Spaces:
Running
Running
Commit
·
99723c5
1
Parent(s):
4c47563
comment NLTK package
Browse files- app/search/bm25_search.py +39 -59
app/search/bm25_search.py
CHANGED
@@ -1,46 +1,38 @@
|
|
1 |
-
# bm25_search.py
|
2 |
import asyncio
|
3 |
from rank_bm25 import BM25Okapi
|
4 |
-
import nltk
|
5 |
import string
|
6 |
from typing import List, Set, Optional
|
7 |
-
from nltk.corpus import stopwords
|
8 |
-
from nltk.stem import WordNetLemmatizer
|
9 |
import os
|
10 |
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
print(f"Error downloading {resource}: {str(e)}")
|
26 |
|
27 |
class BM25_search:
|
28 |
-
# Class variable to track if resources have been downloaded
|
29 |
nltk_resources_downloaded = False
|
30 |
|
31 |
def __init__(self, remove_stopwords: bool = True, perform_lemmatization: bool = False):
|
32 |
"""
|
33 |
Initializes the BM25search.
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
# Ensure NLTK resources are downloaded only once
|
40 |
-
if not BM25_search.nltk_resources_downloaded:
|
41 |
-
|
42 |
-
download_nltk_resources()
|
43 |
-
BM25_search.nltk_resources_downloaded = True # Mark as downloaded
|
44 |
|
45 |
self.documents: List[str] = []
|
46 |
self.doc_ids: List[str] = []
|
@@ -48,20 +40,22 @@ class BM25_search:
|
|
48 |
self.bm25: Optional[BM25Okapi] = None
|
49 |
self.remove_stopwords = remove_stopwords
|
50 |
self.perform_lemmatization = perform_lemmatization
|
51 |
-
|
52 |
-
self.
|
|
|
53 |
|
54 |
def preprocess(self, text: str) -> List[str]:
|
55 |
"""
|
56 |
-
Preprocesses the input text by lowercasing
|
57 |
-
|
58 |
"""
|
59 |
text = text.lower().translate(str.maketrans('', '', string.punctuation))
|
60 |
-
tokens = nltk.word_tokenize(text)
|
61 |
-
|
62 |
-
|
63 |
-
if
|
64 |
-
|
|
|
65 |
return tokens
|
66 |
|
67 |
def add_document(self, doc_id: str, new_doc: str) -> None:
|
@@ -69,11 +63,9 @@ class BM25_search:
|
|
69 |
Adds a new document to the corpus and updates the BM25 index.
|
70 |
"""
|
71 |
processed_tokens = self.preprocess(new_doc)
|
72 |
-
|
73 |
self.documents.append(new_doc)
|
74 |
self.doc_ids.append(doc_id)
|
75 |
self.tokenized_docs.append(processed_tokens)
|
76 |
-
# Ensure update_bm25 is awaited if required in async context
|
77 |
self.update_bm25()
|
78 |
print(f"Added document ID: {doc_id}")
|
79 |
|
@@ -101,14 +93,12 @@ class BM25_search:
|
|
101 |
else:
|
102 |
print("No documents to initialize BM25.")
|
103 |
|
104 |
-
|
105 |
def get_scores(self, query: str) -> List[float]:
|
106 |
"""
|
107 |
Computes BM25 scores for all documents based on the given query.
|
108 |
"""
|
109 |
processed_query = self.preprocess(query)
|
110 |
print(f"Tokenized Query: {processed_query}")
|
111 |
-
|
112 |
if self.bm25:
|
113 |
return self.bm25.get_scores(processed_query)
|
114 |
else:
|
@@ -123,9 +113,9 @@ class BM25_search:
|
|
123 |
if self.bm25:
|
124 |
return self.bm25.get_top_n(processed_query, self.documents, n)
|
125 |
else:
|
126 |
-
print("initialized.")
|
127 |
return []
|
128 |
-
|
129 |
def clear_documents(self) -> None:
|
130 |
"""
|
131 |
Clears all documents from the BM25 index.
|
@@ -133,18 +123,12 @@ class BM25_search:
|
|
133 |
self.documents = []
|
134 |
self.doc_ids = []
|
135 |
self.tokenized_docs = []
|
136 |
-
self.bm25 = None
|
137 |
print("BM25 documents cleared and index reset.")
|
138 |
-
|
139 |
def get_document(self, doc_id: str) -> str:
|
140 |
"""
|
141 |
Retrieves a document by its document ID.
|
142 |
-
|
143 |
-
Parameters:
|
144 |
-
- doc_id (str): The ID of the document to retrieve.
|
145 |
-
|
146 |
-
Returns:
|
147 |
-
- str: The document text if found, otherwise an empty string.
|
148 |
"""
|
149 |
try:
|
150 |
index = self.doc_ids.index(doc_id)
|
@@ -153,13 +137,9 @@ class BM25_search:
|
|
153 |
print(f"Document ID {doc_id} not found.")
