Posts SciBERT Transformer for Neuroscience
Post
Cancel

SciBERT Transformer for Neuroscience

Mrunal Jul 14 2022-07-14T00:00:00-07:00
Jul 28 2022-07-28T20:54:34-07:00 11 min

Hugging Face Transformer model for Neuroscience JSON data

The objective of the project is to apply pretrained SciBERT transformer model and Cosine Similarity for recommending reviewers who have published neuroscience research papers on semantically similar research topic as the user’s input abstract query.

Overall Approach

  1. Load the pretrained SciBert model and tokenizer
  2. Vectorize documents by creating embeddings
  3. Semantic Similarity search by Cosine Similarity

For the purpose of creating embeddings, the bioRxiv Neuroscience data is used.

Step 1 : Import Libraries:

1
2
3
4
5
6
7
8
9
10
11
12
13
14

import pandas as pd
import numpy as np
from tqdm import tqdm
import warnings

# Hugging Face Transformer libraries
!pip install transformers
import torch
from transformers import BertTokenizer,  AutoModelForSequenceClassification

# Similarity search: cosine similarity search 
from sklearn.metrics.pairwise import cosine_similarity

warnings.filterwarnings("ignore")

Step 2: load the bioRxiv Neuroscience data in JSON format

1
2
3

data = pd.read_json("bioarxiv_parsed.json") 
print("Data Shape: {}".format(data.shape))

Step 3: Load Pretrained SciBERT model and tokenizer

The SciBERT model is used for creating embeddings for the abstracts in the Neuroscience research papers. Note that in the code snippet below the output_hidden_states is set to True so that we can extract the embeddings.

1
2
3
4
5
6
7
8
9
10
11

# Get the SciBERT pretrained model path from Allen AI repo
pretrained_model = 'allenai/scibert_scivocab_uncased'

# Get the tokenizer from the previous path
sciBERT_tokenizer = BertTokenizer.from_pretrained(pretrained_model, 
                                          do_lower_case=True)

# Get the model
model = AutoModelForSequenceClassification.from_pretrained(pretrained_model,
                                                          output_attentions=False,
                                                          output_hidden_states=True)

model.eval() gives the architecture of the model as seen below:

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
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310

BertForSequenceClassification(
  (bert): BertModel(
    (embeddings): BertEmbeddings(
      (word_embeddings): Embedding(31090, 768, padding_idx=0)
      (position_embeddings): Embedding(512, 768)
      (token_type_embeddings): Embedding(2, 768)
      (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
      (dropout): Dropout(p=0.1, inplace=False)
    )
    (encoder): BertEncoder(
      (layer): ModuleList(
        (0): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (1): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (2): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (3): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (4): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (5): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (6): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (7): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (8): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (9): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (10): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (11): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
            (intermediate_act_fn): GELUActivation()
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
      )
    )
    (pooler): BertPooler(
      (dense): Linear(in_features=768, out_features=768, bias=True)
      (activation): Tanh()
    )
  )
  (dropout): Dropout(p=0.1, inplace=False)
  (classifier): Linear(in_features=768, out_features=2, bias=True)
)

Step 4: Create an embedding for a given text data using SciBERT pre-trained model

This function convert_single_abstract_to_embedding is mostly inspired by the BERT Word Embeddings Tutorial of Chris McCormick and Zoumana Keita. It aims to create an embedding for a given text data using a pre-trained model.

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

def convert_single_abstract_to_embedding(tokenizer, model, in_text, MAX_LEN = 510):
    
    input_ids = tokenizer.encode(
                        in_text, 
                        add_special_tokens = True, 
                        max_length = MAX_LEN,                           
                   )    

    results = pad_sequences([input_ids], maxlen=MAX_LEN, dtype="long", 
                              truncating="post", padding="post")
    
    # Remove the outer list.
    input_ids = results[0]

    # Create attention masks    
    attention_mask = [int(i>0) for i in input_ids]
    
    # Convert to tensors.
    input_ids = torch.tensor(input_ids)
    attention_mask = torch.tensor(attention_mask)

    # Add an extra dimension for the "batch" (even though there is only one 
    # input in this batch.)
    input_ids = input_ids.unsqueeze(0)
    attention_mask = attention_mask.unsqueeze(0)
    
    # Put the model in "evaluation" mode, meaning feed-forward operation.
    model.eval()

 
    # Run the text through BERT, and collect all of the hidden states produced
    # from all 12 layers. 
    with torch.no_grad():        
        logits, encoded_layers = model(
                                    input_ids = input_ids, 
                                    token_type_ids = None, 
                                    attention_mask = attention_mask,
                                    return_dict=False)

    layer_i = 12 # The last BERT layer before the classifier.
    batch_i = 0 # Only one input in the batch.
    token_i = 0 # The first token, corresponding to [CLS]
        
