Spaces:
Runtime error
Runtime error
Upload 3 files
Browse files- README.md +8 -5
- app.py +154 -0
- requirements.txt +7 -0
README.md
CHANGED
@@ -1,12 +1,15 @@
|
|
1 |
---
|
2 |
-
title:
|
3 |
-
emoji:
|
4 |
-
colorFrom:
|
5 |
-
colorTo:
|
6 |
sdk: gradio
|
7 |
sdk_version: 3.35.2
|
8 |
app_file: app.py
|
9 |
pinned: false
|
|
|
10 |
---
|
11 |
|
12 |
-
|
|
|
|
|
|
1 |
---
|
2 |
+
title: NBDT Reviewer Recommendation System
|
3 |
+
emoji: π
|
4 |
+
colorFrom: indigo
|
5 |
+
colorTo: blue
|
6 |
sdk: gradio
|
7 |
sdk_version: 3.35.2
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
+
models: [biodatlab/MIReAD-Neuro]
|
11 |
---
|
12 |
|
13 |
+
This space is a demo for a Reviewer Recommendation System for the Neurons, Behavior, Data Analysis and Theory Journal.
|
14 |
+
The index being used here includes papers from a variety of authors who have published in the NBDT Journal across various years.
|
15 |
+
The embedding model in use here is [biodatlab/MIReAD-Neuro-Large](https://huggingface.co/biodatlab/MIReAD-Neuro-Large).
|
app.py
ADDED
@@ -0,0 +1,154 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from langchain.vectorstores import FAISS
|
3 |
+
from langchain.embeddings import HuggingFaceEmbeddings
|
4 |
+
import torch
|
5 |
+
|
6 |
+
|
7 |
+
def create_miread_embed(sents, bundle):
|
8 |
+
tokenizer = bundle[0]
|
9 |
+
model = bundle[1]
|
10 |
+
model.cpu()
|
11 |
+
tokens = tokenizer(sents,
|
12 |
+
max_length=512,
|
13 |
+
padding=True,
|
14 |
+
truncation=True,
|
15 |
+
return_tensors="pt"
|
16 |
+
)
|
17 |
+
device = torch.device('cpu')
|
18 |
+
tokens = tokens.to(device)
|
19 |
+
with torch.no_grad():
|
20 |
+
out = model.bert(**tokens)
|
21 |
+
feature = out.last_hidden_state[:, 0, :]
|
22 |
+
return feature.cpu()
|
23 |
+
|
24 |
+
|
25 |
+
def get_matches(query, k):
|
26 |
+
matches = vecdb.similarity_search_with_score(query, k=k)
|
27 |
+
return matches
|
28 |
+
|
29 |
+
|
30 |
+
def inference(query, k=30):
|
31 |
+
matches = get_matches(query, k)
|
32 |
+
j_bucket = {}
|
33 |
+
n_table = []
|
34 |
+
a_table = []
|
35 |
+
scores = [round(match[1].item(), 3) for match in matches]
|
36 |
+
min_score = min(scores)
|
37 |
+
max_score = max(scores)
|
38 |
+
def normaliser(x): return round(1 - (x-min_score)/max_score, 3)
|
39 |
+
for i, match in enumerate(matches):
|
40 |
+
doc = match[0]
|
41 |
+
score = round(normaliser(round(match[1].item(), 3)),3)
|
42 |
+
title = doc.metadata['title']
|
43 |
+
author = doc.metadata['authors'][0].title()
|
44 |
+
date = doc.metadata.get('date', 'None')
|
45 |
+
link = doc.metadata.get('link', 'None')
|
46 |
+
submitter = doc.metadata.get('submitter', 'None')
|
47 |
+
# journal = doc.metadata.get('journal', 'None').strip()
|
48 |
+
journal = doc.metadata['journal']
|
49 |
+
if (journal == None or journal.strip() == ''):
|
50 |
+
journal = 'None'
|
51 |
+
else:
|
52 |
+
journal = journal.strip()
|
53 |
+
# For journals
|
54 |
+
if journal not in j_bucket:
|
55 |
+
j_bucket[journal] = score
|
56 |
+
else:
|
57 |
+
j_bucket[journal] += score
|
58 |
+
|
59 |
+
# For authors
|
60 |
+
record = [i+1,
|
61 |
+
score,
|
62 |
+
author,
|
63 |
+
title,
|
64 |
+
link,
|
65 |
+
date]
|
66 |
+
n_table.append(record)
|
67 |
+
|
68 |
+
# For abstracts
|
69 |
+
record = [i+1,
|
70 |
+
title,
|
71 |
+
author,
|
72 |
+
submitter,
|
73 |
+
journal,
|
74 |
+
date,
|
75 |
+
link,
|
76 |
+
score
|
77 |
+
