_actor

def get_decomposed_elements(tgrl_text):

  # Output format : decomposed_elements =
  #  {"main_element_1": ["sub_element_1", "sub_element_1"]}

  new_tgrl_text = tgrl_text
  decomposed_elements = {}

  main_elements_1 = re.findall("\w+(?=\s+decomposedBy)", new_tgrl_text)

  for main_element in main_elements_1:
      sub_element_1 = (re.findall(main_element+"\s*(?: decomposedBy )([A-Za-z\s]*)", new_tgrl_text)[0])
      sub_element_1 = sub_element_1.replace(" ", "")
      sub_element_2 = (re.findall(main_element+"\s*(?: decomposedBy )"+ sub_element_1 +",\s*([A-Za-z\s]*)", new_tgrl_text)[0])
      new_tgrl_text = new_tgrl_text.replace(main_element+" decomposedBy "+sub_element_1+", "+sub_element_2+";", '')
      decomposed_elements[main_element] = [sub_element_1, sub_element_2]

  # Replace elements IDs with names
  new_decomposed_elements = {}

  for key, _ in decomposed_elements.items():

      new_key = re.findall("(?:"+key+"\s*{\s*name\s=\s\")([A-Za-z\s]*)", tgrl_text)[0]
      new_values = []

      for element in decomposed_elements[key]:
          new_value = re.findall("(?:"+element+"\s*{\s*name\s=\s\")([A-Za-z\s;.,!?:-]*)", tgrl_text)[0]
          new_values.append(new_value)

      new_decomposed_elements[new_key] = new_values

  return new_decomposed_elements

def get_contributing_elements(tgrl_text):

  # Output format : contributing_elements_per_actor =
  #  {"actor_1": [["element_1", "element_2", "contribution_value"], ["element_1", "element_3", "contribution_value"]],
  #   "actor_2": [["element_1", "element_2", "contribution_value"]]}

  new_tgrl_text = tgrl_text
  contributing_elements = []

  main_elements_1 = re.findall("\w+(?=\s+contributesTo)", new_tgrl_text)
  for main_element in main_elements_1:
      sub_element_1 = (re.findall(main_element+"(?: contributesTo )([A-Za-z\s]*)", new_tgrl_text)[0])
      sub_element_1 = sub_element_1.replace(" ", "")
      contribution = (re.findall(main_element+"(?: contributesTo )"+ sub_element_1 +"\s{(-*[0-9A-Za-z]*)", new_tgrl_text)[0])
      new_tgrl_text = new_tgrl_text.replace(main_element+" contributesTo "+sub_element_1, '')
      contributing_elements.append([main_element, sub_element_1, contribution])

  contributing_elements_IDs_per_actor = {}
  for element in contributing_elements:
      corresponding_actor_1 = tgrl_text.rfind('actor', 0, tgrl_text.index(" "+element[0]))
      corresponding_actor_1 = re.split(' |{', tgrl_text[corresponding_actor_1:])[1]
      if corresponding_actor_1 not in contributing_elements_IDs_per_actor:
          contributing_elements_IDs_per_actor[corresponding_actor_1] = []
      contributing_elements_IDs_per_actor[corresponding_actor_1].append(element)

  # Replace elements IDs with names
  contributing_elements_per_actor = {}
  for key, values in contributing_elements_IDs_per_actor.items():
      contributing_elements = []
      for elements in contributing_elements_IDs_per_actor[key]:
          element_name_1 = re.findall("(?:"+elements[0]+"\s*{\s*name\s=\s\")([A-Za-z\s;.,!?:-]*)", tgrl_text)[0]
          element_name_2 = re.findall("(?:"+elements[1]+"\s*{\s*name\s=\s\")([A-Za-z\s;.,!?:-]*)", tgrl_text)[0]
          contributing_elements.append([element_name_1, element_name_2, elements[2]])
      contributing_elements_per_actor[key] = contributing_elements

  return contributing_elements_per_actor

def extract_elements(tgrl_text):

