From d258640ab315e7f598415a6d0047f36364356dd6 Mon Sep 17 00:00:00 2001
From: Blink The Things <blinkthethings@gmail.com>
Date: Tue, 3 Nov 2020 12:20:10 -0500
Subject: [PATCH] Add syntactic dependency and word position to state

---
 markov.py | 23 ++++++++++++++++++++---
 1 file changed, 20 insertions(+), 3 deletions(-)

diff --git a/markov.py b/markov.py
index 40a9266..b46a9de 100644
--- a/markov.py
+++ b/markov.py
@@ -41,18 +41,30 @@ words_doc = nlp(words_text)
 
 words = {}
 for sent in words_doc.sents:
+    cnt = 0
     for token in sent:
         if token.pos_ in ('SPACE', 'PUNCT', 'X'):
             continue
 
-        state = token.tag_
+        cnt += 1
+
         word = token.text
 
+        state = f'{token.tag_},{token.dep_}'
+
         if state in words:
             words[state].append(word)
         else:
             words[state] = [word]
 
+        state = f'{token.tag_},{token.dep_},{str(cnt)}'
+
+        if state in words:
+            words[state].append(word)
+        else:
+            words[state] = [word]
+
+
 pos_text = ''
 with open(args.pos_file, mode='r') as f:
     pos_text = f.read()
@@ -63,11 +75,13 @@ edges = []
 for sent in pos_doc.sents:
     curr_state = 'START'
 
+    cnt = 0
     for token in sent:
         if token.pos_ in ('SPACE', 'PUNCT', 'X'):
             continue
 
-        next_state = token.tag_
+        cnt += 1
+        next_state = f'{token.tag_},{token.dep_},{str(cnt)}'
 
         edges.append((curr_state, next_state))
 
@@ -110,7 +124,10 @@ for _ in range(10):
             sents.append(' '.join(choices))
             break
 
-        word = rng.choice(words[next_choice])
+        try:
+            word = rng.choice(words[next_choice])
+        except KeyError:
+            word = rng.choice(words[','.join(next_choice.split(',')[:-1])])
 
         choices.append(word)