|
154 |
return ""
|
155 |
|
156 |
-
|
157 |
async def initialize_bm25_search(remove_stopwords: bool = True, perform_lemmatization: bool = False) -> BM25_search:
|
158 |
"""
|
159 |
-
Initializes the BM25search
|
160 |
"""
|
161 |
-
|
162 |
-
await loop.run_in_executor(None, download_nltk_resources)
|
163 |
return BM25_search(remove_stopwords, perform_lemmatization)
|
164 |
-
|
165 |
-
|
|
|
|
|
1 |
import asyncio
|
2 |
from rank_bm25 import BM25Okapi
|
3 |
+
# import nltk
|
4 |
import string
|
5 |
from typing import List, Set, Optional
|
6 |
+
# from nltk.corpus import stopwords
|
7 |
+
# from nltk.stem import WordNetLemmatizer
|
8 |
import os
|
9 |
|
10 |
+
# Commented out this function that downloads NLTK resources.
|
11 |
+
# def download_nltk_resources():
|
12 |
+
# """
|
13 |
+
# Downloads required NLTK resources synchronously.
|
14 |
+
# """
|
15 |
+
# resources = ['punkt', 'stopwords', 'wordnet', 'omw-1.4']
|
16 |
+
# nltk_data_path = "/tmp/nltk_data"
|
17 |
+
# os.makedirs(nltk_data_path, exist_ok=True)
|
18 |
+
# nltk.data.path.append(nltk_data_path)
|
19 |
+
# for resource in resources:
|
20 |
+
# try:
|
21 |
+
# nltk.download(resource, download_dir=nltk_data_path, quiet=True)
|
22 |
+
# except Exception as e:
|
23 |
+
# print(f"Error downloading {resource}: {str(e)}")
|
|
|
24 |
|
25 |
class BM25_search:
|
|
|
26 |
nltk_resources_downloaded = False
|
27 |
|
28 |
def __init__(self, remove_stopwords: bool = True, perform_lemmatization: bool = False):
|
29 |
"""
|
30 |
Initializes the BM25search.
|
31 |
+
"""
|
32 |
+
# Commented out NLTK resource initialization
|
33 |
+
# if not BM25_search.nltk_resources_downloaded:
|
34 |
+
# download_nltk_resources()
|
35 |
+
# BM25_search.nltk_resources_downloaded = True
|
|
|
|
|
|
|
|
|
|
|
36 |
|
37 |
self.documents: List[str] = []
|
38 |
self.doc_ids: List[str] = []
|
|
|
40 |
self.bm25: Optional[BM25Okapi] = None
|
41 |
self.remove_stopwords = remove_stopwords
|
42 |
self.perform_lemmatization = perform_lemmatization
|
43 |
+
# Commented out NLTK-specific tools
|
44 |
+
# self.stop_words: Set[str] = set(stopwords.words('english')) if remove_stopwords else set()
|
45 |
+
# self.lemmatizer = WordNetLemmatizer() if perform_lemmatization else None
|
46 |
|
47 |
def preprocess(self, text: str) -> List[str]:
|
48 |
"""
|
49 |
+
Preprocesses the input text by lowercasing and removing punctuation.
|
50 |
+
NLTK-related tokenization, stopword removal, and lemmatization are commented out.
|
51 |
"""
|
52 |
text = text.lower().translate(str.maketrans('', '', string.punctuation))
|
53 |
+
# tokens = nltk.word_tokenize(text) # Commented out NLTK tokenization
|
54 |
+
tokens = text.split() # Basic tokenization as a fallback
|
55 |
+
# if self.remove_stopwords:
|
56 |
+
# tokens = [token for token in tokens if token not in self.stop_words]
|
57 |
+
# if self.perform_lemmatization and self.lemmatizer:
|
58 |
+
# tokens = [self.lemmatizer.lemmatize(token) for token in tokens]
|
59 |
return tokens
|
60 |
|
61 |
def add_document(self, doc_id: str, new_doc: str) -> None:
|
|
|
63 |
Adds a new document to the corpus and updates the BM25 index.
|
64 |
"""
|
65 |
processed_tokens = self.preprocess(new_doc)
|
|
|
66 |
self.documents.append(new_doc)
|
67 |
self.doc_ids.append(doc_id)
|
68 |
self.tokenized_docs.append(processed_tokens)
|
|
|
69 |
self.update_bm25()
|
70 |
print(f"Added document ID: {doc_id}")
|
71 |
|
|
|
93 |
else:
|
94 |
print("No documents to initialize BM25.")
|
95 |
|
|
|
96 |
def get_scores(self, query: str) -> List[float]:
|
97 |
"""
|
98 |
Computes BM25 scores for all documents based on the given query.
|
99 |
"""
|
100 |
processed_query = self.preprocess(query)
|
101 |
print(f"Tokenized Query: {processed_query}")
|
|
|
102 |
if self.bm25:
|
103 |
return self.bm25.get_scores(processed_query)
|
104 |
else:
|
|
|
113 |
if self.bm25:
|
114 |
return self.bm25.get_top_n(processed_query, self.documents, n)
|
115 |
else:
|
116 |
+
print("BM25 is not initialized.")
|
117 |
return []
|
118 |
+
|
119 |
def clear_documents(self) -> None:
|
120 |
"""
|
121 |
Clears all documents from the BM25 index.
|
|
|
123 |
self.documents = []
|
124 |
self.doc_ids = []
|
125 |
self.tokenized_docs = []
|
126 |
+
self.bm25 = None
|
127 |
print("BM25 documents cleared and index reset.")
|
128 |
+
|
129 |
def get_document(self, doc_id: str) -> str:
|
130 |
"""
|
131 |
Retrieves a document by its document ID.
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
"""
|
133 |
try:
|
134 |
index = self.doc_ids.index(doc_id)
|
|
|
137 |
print(f"Document ID {doc_id} not found.")
|
138 |
return ""
|
139 |
|
|
|
140 |
async def initialize_bm25_search(remove_stopwords: bool = True, perform_lemmatization: bool = False) -> BM25_search:
|
141 |
"""
|
142 |
+
Initializes the BM25search.
|
143 |
"""
|
144 |
+
# Removed NLTK resource download from async context
|
|
|
145 |
return BM25_search(remove_stopwords, perform_lemmatization)
|
|
|
|