adding rectangles on chart showing steep derivatives

app.py

@@ -7,6 +7,7 @@ import argparse
 
 import gensim.downloader as api
 import matplotlib.pyplot as plt
+import matplotlib.patches as patches
 import nltk
 import numpy as np
 import pandas as pd
@@ -183,6 +184,47 @@ def stats(text):
     results = readability.getmeasures(text, lang='en')
     return results
 
+def derive(x:list, y:list):
+  all_derivs = []
+  for idx, point in enumerate(x):
+    if idx != len(x) - 1:
+      next_x = x[idx + 1]
+      next_y = y[idx + 1]
+      h = next_x - point
+      if h != 0:
+        deriv = (next_y - y[idx])/h
+      else:
+        deriv = 0
+      all_derivs.append(abs(deriv))
+  return all_derivs
+    #(f(x+h) - f(x))/h
+
+
+def generate_patches(x:list, y:list, range, deriv_threshold=2):
+  derivs = derive(x,y)
+  print('derivs', derivs)
+  in_patch = False
+  patches = []
+  start = []
+  end = []
+  for idx, der in enumerate(derivs):
+    if der > deriv_threshold:
+      if not in_patch: 
+        start.append(x[idx])
+        in_patch = True
+    else:
+      if in_patch:
+        end.append(x[idx])
+        in_patch = False
+      else:
+        continue
+
+  print(start, end)
+  if len(start) != len(end): 
+    #not doing len(x)-1 because the derivitive can't be taken at ending point so in derive() the x length is already -1 of original
+    end.append(len(x))
+  return list(zip(start,end))
+
 
 def predict(text, tokenizer=tokenizer):
     model.eval()
@@ -391,8 +433,14 @@ def sliding_window(text):
     plt.suptitle('Difficulty Score Across Text', fontsize=14, fontweight='bold')
     plt.style.use('ggplot')
     ax.set_facecolor('w')
-    fig = plt.gcf()
+    shaded_areas = generate_patches(x, y, .42)
 
+    for area in shaded_areas:
+        print(range_chart[0], range_chart[1])
+        ax.add_patch(patches.Rectangle((area[0],range_chart[0]), area[1]-area[0], range_chart[1]-range_chart[0], alpha=0.2))
+    
+    fig = plt.gcf()
+        
     mapd = [('', 0)]
     maxy = max(inter_scores)
     miny = min(inter_scores)