diff --git a/CommentTuParles/parles.py b/CommentTuParles/parles.py
index f4d3361..310f120 100644
--- a/CommentTuParles/parles.py
+++ b/CommentTuParles/parles.py
@@ -1,6 +1,9 @@
 import json
+import operator
 import os
-from collections import defaultdict
+from collections import defaultdict, Counter
+
+import spacy
 
 
 def are_invalid(lyrics: str):
@@ -16,8 +19,7 @@ def are_invalid(lyrics: str):
 def cleanup(lyrics):
     lines = lyrics.split("\n")
     filtered = [l for l in lines if
-                l is not None and
-                len(l) and
+                l is not None and len(l) and
                 l[0] not in ["#", "["] and
                 "Paroles de la chanson" not in l and
                 "Paroles de Même" not in l]
@@ -27,33 +29,64 @@ def cleanup(lyrics):
 
 def analyse():
     path = "../data/"
-    files = os.listdir(path)
     stats = {
         "avg_lines": defaultdict(lambda: 0),
         "avg_newlines": defaultdict(lambda: 0),
         "missing": defaultdict(lambda: 0)
     }
 
+    print("Go")
+    files = os.listdir(path)
+    print("Listed data files.")
+    nlp = spacy.load("fr_core_news_md")
+    print("Loaded spaCy.")
+
     for filename in files:
         with open(path + filename) as f:
             content = json.load(f)
+            print("Loaded", filename)
             artist = content["name"]
             songs = content["songs"]
 
             print("\n## %s ##" % artist)
+            keywords = []
 
+            occurences = {}
             for song in songs:
-                # title = song["title"]
                 if song["lyrics"] is not None:
+                    keyword = "anar"
+                    occurences[song["title"]] = song["lyrics"].lower().count(keyword)
+
+            if any([i > 0 for i in occurences.values()]):
+                print("%s chez %s: %i" % (keyword, artist, len(occurences)))
+                for pair in sorted(occurences.items(), key=operator.itemgetter(1), reverse=True):
+                    print(pair)
+            else:
+                print("%s ne parle pas de %s" % (artist, keyword))
+
+            for song in songs:
+                # title = song["title"]
+                if song["lyrics"] is None:
+                    stats["missing"][artist] += 1
+                else:
                     lyrics: str = song["lyrics"]
+
                     if are_invalid(lyrics):
                         stats["missing"][artist] += 1
                     else:
                         lyrics = cleanup(lyrics)
                         stats["avg_lines"][artist] += len(lyrics)
                         stats["avg_newlines"][artist] += lyrics.count("\n")
-                else:
-                    stats["missing"][artist] += 1
+
+                        doc = nlp(lyrics)
+                        keywords.extend([token.text for token in doc if
+                                         token.is_stop != True and
+                                         token.is_punct != True and
+                                         token.pos_ in ["PRON", "NOUN", "ADJ", "VERB", "INTJ", "X"]])
+
+            # five most common noun tokens
+            noun_freq = Counter(keywords)
+            common_nouns = noun_freq.most_common(20)
 
             final_missing = stats["missing"][artist]
             final_songs = len(songs) - final_missing
@@ -66,6 +99,7 @@ def analyse():
                 stats["avg_lines"][artist],
                 stats["avg_newlines"][artist],
                 final_missing))
+            print(common_nouns)
 
 
 if __name__ == '__main__':
diff --git a/CommentTuParles/requirements.txt b/CommentTuParles/requirements.txt
new file mode 100644
index 0000000..e394bb0
--- /dev/null
+++ b/CommentTuParles/requirements.txt
@@ -0,0 +1 @@
+spacy==2.2.3