Spaces:
Sleeping
Sleeping
Update components/pubmed_search.py
Browse files- components/pubmed_search.py +53 -25
components/pubmed_search.py
CHANGED
@@ -1,7 +1,6 @@
|
|
1 |
from Bio import Entrez
|
2 |
-
import os
|
3 |
-
|
4 |
-
# ---------------------------- Configuration ----------------------------
|
5 |
|
6 |
# ---------------------------- Helper Functions ----------------------------
|
7 |
|
@@ -14,50 +13,79 @@ def log_error(message: str):
|
|
14 |
except:
|
15 |
print("Couldn't write to error log file.") #If logging fails, still print to console
|
16 |
|
17 |
-
# ----------------------------
|
18 |
|
19 |
def search_pubmed(query: str) -> list:
|
20 |
"""Searches PubMed and returns a list of article IDs using semantic search."""
|
|
|
21 |
try:
|
22 |
Entrez.email = ENTREZ_EMAIL
|
23 |
print(f"Entrez Email: {Entrez.email}") # DEBUG: Check the email being used
|
24 |
print(f"PubMed Query: {query}") # DEBUG: Check the query being sent
|
25 |
|
26 |
-
|
27 |
-
from sentence_transformers import SentenceTransformer, util
|
28 |
-
|
29 |
-
model = SentenceTransformer('all-mpnet-base-v2') #Model by all-mpnet-base-v2
|
30 |
-
|
31 |
# Fetch PubMed IDs
|
32 |
-
|
33 |
-
|
34 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
|
|
|
36 |
# Fetch abstracts for all IDs:
|
37 |
abstracts = []
|
38 |
for article_id in id_list:
|
39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
|
41 |
-
|
42 |
-
|
43 |
-
abstract_embeddings = model.encode(abstracts)
|
44 |
|
45 |
-
|
46 |
-
similarities = util.cos_sim(query_embedding, abstract_embeddings)[0]
|
47 |
|
48 |
-
|
49 |
-
ranked_articles = sorted(zip(id_list, similarities), key=lambda x: x[1], reverse=True)
|
50 |
|
51 |
-
|
52 |
-
|
|
|
53 |
|
54 |
-
print(f"PubMed Results: {ranked_ids}") # DEBUG: Check the results
|
55 |
-
return ranked_ids
|
56 |
except Exception as e:
|
57 |
log_error(f"PubMed search error: {e}")
|
58 |
return [f"Error during PubMed search: {e}"]
|
59 |
|
60 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
|
62 |
def medai_agent(query: str) -> str:
|
63 |
"""Orchestrates the medical literature review and presents abstract."""
|
|
|
1 |
from Bio import Entrez
|
2 |
+
import os
|
3 |
+
from sentence_transformers import SentenceTransformer, util
|
|
|
4 |
|
5 |
# ---------------------------- Helper Functions ----------------------------
|
6 |
|
|
|
13 |
except:
|
14 |
print("Couldn't write to error log file.") #If logging fails, still print to console
|
15 |
|
16 |
+
# ---------------------------- Functions ----------------------------
|
17 |
|
18 |
def search_pubmed(query: str) -> list:
|
19 |
"""Searches PubMed and returns a list of article IDs using semantic search."""
|
20 |
+
ENTREZ_EMAIL = os.environ.get("ENTREZ_EMAIL", "[email protected]")
|
21 |
try:
|
22 |
Entrez.email = ENTREZ_EMAIL
|
23 |
print(f"Entrez Email: {Entrez.email}") # DEBUG: Check the email being used
|
24 |
print(f"PubMed Query: {query}") # DEBUG: Check the query being sent
|
25 |
|
26 |
+
"" Added code ""
|
|
|
|
|
|
|
|
|
27 |
# Fetch PubMed IDs
|
28 |
+
try:
|
29 |
+
handle = Entrez.esearch(db="pubmed", term=query, retmax="5")
|
30 |
+
record = Entrez.read(handle)
|
31 |
+
id_list = record["IdList"]
|
32 |
+
print(f"Original PubMed Results: {id_list}")
|
33 |
+
|
34 |
+
except Exception as e:
|
35 |
+
log_error(f"First step of PubMed search error: {e}")
|
36 |
+
return [f"First step of Error during PubMed search: {e}"]
|
37 |
|
38 |
+
"" Added code ""
|
39 |
# Fetch abstracts for all IDs:
|
40 |
abstracts = []
|
41 |
for article_id in id_list:
|
42 |
+
abstract_text = fetch_abstract(article_id)
|
43 |
+
if "Error" in abstract_text:
|
44 |
+
print(f"Skipping article {article_id} due to abstract error: {abstract_text}")
|
45 |
+
continue
|
46 |
+
abstracts.append(abstract_text)
|
47 |
+
|
48 |
+
"" Added code ""
|
49 |
+
# Semantic Search Using Sentence Transformers:
|
50 |
+
try:
|
51 |
+
model = SentenceTransformer('all-mpnet-base-v2') #Model by all-mpnet-base-v2
|
52 |
+
# Generate embeddings for abstracts and the query:
|
53 |
+
query_embedding = model.encode(query)
|
54 |
+
abstract_embeddings = model.encode(abstracts)
|
55 |
+
|
56 |
+
# Calculate cosine similarities
|
57 |
+
similarities = util.cos_sim(query_embedding, abstract_embeddings)[0]
|
58 |
+
|
59 |
+
# Sort by similarity (higher is better)
|
60 |
+
ranked_articles = sorted(zip(id_list, similarities), key=lambda x: x[1], reverse=True)
|
61 |
|
62 |
+
# Extract ranked IDs:
|
63 |
+
ranked_ids = [article_id for article_id, similarity in ranked_articles]
|
|
|
64 |
|
65 |
+
print(f"Semantic search code: {ranked_ids}") # DEBUG: Check the results
|
|
|
66 |
|
67 |
+
return ranked_ids
|
|
|
68 |
|
69 |
+
except Exception as e:
|
70 |
+
log_error(f"Semantic search code. : {e}")
|
71 |
+
return [f"Error during semantic search code: {e}"]
|
72 |
|
|
|
|
|
73 |
except Exception as e:
|
74 |
log_error(f"PubMed search error: {e}")
|
75 |
return [f"Error during PubMed search: {e}"]
|
76 |
|
77 |
+
def fetch_abstract(article_id: str) -> str:
|
78 |
+
"""Fetches the abstract for a given PubMed article ID."""
|
79 |
+
ENTREZ_EMAIL = os.environ.get("ENTREZ_EMAIL", "[email protected]")
|
80 |
+
try:
|
81 |
+
Entrez.email = ENTREZ_EMAIL
|
82 |
+
handle = Entrez.efetch(db="pubmed", id=article_id, rettype="abstract", retmode="text")
|
83 |
+
abstract = handle.read()
|
84 |
+
handle.close()
|
85 |
+
return abstract
|
86 |
+
except Exception as e:
|
87 |
+
log_error(f"Error fetching abstract for {article_id}: {e}")
|
88 |
+
return f"Error fetching abstract for {article_id}: {e}"
|
89 |
|
90 |
def medai_agent(query: str) -> str:
|
91 |
"""Orchestrates the medical literature review and presents abstract."""
|