]
|
78 |
+
a_table.append(record)
|
79 |
+
|
80 |
+
del j_bucket['None']
|
81 |
+
j_table = sorted([[journal, round(score,3)] for journal,
|
82 |
+
score in j_bucket.items()],
|
83 |
+
key=lambda x: x[1], reverse=True)
|
84 |
+
j_table = [[i+1, item[0], item[1]] for i, item in enumerate(j_table)]
|
85 |
+
j_output = gr.Dataframe.update(value=j_table, visible=True)
|
86 |
+
n_output = gr.Dataframe.update(value=n_table, visible=True)
|
87 |
+
a_output = gr.Dataframe.update(value=a_table, visible=True)
|
88 |
+
|
89 |
+
return [a_output, j_output, n_output]
|
90 |
+
|
91 |
+
|
92 |
+
model_name = "biodatlab/MIReAD-Neuro-Large"
|
93 |
+
model_kwargs = {'device': 'cpu'}
|
94 |
+
encode_kwargs = {'normalize_embeddings': False}
|
95 |
+
faiss_embedder = HuggingFaceEmbeddings(
|
96 |
+
model_name=model_name,
|
97 |
+
model_kwargs=model_kwargs,
|
98 |
+
encode_kwargs=encode_kwargs
|
99 |
+
)
|
100 |
+
|
101 |
+
vecdb = FAISS.load_local("nbdt_index", faiss_embedder)
|
102 |
+
|
103 |
+
|
104 |
+
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
105 |
+
gr.Markdown("# NBDT Recommendation Engine for Editors")
|
106 |
+
gr.Markdown("NBDT Recommendation Engine for Editors is a tool for neuroscience authors/abstracts/journalsrecommendation built for NBDT journal editors. \
|
107 |
+
It aims to help an editor to find similar reviewers, abstracts, and journals to a given submitted abstract.\
|
108 |
+
To find a recommendation, paste a `title[SEP]abstract` or `abstract` in the text box below and click \"Find Matches\".\
|
109 |
+
Then, you can hover to authors/abstracts/journals tab to find a suggested list.\
|
110 |
+
The data in our current demo includes authors associated with the NBDT Journal. We will update the data monthly for an up-to-date publications.")
|
111 |
+
|
112 |
+
abst = gr.Textbox(label="Abstract", lines=10)
|
113 |
+
|
114 |
+
k = gr.Slider(1, 100, step=1, value=50,
|
115 |
+
label="Number of matches to consider")
|
116 |
+
|
117 |
+
action_btn = gr.Button(value="Find Matches")
|
118 |
+
|
119 |
+
with gr.Tab("Authors"):
|
120 |
+
n_output = gr.Dataframe(
|
121 |
+
headers=['No.', 'Score', 'Name', 'Title', 'Link', 'Date'],
|
122 |
+
datatype=['number', 'number', 'str', 'str', 'str', 'str'],
|
123 |
+
col_count=(6, "fixed"),
|
124 |
+
wrap=True,
|
125 |
+
visible=False
|
126 |
+
)
|
127 |
+
with gr.Tab("Abstracts"):
|
128 |
+
a_output = gr.Dataframe(
|
129 |
+
headers=['No.', 'Title', 'Author', 'Corresponding Author',
|
130 |
+
'Journal', 'Date', 'Link', 'Score'],
|
131 |
+
datatype=['number', 'str', 'str', 'str',
|
132 |
+
'str', 'str', 'str', 'number'],
|
133 |
+
col_count=(8, "fixed"),
|
134 |
+
wrap=True,
|
135 |
+
visible=False
|
136 |
+
)
|
137 |
+
with gr.Tab("Journals"):
|
138 |
+
j_output = gr.Dataframe(
|
139 |
+
headers=['No.', 'Name', 'Score'],
|
140 |
+
datatype=['number', 'str', 'number'],
|
141 |
+
col_count=(3, "fixed"),
|
142 |
+
wrap=True,
|
143 |
+
visible=False
|
144 |
+
)
|
145 |
+
|
146 |
+
action_btn.click(fn=inference,
|
147 |
+
inputs=[
|
148 |
+
abst,
|
149 |
+
k,
|
150 |
+
],
|
151 |
+
outputs=[a_output, j_output, n_output],
|
152 |
+
api_name="neurojane")
|
153 |
+
|
154 |
+
demo.launch(debug=True)
|
requirements.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
sentence-transformers
|
2 |
+
torch
|
3 |
+
datasets
|
4 |
+
sentencepiece
|
5 |
+
langchain
|
6 |
+
faiss-cpu
|
7 |
+
accelerate
|