  # Extract actors
  actors = re.findall("(?:.*?actor\s\S+\s?{\s?name\s?=\s?\")([A-Za-z\s;.,!?:-]*)(?:\")", tgrl_text)
  # Extract goals
  goals = re.findall("(?:.*?goal\s\S+\s?{\s?name\s?=\s?\")([A-Za-z\s;.,!?:-]*)(?:\")", tgrl_text)
  # Extract softGoals
  softGoals = re.findall("(?:.*?softGoal\s\S+\s?{\s?name\s?=\s?\")([A-Za-z\s;.,!?:-]*)(?:\")", tgrl_text)
  # Extract tasks
  tasks = re.findall("(?:.*?task\s\S+\s?{\s?name\s?=\s?\")([A-Za-z\s;.,!?:-]*)(?:\")", tgrl_text)
  # Extract resources
  resources = re.findall("(?:.*?resource\s\S+\s?{\s?name\s?=\s?\")([A-Za-z\s;.,!?:-]*)(?:\")", tgrl_text)

  all_elements = {
    "actors": actors,
    "goals": goals,
    "softGoals": softGoals,
    "tasks": tasks,
    "resources": resources
  }

  ####### get elements per actor #######
  elements_per_actor = get_elements_per_actor(all_elements, tgrl_text)

  ####### get decomposed elements #######
  decomposed_elements = get_decomposed_elements(tgrl_text)

  ####### get contributing elements #######
  contributing_elements_per_actor = get_contributing_elements(tgrl_text)

  return all_elements, elements_per_actor, decomposed_elements, contributing_elements_per_actor


def get_long_elements(elements, size_threshold): # Using RegEx 

  long_elements = []

  for key, value in elements.items():
    for i in range(0, len(elements[key])):
        if len(re. findall(r'\w+', elements[key][i])) > size_threshold:
            long_elements.append(elements[key][i])

  if long_elements:
    output = ""
    for long_element in long_elements:
      refactored_element = prompt(
          '''You are a specialist in English linguistics.
          You will be provided with a sentence, and your task is to summarize it in''' + str(size_threshold) + ''' words or fewer.
          Comply with the following conditions:
          (1) Do not convert a verb phrase to a noun phrase, and vice versa.
          (2) Change as few words as possible.
          Answer with the new sentence only.''',
          long_element)
      output = output + '"' + long_element + '" should be refactored to "' + refactored_element + '"\n'
    #long_elements = "\n".join(long_elements)
    return "Lengthy elements:\n" + output
  else:
    return ""
# #####################################

# ######### Complex Sentences #########
def is_complex_sentence(sentence):

  nlp = stanza.Pipeline(lang='en', processors='tokenize,pos,constituency')
  doc = nlp(sentence)
  for sentence in doc.sentences:
      unique_constituent_labels = Tree.get_unique_constituent_labels(sentence.constituency)
      if 'SBAR' in unique_constituent_labels:
        return True
      else:
        return False

def get_complex_sentences(elements):
    
    complex_sentences = []

    for key, value in elements.items():
        for i in range(0, len(elements[key])):
            if is_complex_sentence(elements[key][i]):
                complex_sentences.append(elements[key][i])

    if complex_sentences:
        output = ""
        for complex_sentence in complex_sentences:
            refactored_element = prompt(
            '''
            You are a specialist in English linguistics.
            A complex sentence is a sentence with one independent clause and at least one dependent clause. A simple sentence has a single independent clause.
            You will be provided with a complex sentence, and your task is to make it a simple sentence.
            Do not convert a verb phrase to a noun phrase, and vice versa.
            Answer with the new sentence only.
            ''', complex_sentence)
            output = output + '"' + complex_sentence + '" should be refactored to "' + refactored_element + '"\n'
        return "Complex elements:\n" + output
    else:
        return ""

# #####################################
            
# ########## Punctuations ######### 
def get_punctuations(elements):

  punctuations = []

  for key, value in elements.items():
    for i in range(0, len(elements[key])):
      if len(re.findall("[^\s\w\d-]", elements[key][i])) > 0:
          punctuations.append(elements[key][i])

  if punctuations:
    output = ""
    for punctuation in punctuations:
      refactored_element = prompt(
        '''
        You are a specialist in English linguistics.
        You will be provided with a sentence, and your task is to remove all punctuation marks.
        Answer with the new sentence only.''', punctuation)
      output = output + '"' + punctuation + '" should be refactored to "' + refactored_element + '"\n'
    #punctuations = "\n".join(punctuations)
    return "Punctuation-marked elements:\n" + output
  else:
    return ""
# #################################

