Legal_Document_Similarity

The Legal_Document_Similarity class is designed for comparing legal documents to a corpus of legal text using TF-IDF vectorization and cosine similarity, allowing for the identification of similar legal content within documents.

git clone https://github.com/ChSatyaSavith/Legal_Document_Similarity.git

Installation

Dependencies

Legal_Document_Similarity requires:-

• numpy (>=1.23.5)
• pandas (>=1.5.2)
• scikit-learn (>=1.2.2)
• nltk (>=3.8.1)
• python-docx (>=0.8.11)
• PyPDF2 (>=3.0.1)
• reportlab (>=3.6.8)

Installing requirements using pip

pip install -r requirements.txt

Documentation

update_embeddings(pdf_path,law_book_sentences='law_book.txt',book_path = True,folder_path = False)

• Either pass the path of folder which contains law documents in which case set book_path = False and set folder_path = True or pass the path of law document in which case set book_path = True and set folder_path = False
• law_book_sentences contains the path of where the law book will be saved at
• Only call update_embeddings if you want to update the law book otherwise there is no need to call it

read_docx(folder_path)

• Pass the pdf path of the generated document

checkSimilarity()

• Checks the similarity between the generated document and the law book
• Create a dataframe which has three columns Generated_Document, Similar_Line in Acts and Similarity_Score
• Also creates a variable percentage_match which tells us the percentage of match between the generated document and the law book

save_Acts(path)

• Saves the dataframe in question, answer and similarity score format at the specified path

Sample Code

from Similarity import LegalDocumentSimilarity
obj = LegalDocumentSimilarity()
obj.read_docx(pdf_path)
obj.checkSimilarity()
obj.save_Acts('output.pdf')