Update app.py

app.py

@@ -195,7 +195,7 @@ def getTopic(encoded_input):
         logits = modelTopic(**encoded_input).logits
     predicted_class_id = logits.argmax().item()
     return modelTopic.config.id2label[predicted_class_id]
-def selectedCorpusForNextQuarterModel(x,quarter):
+def selectedCorpusForNextQuarterModel(x,quarter,year):
     number_word_dict = {
     "1": "first",
     "2": "second",
@@ -205,9 +205,9 @@ def selectedCorpusForNextQuarterModel(x,quarter):
     }
     tokens=tokenizerTopic(x, padding=True, truncation=True, return_tensors='pt')
     splitSize=256
-    chunksInput_ids=[tokens["input_ids"][r*splitSize:(r+1)*splitSize] for r in range(math.ceil(len(tokens["input_ids"])/splitSize))]
-    chunksToken_type_ids=[tokens["token_type_ids"][r*splitSize:(r+1)*splitSize] for r in range(math.ceil(len(tokens["token_type_ids"])/splitSize))]
-    chunksAttention_mask=[tokens["attention_mask"][r*splitSize:(r+1)*splitSize] for r in range(math.ceil(len(tokens["attention_mask"])/splitSize))]
+    chunksInput_ids=[tokens["input_ids"][0][r*splitSize:(r+1)*splitSize] for r in range(math.ceil(len(tokens["input_ids"][0])/splitSize))]
+    chunksToken_type_ids=[tokens["token_type_ids"][0][r*splitSize:(r+1)*splitSize] for r in range(math.ceil(len(tokens["token_type_ids"][0])/splitSize))]
+    chunksAttention_mask=[tokens["attention_mask"][0][r*splitSize:(r+1)*splitSize] for r in range(math.ceil(len(tokens["attention_mask"][0])/splitSize))]
     l=[]
     for idx in range(len(chunksInput_ids)):
         l.append({"input_ids":torch.tensor([list(chunksInput_ids[idx])]).to("cuda"),
@@ -218,16 +218,17 @@ def selectedCorpusForNextQuarterModel(x,quarter):
     selectedTopics = ["Stock Movement", "Earnings", "IPO", "Stock Commentary", "Currencies", "M&A | Investments", "Financials", "Macro", "Analyst Update", "Company | Product News"]
     result = [tokenizerTopic.decode(x["input_ids"][0], skip_special_tokens=True) for x in l if getTopic(x) in selectedTopics]
     result=[x for x in result if len(x)>10]
-    des=f"the {number_word_dict[str(quarter)]} quarter results of the {usedData['quad-date'].iloc[i]}"
-    courpus=result.split("\n")
+    des=f"the {number_word_dict[str(quarter)]} quarter results of the {year}"
+    courpus=result
     embeddings_1 = similarityModel.encode([des]+courpus, normalize_embeddings=True,device='cuda',show_progress_bar=False) 
     sents=[des]+courpus
     rest=[sents[f] for f in [list(cosine_similarity(embeddings_1)[0][1:]).index(value)+1 for value in sorted(list(cosine_similarity(embeddings_1)[0][1:]),reverse=True)][:3]]
-    return selectedCourpusForTraing.append(",".join(rest))
+    return ",".join(rest)
 
 st.header("Transcript Analysis", divider='rainbow')
 mainTranscript = st.text_area("Enter the transcript:", height=100)
-quarter = st.text_input('Enter your quarter', 'quarter of transcript')
+quarter = st.text_input('Enter your quarter', '')
+year = st.text_input('Enter your quarter', '')
 if st.button("Analyze"):
     transcript=replace_abbreviations(mainTranscript)
     transcript=replace_abbreviations(transcript)
@@ -245,7 +246,7 @@ if st.button("Analyze"):
     st.markdown(f'<span style="color:{sentiment_color}">{sentiment}</span>', unsafe_allow_html=True)
     st.subheader("Next Quarter Perdiction", divider='rainbow')
     # increase_decrease = [increase_decrease_model(x)[0]['label'] for x in chunks]
-    increase_decrease=increase_decrease_model(selectedCorpusForNextQuarterModel(mainTranscript,quarter))[0]['label']
+    increase_decrease=increase_decrease_model(selectedCorpusForNextQuarterModel(mainTranscript,quarter,year))[0]['label']
     increase_decrease=max(increase_decrease,key=increase_decrease.count)
     increase_decrease_color = "green" if increase_decrease == "Increase" else "red"
     st.markdown(f'<span style="color:{increase_decrease_color}">{increase_decrease}</span>', unsafe_allow_html=True)