    # Extract the embedding.
    embedding = encoded_layers[layer_i][batch_i][token_i]

    # Move to the CPU and convert to numpy ndarray.
    embedding = embedding.detach().cpu().numpy()

    return(embedding)

Now we can use the model and tokenizer to generate an embedding for the 3rd input_abstract as a way of testing as seen in the code snippet below:

1
2
3
4
5
6
7
8

    from keras_preprocessing.sequence import pad_sequences

    input_abstract = data.abstract.iloc[3]

    abstract_embedding = convert_single_abstract_to_embedding(sciBERT_tokenizer, model, input_abstract)

    print('Embedding shape: {}'.format(abstract_embedding.shape))

Note to run the above code snippet make sure you have installed keras and tensorflow. You can install both of them in the jupyter notebook in the following way:

1
2
3

    !pip install keras
    
    !pip3 install tensorflow

The output of the 3rd input abstract embedding shape is:

1
2

    Embedding shape: (768,)
    Embedding is composed of 768 values.

Step 5: Create Embedding for all the abstracts

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

   def convert_all_abstract_text_to_embedding(df):
    
    # The list of all the embeddings
    embeddings = []
    
    # Get overall text data
    overall_text_data = data.abstract.values
    
    # Loop over all the comment and get the embeddings
    for abstract in tqdm(overall_text_data):
        
        # Get the embedding 
        embedding = convert_single_abstract_to_embedding(sciBERT_tokenizer, model, abstract)
        
        #add it to the list
        embeddings.append(embedding)
        
    print("Conversion Done!")
    
    return embeddings

Note that creating embeddings for all the abstracts in the 3948 BioRxiv Neuroscience research papers takes atleast 2 hours in aws sagemaker.

1
2

   # This task can take a lot of time depending on the sample_size value 
   embeddings = convert_all_abstract_text_to_embedding(data)

Step 6: Save the embeddings for future use

1
2
3

   
   embeddings = np.array(embeddings)
   np.save('embeddings.npy', embeddings)

Step 7: Load the saved .npy embeddings

   embeddings = np.load('embeddings.npy')

Step 8: Create a new column that will contain embedding of each body text

1
2
3
4
5
6
7
8
9

   
def create_final_embeddings(df, embeddings):
    
    df["embeddings"] = embeddings
    df["embeddings"] = df["embeddings"].apply(lambda emb: np.array(emb))
    df["embeddings"] = df["embeddings"].apply(lambda emb: emb.reshape(1, -1))
    
    return df

To see the output:

    data = create_final_embeddings(data, embeddings)
    data.head(3)

embeddigns column

References

  1. @inproceedings{beltagy-etal-2019-scibert, title = “SciBERT: A Pretrained Language Model for Scientific Text”, author = “Beltagy, Iz and Lo, Kyle and Cohan, Arman”, booktitle = “EMNLP”, year = “2019”, publisher = “Association for Computational Linguistics”, url = “https://www.aclweb.org/anthology/D19-1371” }

  2. @article{johnson2019billion, title={Billion-scale similarity search with {GPUs}}, author={Johnson, Jeff and Douze, Matthijs and J{'e}gou, Herv{'e}}, journal={IEEE Transactions on Big Data}, volume={7}, number={3}, pages={535–547}, year={2019}, publisher={IEEE} }

  3. “Bert Word Embeddings Tutorial.” BERT Word Embeddings Tutorial · Chris McCormick, 14 May 2019, https://mccormickml.com/2019/05/14/BERT-word-embeddings-tutorial/#3-extracting-embeddings.

  4. Keita, Zoumana. “Scientific Documents Similarity Search with Deep Learning Using Transformers (Scibert).” Medium, Towards Data Science, 17 Jan. 2022, https://towardsdatascience.com/scientific-documents-similarity-search-with-deep-learning-using-transformers-scibert-d47c4e501590.

  5. @article{Beltagy2020Longformer, title={Longformer: The Long-Document Transformer}, author={Iz Beltagy and Matthew E. Peters and Arman Cohan}, journal={arXiv:2004.05150}, year={2020}, }

GSoC Blogging
GSoC
This post is licensed under CC BY 4.0 by the author.
Recent Update
Trending Tags
NLP GSoC Evaluation Report
Contents

PubMed Data

Vector Similarity Search

Trending Tags

NLP GSoC Evaluation Report