# ########## Incorrect Actor Syntax ##########
def check_verb_or_noun_phrase(sentence):

  result = prompt(
        '''
        You are a specialist in English linguistics.
        You will be provided with a sentence, and your task is to determine whether the sentence is a noun phrase or a verb phrase.
        Answer with "noun phrase" or "verb phrase" and your reasons.
        Use JSON format with keys "answer" and "reason".''', sentence)
  result = json.loads(result)
  return result["answer"]
    
# def find_non_NPs(sentences):

#   pipeline = TokenClassificationPipeline(model=pos_model, tokenizer=pos_tokenizer)

#   outputs = pipeline(sentences)

#   Non_NPs = []

#   for idx, output in enumerate(outputs):
#     if output[0]['entity'].startswith('V'):
#       Non_NPs.append(sentences[idx])
    
#   return Non_NPs

def check_actor_syntax(actors):

  incorrect_actors_syntax = []
  for actor in actors:
    result = check_verb_or_noun_phrase(actor)
    if result == "verb phrase":
      incorrect_actors_syntax.append(actor)


  if incorrect_actors_syntax:
    output = ""
    for incorrect_actor_syntax in incorrect_actors_syntax:
      refactored_element = prompt(
        '''
        You are a specialist in English linguistics.
        You will be provided with a sentence that is a verb phrase, and your task is to make it a noun pharse representing an actor.
        A noun phrase should start with a noun.
        Example of actors: System, PC User, and Privacy Officer.
        Answer with the new sentence only.''', incorrect_actor_syntax)
      output = output + '"' + incorrect_actor_syntax + '" should be refactored to "' + refactored_element + '"\n'
    #incorrect_actor_syntax = "\n".join(incorrect_actor_syntax)
    return "Incorrect actors syntax:\n" + output
  else:
    return ""
# ############################################

# ########## Incorrect Goal Syntax ###########
def check_goal_syntax(goals):

  incorrect_goals_syntax = []
  for goal in goals:
    result = check_verb_or_noun_phrase(goal)
    if result == "verb phrase":
      incorrect_goals_syntax.append(goal)

  if incorrect_goals_syntax:
    output = ""
    for incorrect_goal_syntax in incorrect_goals_syntax:
      refactored_element = prompt(
        '''
        You are a specialist in English linguistics.
        You will be provided with a sentence that is not a noun phrase, and your task is to make it a noun pharse representing a goal.
        A noun phrase should start with a noun.
        For example: high data quality, fast response time, and course registration.
        Answer with the new sentence only.''', incorrect_goal_syntax)
      output = output + '"' + incorrect_goal_syntax + '" should be refactored to "' + refactored_element + '"\n'
    #incorrect_goal_syntax = "\n".join(incorrect_goal_syntax)
    return "Incorrect goals syntax:\n" + output
  else:
    return ""
# ############################################

# ########## Incorrect Softgoal Syntax ###########
def check_softgoal_syntax(softgoals):

  incorrect_softgoals_syntax = []
  for softgoal in softgoals:
    result = check_verb_or_noun_phrase(softgoal)
    if result == "verb phrase":
      incorrect_softgoals_syntax.append(softgoal)

  if incorrect_softgoals_syntax:

    output = ""
    for incorrect_softgoal_syntax in incorrect_softgoals_syntax:
      refactored_element = prompt(
       '''
        You are a specialist in English linguistics.
        You will be provided with a sentence that is not a noun phrase, and your task is to make it a noun pharse representing a goal.
        A noun phrase should start with a noun.
        For example: high data quality, fast response time, and course registration.
        Answer with the new sentence only.''', incorrect_softgoal_syntax)
      output = output + '"' + incorrect_softgoal_syntax + '" should be refactored to "' + refactored_element + '"\n'
    #incorrect_softgoal_syntax = "\n".join(incorrect_softgoal_syntax)
    return "Incorrect softgoals syntax:\n" + output
  else:
    return ""
# ############################################

