CopilotForDocs/main.py at main · lavina98/CopilotForDocs · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
from pypdf import PdfReader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceBgeEmbeddings
import re
from keybert import KeyBERT
import nltk
from nltk import WordNetLemmatizer, sent_tokenize, word_tokenize
from dotenv import load_dotenv
import os
import pinecone
from transformers import pipeline

def main():
    #nltk.download('wordnet')
    #nltk.download('punkt')
    load_dotenv()
    api_key = os.getenv("PINECONE_API_KEY")
    pinecone.init(api_key=api_key, environment='us-west4-gcp-free')
    if 'honda-city-manual' not in pinecone.list_indexes():
        load_data()

    index = pinecone.Index("honda-city-manual")
    #load_data()
    question = "What does honda do with the data it acquires"
    keyword_model =  KeyBERT()
    keyword = keyword_model.extract_keywords(question, keyphrase_ngram_range=(1, 1), stop_words=None)
    #print(keyword)
    question_keyword = ' '.join([element[0] for element in keyword])

    model_name = "BAAI/bge-base-en-v1.5"
    model_kwargs = {'device': 'cpu'}
    encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity

    model = HuggingFaceBgeEmbeddings(
        model_name=model_name,
        model_kwargs=model_kwargs,
        encode_kwargs=encode_kwargs,
    )

    embedded_keys = model.embed_query(question_keyword)
    results_obj = index.query(queries=[embedded_keys], top_k=2, include_metadata=True)
    results = results_obj['results'][0]['matches']

    context= ''
    for result in results:
       context += result['metadata']['data'] + '.'

    qa_model_name = "deepset/roberta-base-squad2"
    nlp = pipeline('question-answering', model=qa_model_name, tokenizer=qa_model_name)
    QA_input = {
        'question': question,
        'context': context
    }
    res = nlp(QA_input)
    print(question)
    print(res)

def load_data():
    directory = os.path.dirname(__file__)
    pdfPath = os.path.join(directory, 'HondaCityManual.pdf')
    reader = PdfReader(pdfPath)
    num_pages = len(reader.pages)
    pages = [reader.pages[i].extract_text() for i in range(0, 10)]
    cleaned_text_with_newlines = [re.sub(r"[^a-zA-Z0-9\s\.\?,!]", "", page) for page in pages]
    cleaned_text = [re.sub(r"\n", " ", page) for page in cleaned_text_with_newlines]
    #print(cleaned_text[1])
    lemmatizer = WordNetLemmatizer()

    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size = 500,
        chunk_overlap = 20,
        length_function = len,
        is_separator_regex=False
    )
    chunks = []
    for page in cleaned_text:
        chunks.extend(text_splitter.split_text(page))

    lemmatized_chunks = []
    for chunk in chunks:
        lemmatized_chunk = ' '. join([lemmatizer.lemmatize(word.lower()) for word in word_tokenize(chunk)])
        lemmatized_chunks.append(lemmatized_chunk)

    keyword_model =  KeyBERT()
    keywords = keyword_model.extract_keywords(lemmatized_chunks, keyphrase_ngram_range=(1, 1), stop_words=None)

    keys = []
    for arr in keywords:
        els = [element[0] for element in arr]
        els.sort()
        keys.append(' '.join(el for el in els))

    model_name = "BAAI/bge-base-en-v1.5"
    model_kwargs = {'device': 'cpu'}
    encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity

    model = HuggingFaceBgeEmbeddings(
        model_name=model_name,
        model_kwargs=model_kwargs,
        encode_kwargs=encode_kwargs,
    )

    embedded_keys = model.embed_documents(keys)

    key_embeddings =[]
    for i in range(len(embedded_keys)):
        key_embeddings.append((f'{keys[i]}', embedded_keys[i] , { 'data':chunks[i]}))


    load_dotenv()
    api_key = os.getenv("PINECONE_API_KEY")
    pinecone.init(api_key=api_key, environment='us-west4-gcp-free')

    pinecone.create_index("honda-city-manual", metric="cosine", dimension=768)
    index = pinecone.Index("honda-city-manual")
    index.upsert(key_embeddings)


def get_unique_char_set(pages):
    unique_set = set()
    for page in pages:
        unique_set.update(set(page))
    return unique_set

if __name__ == "__main__":
    main()