From d86b2a97aa02e3b68e1a25f565554f9239f384b1 Mon Sep 17 00:00:00 2001
From: Hao Chen
Date: Fri, 10 Apr 2020 10:23:37 -0500
Subject: maybe more efficient

---
 nlp.py    | 33 +++++++++++++++++----------------
 server.py | 37 +++++++++++++++++++++----------------
 2 files changed, 38 insertions(+), 32 deletions(-)

diff --git a/nlp.py b/nlp.py
index e54713e..9f36b58 100644
--- a/nlp.py
+++ b/nlp.py
@@ -48,18 +48,23 @@ with open('./nlp/vocabulary.txt', 'r') as vocab:
     vocab = vocab.read()
 
 # create the CNN model
-def create_model(vocab_size, max_length):
-    model = Sequential()
-    model.add(Embedding(vocab_size, 32, input_length=max_length))
-    model.add(Conv1D(filters=16, kernel_size=4, activation='relu'))
-    model.add(MaxPooling1D(pool_size=2))
-    model.add(Flatten())
-    model.add(Dense(10, activation='relu'))
-    model.add(Dense(1, activation='sigmoid'))
-    opt = keras.optimizers.Adamax(learning_rate=0.002, beta_1=0.9, beta_2=0.999)
+#def create_model(vocab_size, max_length):
+model = Sequential()
+model.add(Embedding(vocab_size, 32, input_length=max_length))
+model.add(Conv1D(filters=16, kernel_size=4, activation='relu'))
+model.add(MaxPooling1D(pool_size=2))
+model.add(Flatten())
+model.add(Dense(10, activation='relu'))
+model.add(Dense(1, activation='sigmoid'))
+opt = keras.optimizers.Adamax(learning_rate=0.002, beta_1=0.9, beta_2=0.999)
+model.compile(loss='binary_crossentropy', optimizer=opt, metrics=[keras.metrics.AUC()])
+model = create_model(23154, 64)
+# load the weights
+## this is done for every prediction??
+checkpoint_path = "./nlp/weights.ckpt"
+model.load_weights(checkpoint_path)
 
-    model.compile(loss='binary_crossentropy', optimizer=opt, metrics=[keras.metrics.AUC()])
-    return model
+#return model
 
 def predict_sent(sent_for_pred):
     max_length = 64
@@ -70,13 +75,9 @@ def predict_sent(sent_for_pred):
     line = [line]
     tokenized_sent = tokenizer.texts_to_sequences(line)
     tokenized_sent = pad_sequences(tokenized_sent, maxlen=max_length, padding='post')
-    model = create_model(23154, 64)
-    # load the weights
-    checkpoint_path = "./nlp/weights.ckpt"
-    model.load_weights(checkpoint_path)
     predict_sent = model.predict(tokenized_sent, verbose=0)
     percent_sent = predict_sent[0,0]
     if round(percent_sent) == 0:
         return 'neg'
     else:
-        return 'pos'
\ No newline at end of file
+        return 'pos'
diff --git a/server.py b/server.py
index f16d930..f36a4b3 100755
--- a/server.py
+++ b/server.py
@@ -541,11 +541,14 @@ def sentences():
     edge=request.args.get('edgeID')
     (tf_name, gene0, cat0)=edge.split("|")
     out3=""
-    out5_pl=""
-    out5_sn=""
+#    out5_pl=""
+#    out5_sn=""
     out_pos = ""
     out_neg = ""
     num_abstract = 0
+    stress_systemic = "<br><br><br><hr>"+"<b>Sentence(s) describing celluar stress (classified using a deep learning model):</b>"
+    stress_cellular = "<b>Sentence(s) describing systemic stress (classified using a deep learning model):</b>"
+    stress_sents={}
     with open(tf_name, "r") as df:
         all_sents=df.read()
     for sent in all_sents.split("\n"):
@@ -555,29 +558,31 @@ def sentences():
                 out3+= "<li> "+ text + " <a href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=" + pmid +"\" target=_new>PMID:"+pmid+"<br></a>"
                 num_abstract += 1
                 if(pmid+cat0 not in pmid_list):
-                    pmid_list.append(pmid+cat0)                
+                    pmid_list.append(pmid+cat0)
                 if(cat0=='stress'):
-                    out5_pl = 'These are analyzed by deep learning to seperate the relevant sentences.'
-                    out5_sn = 'This is analyzed by deep learning to see whether it is relevant or not.'
+#                    out5_pl = 'These are analyzed by deep learning to seperate the relevant sentences.'
+#                    out5_sn = 'This is analyzed by deep learning to see whether it is relevant or not.'
                     out_pred = "<li> "+ text + " <a href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=" + pmid +"\" target=_new>PMID:"+pmid+"<br></a>"
+                    #should we add the html part after the predict_sent function?
                     out4 = predict_sent(out_pred)
-                    if(out4 == 'pos'):
-                        out_rel = "<b>Relevant sentences:</b>"
-                        out_pos += out_pred
-                    else:
-                        out_irrel = "<br><br><br><hr>"+"<b>Irrelevant sentences:</b>"
-                        out_neg += out_pred
+                    stress_sents[out4] +=stress_sents[out4]
+#                    stress_sents["pos"]+=stress_sents["pos"]
+#                    stress_sents["neg"]+=stress_sents["neg"]
+#                    if(out4 == 'pos'):
+#                        out_pos += out_pred
+#                    else:
+#                        out_neg += out_pred
     out1="<h3>"+gene0 + " and " + cat0  + "</h3>\n"
     if len(pmid_list)>1:
         out2 = str(num_abstract) + ' sentences in ' + str(len(pmid_list)) + ' studies' + "<br>"
-        if(out5_pl!=""):
-            out2 += out5_pl
+#        if(out5_pl!=""):
+#            out2 += out5_pl
         out2 += "<hr>\n"
     else:
         out2 = str(num_abstract) + ' sentence in ' + str(len(pmid_list)) + ' study' "<br>"
-        if(out5_sn!=""):
-            out2 += out5_sn    
-        out2 += "<hr>\n"  
+#        if(out5_sn!=""):
+#            out2 += out5_sn
+        out2 += "<hr>\n"
     if(out_neg == "" and out_pos == ""):
         out= out1+ out2 +out3
     elif(out_pos != "" and out_neg!=""):
-- 
cgit v1.2.3