# ########## Incorrect Task Syntax ###########
# def find_NPs(sentences):

#   pipeline = TokenClassificationPipeline(model=pos_model, tokenizer=pos_tokenizer)

#   outputs = pipeline(sentences)

#   NPs = []
    
#   for idx, output in enumerate(outputs):
#     if not output[0]['entity'].startswith('V'):
#       NPs.append(sentences[idx])
    
#   return NPs

def check_task_syntax(tasks):

  incorrect_tasks_syntax = []
  for task in tasks:
    result = check_verb_or_noun_phrase(task)
    if result == "noun phrase":
      incorrect_tasks_syntax.append(task)


  if incorrect_tasks_syntax:
    output = ""
    for incorrect_task_syntax in incorrect_tasks_syntax:
      refactored_element = prompt(
       '''
        You are a specialist in English linguistics.
        You will be provided with a sentence that is not a verb phrase, and your task is to make it a verb pharse representing a task.
        A verb phrase should start with a verb.
        For example: provide maintenance services, help co-workers, and enhance quality.
        Answer with the new sentence only.''', incorrect_task_syntax)
      output = output + '"' + incorrect_task_syntax + '" should be refactored to "' + refactored_element + '"\n'
    #incorrect_task_syntax = "\n".join(incorrect_task_syntax)
    return "Incorrect tasks syntax:\n" + output
  else:
    return ""
# ############################################

# ########## Incorrect Resource Syntax ###########
def check_resource_syntax(resources):

  if len(resources) == 0:
    return ""

  #incorrect_resources_syntax = find_non_NPs(resources)
  incorrect_resources_syntax = []
  for resource in resources:
    result = check_verb_or_noun_phrase(resource)
    if result == "verb phrase":
      incorrect_resources_syntax.append(resource)

  if incorrect_resources_syntax:
    output = ""
    for incorrect_resource_syntax in incorrect_resources_syntax:
      refactored_element = prompt(
       '''
        You are a specialist in English linguistics.
        You will be provided with a sentence that is not a noun phrase, and your task is to make it a noun pharse representing a resource.
        A noun phrase should start with a noun.
        For example: internet, database, and files system.
        Answer with the new sentence only.''', incorrect_resource_syntax)
      output = output + '"' + incorrect_resource_syntax + '" should be refactored to "' + refactored_element + '"\n'

    #incorrect_resource_syntax = "\n".join(incorrect_resource_syntax)
    return "Incorrect resources syntax:\n" + output
  else:
    return ""
# ############################################

# ########## Similarity ###########
def get_similar_elements(elements_per_actor, similarity_threshold):

  # Prepare sentence pair array
  sentence_pairs = []

  for key, value in elements_per_actor.items():

      for i in range(len(elements_per_actor[key])):
          for j in range(i+1,len(elements_per_actor[key])):
              sentence_pairs.append([elements_per_actor[key][i], elements_per_actor[key][j]])

  # Predict semantic similarity
  semantic_similarity_scores = sentences_similarity_model.predict(sentence_pairs, show_progress_bar=True)

  similar_elements = []

  for index, value in enumerate(sentence_pairs):
    if semantic_similarity_scores[index] > similarity_threshold:
      similar_elements.append(value)
      #similar_elements.append('"'+value+'"')
      #semantic_similarity["pair_"+str(index+1)] = [value,semantic_similarity_scores[index]]

  if similar_elements:
    result_string = ""
    for sublist in similar_elements:
      result_string += ' and '.join(f'"{item}"' for item in sublist) + '\n'

    #similar_elements = [' and '.join('"' + ele + '"') for ele in similar_elements]
    #similar_elements = "\n".join(similar_elements)
    return "Similar elements:\n" + result_string
  else:
    return ""

  return semantic_similarity
# #################################

# ########## Misspelling ###########
# def get_misspelled_words(sentence):

#   spell = Speller(only_replacements=True)
    
#   misspelled= []
  
#   for word in sentence.split():
#       correct_word = spell(word)
#       if word != correct_word:
#           misspelled.append([word, correct_word])

