Update src/streamlit_app.py

src/streamlit_app.py

@@ -1,28 +1,105 @@
+# import streamlit as st
+# import nltk
+# import spacy
+# import benepar
+# from nltk import Tree
+
+# # Configure nltk to use /tmp
+# nltk_data_path = "/tmp/nltk_data"
+# nltk.data.path.append(nltk_data_path)
+# nltk.download('punkt', download_dir=nltk_data_path)
+
+# # Load installed spaCy model
+# nlp = spacy.load("en_core_web_sm")
+
+# # Add benepar parser
+# if "benepar" not in nlp.pipe_names:
+#     benepar.download("benepar_en3")
+#     nlp.add_pipe("benepar", config={"model": "benepar_en3"})
+# # Streamlit UI
+# st.set_page_config(page_title="Syntax Parser Comparison", layout="wide")
+# st.title("🌐 Syntax Parser Comparison Tool")
+# st.write("This tool compares Dependency Parsing, Constituency Parsing, and a simulated Abstract Syntax Representation (ASR).")
+
+# # Input
+# sentence = st.text_input("Enter a sentence:", "John eats an apple.")
+
+# if sentence:
+#     doc = nlp(sentence)
+#     sent = list(doc.sents)[0]
+
+#     col1, col2, col3 = st.columns(3)
+
+#     with col1:
+#         st.header("Dependency Parsing")
+#         for token in sent:
+#             st.write(f"{token.text} --> {token.dep_} --> {token.head.text}")
+#         st.code(" ".join(f"({token.text}, {token.dep_}, {token.head.text})" for token in sent))
+
+#     with col2:
+#         st.header("Constituency Parsing")
+#         tree = sent._.parse_string
+#         st.text(tree)
+#         st.code(Tree.fromstring(tree).pformat())
+
+#     with col3:
+#         st.header("Simulated ASR Output")
+#         st.write("Combining phrase structure with dependency head annotations:")
+#         for token in sent:
+#             if token.dep_ in ("nsubj", "obj", "det", "ROOT"):
+#                 st.write(f"[{token.text}] - {token.dep_} --> {token.head.text} ({token.pos_})")
+#         st.markdown("_(ASR is simulated by combining POS tags, dependency heads, and phrase information.)_")
+#         st.code(" ".join(f"[{token.text}: {token.dep_} → {token.head.text}]({token.pos_})" for token in sent))
+
+
+
 import streamlit as st
 import nltk
 import spacy
 import benepar
 from nltk import Tree
+from graphviz import Digraph
 
-# Configure nltk to use /tmp
+# Configure NLTK to use /tmp for downloads
 nltk_data_path = "/tmp/nltk_data"
 nltk.data.path.append(nltk_data_path)
 nltk.download('punkt', download_dir=nltk_data_path)
 
-# Load installed spaCy model
+# Load spaCy model
 nlp = spacy.load("en_core_web_sm")
 
-# Add benepar parser
+# Add benepar if not already added
 if "benepar" not in nlp.pipe_names:
     benepar.download("benepar_en3")
     nlp.add_pipe("benepar", config={"model": "benepar_en3"})
-# Streamlit UI
-st.set_page_config(page_title="Syntax Parser Comparison", layout="wide")
+
+# Streamlit page config
+st.set_page_config(
+    page_title="🌐 Syntax Parser Comparison Tool",
+    page_icon="📝",
+    layout="wide"
+)
+
+# Sidebar Info
+st.sidebar.title("ℹ️ About This Tool")
+st.sidebar.write("""
+Compare **Dependency Parsing**, **Constituency Parsing**,  
+and a simulated **Abstract Syntax Representation (ASR)**.
+""")
+st.sidebar.markdown("---")
+st.sidebar.info("💡 Enter a sentence in the input box to see all 3 parses.")
+
+# Title
 st.title("🌐 Syntax Parser Comparison Tool")
-st.write("This tool compares Dependency Parsing, Constituency Parsing, and a simulated Abstract Syntax Representation (ASR).")
+st.markdown("""
+This tool demonstrates **three parsing styles** side-by-side:
+1. **Dependency Parsing** – Shows head–dependent word relationships.
+2. **Constituency Parsing** – Displays hierarchical phrase structures.
+3. **Abstract Syntax Representation (ASR)** – Merges phrase structure with dependency info.
+""")
 
-# Input
-sentence = st.text_input("Enter a sentence:", "John eats an apple.")
+# Input sentence
+sentence = st.text_input("✏️ Enter a sentence:", "John eats an apple.")
 
 if sentence:
     doc = nlp(sentence)
@@ -30,23 +107,39 @@ if sentence:
 
     col1, col2, col3 = st.columns(3)
 
+    # 1️⃣ Dependency Parsing
     with col1:
-        st.header("Dependency Parsing")
+        st.subheader("🔗 Dependency Parsing")
+        dep_graph = Digraph()
+        dep_graph.attr(rankdir="TB")
         for token in sent:
-            st.write(f"{token.text} --> {token.dep_} --> {token.head.text}")
-        st.code(" ".join(f"({token.text}, {token.dep_}, {token.head.text})" for token in sent))
+            dep_graph.node(token.text, f"{token.text}\n({token.dep_})")
+            if token.head != token:
+                dep_graph.edge(token.head.text, token.text)
+        st.graphviz_chart(dep_graph)
+        with st.expander("Raw Dependency Tuples"):
+            st.code(" ".join(f"({t.text}, {t.dep_}, {t.head.text})" for t in sent))
 
+    # 2️⃣ Constituency Parsing
     with col2:
-        st.header("Constituency Parsing")
-        tree = sent._.parse_string
-        st.text(tree)
-        st.code(Tree.fromstring(tree).pformat())
+        st.subheader("🌳 Constituency Parsing")
+        tree_str = sent._.parse_string
+        with st.expander("Tree String"):
+            st.text(tree_str)
+        st.code(Tree.fromstring(tree_str).pformat(), language="text")
 
+    # 3️⃣ Simulated ASR
     with col3:
-        st.header("Simulated ASR Output")
-        st.write("Combining phrase structure with dependency head annotations:")
+        st.subheader("🧩 Simulated ASR Output")
+        st.markdown("Combines **dependency heads**, **POS tags**, and **phrase info**.")
+        highlighted_output = []
         for token in sent:
             if token.dep_ in ("nsubj", "obj", "det", "ROOT"):
-                st.write(f"[{token.text}] - {token.dep_} --> {token.head.text} ({token.pos_})")
-        st.markdown("_(ASR is simulated by combining POS tags, dependency heads, and phrase information.)_")
-        st.code(" ".join(f"[{token.text}: {token.dep_} → {token.head.text}]({token.pos_})" for token in sent))
+                highlighted_output.append(
+                    f"**[{token.text}]** - {token.dep_} → {token.head.text} ({token.pos_})"
+                )
+        st.write("\n".join(highlighted_output))
+        with st.expander("ASR Encoded String"):
+            st.code(
+                " ".join(f"[{t.text}: {t.dep_} → {t.head.text}]({t.pos_})" for t in sent)
+            )