Update app.py

Code modified to output the simplified sentence using a DAG.

app.py

@@ -49,6 +49,8 @@ from collections import defaultdict
 
 import html
 
+from graphviz import Digraph
+
 PRONOUNS = {
     "he", "she", "it", "they", "him", "her", "them",
     "his", "hers", "its", "their", "theirs",
@@ -2731,6 +2733,57 @@ def expand_paren_lists(sent: str, lists: list[list[str]]) -> list[str]:
     return outs
 ###############################################################################
 
+###############################################################################
+def build_shared_dag(sentences: list[str]):
+    """
+    Nodes are keyed by the remaining word tuple, so any common
+    suffix (or whole remaining span) is a single node.
+    """
+    cache = {}          # tuple[str, ...] -> node_id
+    labels = {}         # node_id -> str
+    edges = set()       # (src, dst)
+
+    def node_for(words: tuple[str, ...]) -> str:
+        if words in cache:
+            return cache[words]
+        nid = f"n{len(cache)}"
+        cache[words] = nid
+        labels[nid] = " ".join(words) if words else "•"
+        return nid
+
+    root = node_for(())
+    for sent in sentences:
+        words = tuple(sent.rstrip(".").split())
+        prev = root
+        for i in range(len(words)):
+            # edge into the shared node for the remaining span
+            nxt = node_for(words[i:])
+            edges.add((prev, nxt))
+            prev = nxt
+        edges.add((prev, node_for(())))  # terminal
+
+    return labels, edges
+
+###############################################################################
+def render_dag(sentences: list[str], path: str = "simp_dag") -> str:
+    labels, edges = build_shared_dag(sentences)
+    g = Digraph("simp", format="svg")
+    g.attr(rankdir="TB", bgcolor="white")
+    g.attr("node", shape="box", style="rounded,filled",
+           fillcolor="#f7f7f7", fontname="Helvetica", fontsize="11")
+    g.attr("edge", color="#555555")
+    for nid, lab in labels.items():
+        if lab == "•":
+            continue
+        g.node(nid, lab)
+    for a, b in edges:
+        if labels.get(a) == "•" or labels.get(b) == "•":
+            continue
+        g.edge(a, b)
+    g.render(path, cleanup=True)
+    return path + ".svg"
+###############################################################################
+
 ###############################################################################
 @spaces.GPU(duration=120)
 def process_text(input_text):
@@ -3033,6 +3086,12 @@ def process_text(input_text):
             if para_simplified:
                 para_simplified = order_sentences_for_discourse(para_simplified)
                 output_paragraphs.append(" ".join(para_simplified))
+
+            svg_path = render_dag(para_simplified)
+            # return the SVG (or a PNG) instead of, or beside, the flat text
+            with open(svg_path) as f:
+                svg = f.read()
+            return f'<div class="simp-out">{svg}</div>'
         
         final_output = "\n\n".join(output_paragraphs)
         print(f"\n=== FINAL OUTPUT ===\n{final_output}")