#   return misspelled

def check_spelling(elements):

  refactored_elements = []

  for key, value in elements.items():
     for i in range(0, len(elements[key])):
        refactored_element = prompt(
        '''
        You are a specialist in English linguistics.
        You will be provided with a sentence and your task is to report any misspilled words and correct the spilling if needed.
        Answer with "correct" or "misspilled". In case the sentence is misspilled, correct it with the right spelling.
        Use a JSON format with keys 'original sentence', 'answer', and 'correct sentence'.
        For example: {'original sentence': 'incraese value', 'answer': 'misspilled', 'correct sentence': 'increase value'}''', elements[key][i])

        refactored_element = refactored_element.replace("'", '"')
        refactored_element = json.loads(refactored_element)

        if refactored_element['answer'] == 'misspilled':
          refactored_elements.append('"' + refactored_element["original sentence"] + '" should be written as "' + refactored_element["correct sentence"] + '"')


  if refactored_elements:
    refactored_elements = "\n".join(refactored_elements)
    return "Misspilled elements:\n" + refactored_elements
  else:
    return ""
# ##################################

# ########## NLI ###########
def do_nli(premise, hypothesis):

  # Tokenization 
  token_ids = []
  seg_ids = []
  mask_ids = []

  premise_id = nli_tokenizer.encode(premise, add_special_tokens = False)
  hypothesis_id = nli_tokenizer.encode(hypothesis, add_special_tokens = False)
  pair_token_ids = [nli_tokenizer.cls_token_id] + premise_id + [nli_tokenizer.sep_token_id] + hypothesis_id + [nli_tokenizer.sep_token_id]
  premise_len = len(premise_id)
  hypothesis_len = len(hypothesis_id)

  segment_ids = torch.tensor([0] * (premise_len + 2) + [1] * (hypothesis_len + 1))  # sentence 0 and sentence 1
  attention_mask_ids = torch.tensor([1] * (premise_len + hypothesis_len + 3))  # mask padded values

  token_ids.append(torch.tensor(pair_token_ids))
  seg_ids.append(segment_ids)
  mask_ids.append(attention_mask_ids)

  # Forward pass 
  token_ids = pad_sequence(token_ids, batch_first=True)
  mask_ids = pad_sequence(mask_ids, batch_first=True)
  seg_ids = pad_sequence(seg_ids, batch_first=True)

  with torch.no_grad():
    output = nli_model(token_ids,
                  token_type_ids=seg_ids,
                  attention_mask=mask_ids)

  # Output predication 
  result = ""
  prediction = np.argmax(output.logits.cpu().numpy()).flatten().item()
  if prediction == 0:
    result = "Entailment"
    #print("Entailment")
  elif prediction == 1:
    result = "Contradiction"
    #print("Contradiction")
  elif prediction == 2:
    result = "Neutral"
    #print("Neutral")
  
  return result

# Entailment
def check_entailment(decomposed_elements):

  sentence_pairs = []
  non_matching_elements = []

  for key, value in decomposed_elements.items():
      #print(key, value)
      for i in decomposed_elements[key]:
          #print(key, i)
          sentence_pairs.append([key, i])
              
  for sentence_pair in sentence_pairs:
    result = do_nli(sentence_pair[0], sentence_pair[1])
    print(result)
    if result != "Entailment":
      non_matching_elements.append(sentence_pair)
  
  if non_matching_elements:
    non_matching_elements = [' and '.join(ele) for ele in non_matching_elements]
    non_matching_elements = "\n".join(non_matching_elements)
    return "The following elements are miss matching:\n" + non_matching_elements
  else:
    return "There are no miss matched elements."

