Upload 3 files

.gitattributes

@@ -32,3 +32,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+earnings_calls_sentencewise.csv filter=lfs diff=lfs merge=lfs -text

app.py

@@ -12,10 +12,13 @@ from transformers import (
 import streamlit as st
 import openai
 
+@st.experimental_singleton
+def get_data():
+    data = pd.read_csv("earnings_calls_sentencewise.csv")
+    return data
 
 # Initialize models from HuggingFace
 
-
 @st.experimental_singleton
 def get_t5_model():
     return pipeline("summarization", model="t5-small", tokenizer="t5-small")
@@ -66,6 +69,26 @@ def format_query(query_results):
     context = [result["metadata"]["Text"] for result in query_results["matches"]]
     return context
 
+def sentence_id_combine(data, query_results, lag=2):
+    # Extract sentence IDs from query results
+    ids = [result["metadata"]["Sentence_id"] for result in query_results["matches"]]
+    # Generate new IDs by adding a lag value to the original IDs
+    new_ids = [id + i for id in ids for i in range(-lag, lag + 1)]
+    # Remove duplicates and sort the new IDs
+    new_ids = sorted(set(new_ids))
+    # Create a list of lookup IDs by grouping the new IDs in groups of lag*2+1
+    lookup_ids = [
+        new_ids[i : i + (lag * 2 + 1)] for i in range(0, len(new_ids), lag * 2 + 1)
+    ]
+    # Create a list of context sentences by joining the sentences corresponding to the lookup IDs
+    context_list = [
+        " ".join(data.Text.iloc[lookup_id].to_list()) for lookup_id in lookup_ids
+    ]
+    return context_list
+
+def text_lookup(data, sentence_ids):
+    context = " ".join(data.iloc[sentence_ids].to_list())
+    return context
 
 def gpt3_summary(text):
     response = openai.Completion.create(
@@ -136,7 +159,12 @@ elif encoder_model == "SGPT":
 
 query_results = query_pinecone(query_text, num_results, retriever_model, pinecone_index)
 
-context_list = format_query(query_results)
+window = int(st.number_input("Sentence Window Size", 1, 3, value=1))
+
+data = get_data()
+
+#context_list = format_query(query_results)
+context_list = sentence_id_combine(data, query_results, lag=window)
 
 
 st.subheader("Answer:")

earnings_calls_sentencewise.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a24373c9cb8d68b4681f7590b5d94916ef748bd259636d93728e99b8e50678a5
+size 12926317