Refineverse_testBG / TextSummarizationFeature.py
bsjghee001's picture
Duplicate from Nik97/Refineverse
910fc51
# Contains only NLP text summarization function codes!
import re # Python's built-in library for regular expressions (or Regex)
import sqlite3
from flask import g
from transformers import pipeline
def summarize(Entered_story):
# Check if the input is empty
if not Entered_story.strip():
raise ValueError("Empty input!")
# Validate that the input is in the correct format
if not validate_story(Entered_story):
raise ValueError("Incorrect format!")
# Before we do anything, make sure the input is long enough for summarization.
if len(Entered_story) < 200:
raise ValueError("Invalid length!")
# Set the pipeline to use the correct NLP type and model
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
# Take note: The max_length & min_length variables refer to the OUTPUT length!
summary = summarizer(Entered_story, max_length=100, min_length=30, do_sample=False)[0]["summary_text"]
return summary
# User Input Format Validation Function for english only
def validate_story(Entered_story):
pattern = r'As a (?P<role>[^,.]+), I want to (?P<goal>[^,.]+)(,|.)+so that (?P<benefit>.+)'
match = re.search(pattern, Entered_story, flags=re.DOTALL)
return bool(match)
# User Input Format Validation Function for all 4 languages
#def validate_story(Entered_story):
# Pattern for English language
#pattern_en = r'As a (?P<role>[^,.]+), I want to (?P<goal>[^,.]+)(,|.)+so that (?P<benefit>.+)'
# Pattern for Chinese language
#pattern_zh = r'作为(?P<role>[^,.]+),我想要(?P<goal>[^,.]+)(,|。)+以便(?P<benefit>.+)'
# Pattern for Malay language
#pattern_ms = r'Sebagai(?P<role>[^,.]+), saya mahu(?P<goal>[^,.]+)(,|.)+supaya(?P<benefit>.+)'
# Pattern for Tamil language
#pattern_ta = r'என(?P<role>[^,.]+) எனக்கு வேண்டும்(?P<goal>[^,.]+)(,|.)+அதனால்(?P<benefit>.+) பயன்படுத்தி வைக்கும்'
# Pattern for Thai language
#pattern_th = r'ในฐานะ(?P<role>[^,.]+) ฉันต้องการ(?P<goal>[^,.]+)(,|.)+เพื่อที่ฉัน(?P<benefit>.+)'
# Try each pattern to see if there is a match
#match_en = re.search(pattern_en, Entered_story, flags=re.DOTALL)
#match_zh = re.search(pattern_zh, Entered_story, flags=re.DOTALL)
#match_ms = re.search(pattern_ms, Entered_story, flags=re.DOTALL)
#match_ta = re.search(pattern_ta, Entered_story, flags=re.DOTALL)
#match_th = re.search(pattern_th, Entered_story, flags=re.DOTALL)
# Return True if at least one pattern matches, otherwise False
#return bool(match_en or match_zh or match_ms or match_ta or match_th)
# Function to grab all contents in the "TextSummarization" table (except for unique ids)
# If adding any additional attributes to the table, this has to be updated accordingly
def getTextSumContents():
db = getattr(g, '_database', None) # Gets the _database attribute from the 'g' object. If it does not exist, returns 'None'
if db is None:
db = g._database = sqlite3.connect('Refineverse.db') # If db is None, create a new connection for db and g._database.
cursor = db.cursor() # Creates a cursor object to handle data
cursor.execute("SELECT Entered_story, summary FROM TextSummarization") # The cursor executes the query
rows = cursor.fetchall() # Stores the results of fetchall() into a variable
return rows
# Function to insert a new row into the "TextSummarization" table
# Using "with" for the connection here seems important, as otherwise it results in an exception
def insertTextSumRow( Entered_story, summary):
with sqlite3.connect('Refineverse.db') as conn: # 'With' will automatically take care of closing and opening the connection
cursor = conn.cursor()
cursor.execute("INSERT INTO TextSummarization (Entered_story, summary) VALUES (?, ?)", (Entered_story, summary))
conn.commit()