  return result

# Contradiction
def check_for_linguistic_conflict(pairs):

    pairs = ",".join(str(element) for element in pairs)

    contradicting_pairs = []
    result = prompt(
        '''
        You are a specialist in English linguistics.
        You will be provided with a list of sentencses pair, and your task is to determine whether each pair can be conflicting or not.
        For example: "Inrease quality of service" AND "Cut expenses" are conflicting because increasing quality usually requires spending money.
        For each pair, answer with "yes" or "no" with your reason in short.
        Use a list of dictionaries format with keys "pair" and "answer". Omit "reason" from your response.''', pairs)

    result = result.replace("'", '"')

    results = json.loads(result)
    for result in results:
      if result["answer"] == "yes":
        contradicting_pairs.append(result["pair"])

    return contradicting_pairs

def find_paths_between_elements(elements, start_element, end_element, visited, path=[]):

    visited[start_element] = True
    path.append(start_element)

    if start_element == end_element:
        yield list(path)
    else:
        for contrib in elements:
            if contrib[1] in visited: ## added
              if contrib[0] == start_element and not visited[contrib[1]]:
                yield from find_paths_between_elements(elements, contrib[1], end_element, visited, path)

    path.pop()
    visited[start_element] = False

def check_contradiction(elements_per_actor, contributing_elements):

  pairs_to_check_1 = []
  pairs_to_check_2 = []
  pairs_to_check_3 = []

  all_values_contributing_elements = []
  for values_list in contributing_elements.values():
      all_values_contributing_elements.extend(values_list)


  sentence_pairs = []
  contradicting_elements = []

  # case 1: contradicting elements contributing similarly to other elements
  for key, value in elements_per_actor.items():

      for i in range(len(elements_per_actor[key])):
          for j in range(i+1,len(elements_per_actor[key])):
              sentence_pairs.append([elements_per_actor[key][i], elements_per_actor[key][j]])

  for sentence_pair in sentence_pairs:
    contribution_scores = []

    for contributing_element in all_values_contributing_elements:

      if contributing_element[0] == sentence_pair[0] or contributing_element[0] == sentence_pair[1]:

        if contributing_element[2] == "make":
          contribution_score = 75
        elif contributing_element[2] == "help":
          contribution_score = 50
        elif contributing_element[2] == "somePositive":
          contribution_score = 25
        elif contributing_element[2] == "unknown":
          contribution_score = 0
        elif contributing_element[2] == "someNegative":
          contribution_score = -25
        elif contributing_element[2] == "break":
          contribution_score = -50
        elif contributing_element[2] == "hurt":
          contribution_score = -75
        else:
          contribution_score = int(contributing_element[2])

        contribution_scores.append((contributing_element[0], contribution_score))

    if len(contribution_scores) < 2:
      pairs_to_check_1.append([sentence_pair[0].replace("'", ""), sentence_pair[1].replace("'", "")])
    else:
      flag = 0
      for pair in itertools.combinations(contribution_scores, r=2):
        if pair[0][0] != pair[1][0]:
          if pair[0][1] * pair[1][1] < 0:
            flag = 1

      if flag == 0:
        pairs_to_check_2.append([sentence_pair[0].replace("'", ""), sentence_pair[1].replace("'", "")])

  # case 2: contradicting elements contributing similarly to each other, taking into considration the full path between the two elements
  for key, value in elements_per_actor.items():
    for element1 in value:
        for element2 in value:
            if element1 != element2:
                visited = {e: False for e in value}

                for path in find_paths_between_elements(all_values_contributing_elements, element1, element2, visited):

                    first_edge_value = next((contrib[2] for contrib in all_values_contributing_elements if contrib[0] == path[0] and contrib[1] == path[1]), None)
                    last_edge_value = next((contrib[2] for contrib in all_values_contributing_elements if contrib[0] == path[-2] and contrib[1] == path[-1]), None)

                    if first_edge_value is not None and last_edge_value is not None and int(first_edge_value) * int(last_edge_value) > 0:
                        pairs_to_check_3.append([sentence_pair[0].replace("'", ""), sentence_pair[1].replace("'", "")])

  pairs_to_check = pairs_to_check_1 + pairs_to_check_2 + pairs_to_check_3

  # Initialize an empty list to store the divided lists
  divided_lists = []

  # Iterate over the long list and create sublists of 30 items each
  for i in range(0, len(pairs_to_check), 30):
      sublist = pairs_to_check[i:i + 30]
      divided_lists.append(sublist)

  for divided_list in divided_lists:
    contradicting_elements = contradicting_elements + check_for_linguistic_conflict(divided_list)

  if contradicting_elements:
    # Using a set to store unique sublists
    contradicting_elements = set(tuple(sublist) for sublist in contradicting_elements)
    # Converting back to a list of lists
    contradicting_elements = [list(sublist) for sublist in contradicting_elements]

    contradicting_elements = [' and '.join(ele) for ele in contradicting_elements]
    contradicting_elements = "\n".join(contradicting_elements)

    return "Conflicting elements:\n" + contradicting_elements
  else:
    return ""
# ##########################

# ************************* User Interface *************************

def detect_bad_smells(tgrl_file, selected_bad_smells, size_threshold, similarity_threshold):

  output = ""

  tgrl_text = parse_tgrl(tgrl_file)

  all_elements, elements_per_actor, decomposed_elements, contributing_elements = extract_elements(tgrl_text)

  if 'Lengthy element' in selected_bad_smells:
    result = get_long_elements(all_elements, size_threshold)
    if result != "":
      output = output + result + "\n\n"

  if 'Complex element' in selected_bad_smells:
    result = get_complex_sentences(all_elements)
    if result != "":
      output = output + result + "\n\n"

  if 'Punctuation-marked element' in selected_bad_smells:
    result = get_punctuations(all_elements)
    if result != "":
        warning = "WARNING: Avoid using punctuation to imply priority or urgency. The use of punctuation can lead to misinterpretation and inefficiencies in communicating the goals/requirements.\n\n"
        output = warning + output + result + "\n\n"

  if 'Incorrect actor syntax' in selected_bad_smells:
    result = check_actor_syntax(all_elements['actors'])
    if result != "":
      output = output + result + "\n\n"

  if 'Incorrect goal syntax' in selected_bad_smells:
    result = check_goal_syntax(all_elements['goals'])
    if result != "":
      output = output + result + "\n\n"

  if 'Incorrect softgoal syntax' in selected_bad_smells:
    result = check_softgoal_syntax(all_elements['softGoals'])
    if result != "":
      output = output + result + "\n\n"

  if 'Incorrect task syntax' in selected_bad_smells:
    result = check_task_syntax(all_elements['tasks'])
    if result != "":
      output = output + result + "\n\n"

  if 'Incorrect resource syntax' in selected_bad_smells:
    result = check_resource_syntax(all_elements['resources'])
    if result != "":
      output = output + result + "\n\n"

  if 'Similar elements' in selected_bad_smells:
    result = get_similar_elements(elements_per_actor, similarity_threshold)
    if result != "":
      output = output + result + "\n\n"

  if 'Misspelled element' in selected_bad_smells:
    result = check_spelling(all_elements)
    if result != "":
      output = output + result + "\n\n"

  if 'Goal/Task and Sub-goal/Sub-task mismatch' in selected_bad_smells:
    result = check_entailment(decomposed_elements)
    if result != "":
      output = output + result + "\n\n"

  if 'Conflicting elements' in selected_bad_smells:
    result = check_contradiction(elements_per_actor, contributing_elements)
    if result != "":
      output = output + result + "\n\n"

  return output


interface = gr.Interface(fn = detect_bad_smells,
                         inputs = [gr.File(label="TGRL File"),
                         gr.CheckboxGroup(["Lengthy element", "Complex element", "Punctuation-marked element", "Incorrect actor syntax", "Incorrect goal syntax", "Incorrect softgoal syntax", "Incorrect task syntax", "Incorrect resource syntax", "Similar elements", "Misspelled element", "Goal/Task and Sub-goal/Sub-task mismatch", "Conflicting elements"],
                                           label="Which bad smells you want to detect and refactor?"),
                                   gr.Slider(label= "Length threshold", value = 5, minimum = 2, maximum = 10, step = 1),
                                   gr.Slider(label= "Similarity threshold", value = 0.9, minimum = 0, maximum = 1, step = 0.1)],
                         outputs = [gr.Textbox(label= "Detected and refactored bad smells:")],
                         title = "TGRL Bad Smells Detection and Refactoring",
                         description = "Upload your .xgrl file and we will find the bad smells and refactor them for you!",
                         theme = gr.themes.Soft())


interface.launch(inline = False)