diff --git a/language_classifier/classifier b/language_classifier/classifier
new file mode 100644
index 0000000..425a653
Binary files /dev/null and b/language_classifier/classifier differ
diff --git a/language_classifier/gather_data.py b/language_classifier/gather_data.py
new file mode 100644
index 0000000..4600492
--- /dev/null
+++ b/language_classifier/gather_data.py
@@ -0,0 +1,46 @@
+import pandas as pd
+import glob
+from bs4 import BeautifulSoup
+import os
+
+extension_dict = {'gcc': 'c', 'perl': 'perl', 'clojure': 'clojure', 'hs': 'haskell', 'java': 'java',
+                 'javascript': 'javascript', 'jruby': 'ruby', 'yarv': 'ruby', 'ocaml': 'ocaml',
+                 'sbcl': 'lisp', 'scala': 'scala', 'csharp': 'csharp', 'hack': 'php', 'php': 'php',
+                 'python3': 'python', 'racket': 'scheme', 'tcl': 'tcl'}
+
+def get_test_data():
+    content = []
+    for file in sorted(os.listdir("../data/test/"), key=int):
+        with open("../data/test/" + file) as fh:
+            content.append([fh.read()])
+    test_data = pd.DataFrame(content)
+    return test_data
+
+def get_code_from_html(lang):
+    htmlfiles = glob.glob("../data/html/*.html")
+    texts = []
+    tags = []
+    for file in htmlfiles:
+        soup = BeautifulSoup(open(file))
+        html_tag = soup.find_all('pre', {'class' : '{} highlighted_source'.format(lang)})
+        html_text = [part.get_text() for part in html_tag]
+        for tag in html_tag:
+            tags.append(lang)
+        texts.extend(html_text)
+    return texts, tags
+
+def get_benchmark_code(directory):
+    files = glob.glob("../data/corpus/{}/*.{}".format(directory, directory))
+    texts = []
+    tags = []
+    for file in files:
+        with open(file) as fh:
+            tags.append(extension_dict[directory])
+            texts.append(fh.read())
+    return texts, tags
+
+def get_snippet(filename):
+    content = []
+    with open(filename) as fh:
+        content.append([fh.read()])
+    return content
\ No newline at end of file
diff --git a/language_classifier/lang_classifier.py b/language_classifier/lang_classifier.py
new file mode 100644
index 0000000..93e35a0
--- /dev/null
+++ b/language_classifier/lang_classifier.py
@@ -0,0 +1,113 @@
+import re
+import itertools
+import random
+import pickle
+import pandas as pd
+
+from sklearn.pipeline import Pipeline
+from sklearn.cross_validation import train_test_split
+from sklearn.naive_bayes import MultinomialNB
+from sklearn.base import TransformerMixin
+
+import gather_data as gd
+
+def count_characters(text):
+    return len(text)
+
+def count_words(text):
+    words = [r'\barray\b', r'\bbegin\b', r'\bend\b', r'\bdo\b', r'\bvar\b', r'\bdefn\b', r'\bfunction\b',
+            r'\bclass\b', r'\brequire\b', r'\bval\b', r'\bpublic\b', r'\blet\b', r'\bwhere\b', r'\busing\b',
+            r'\bextend\b', r'\bfunction\b', r'\bval\b', r'\btry\b']
+    results = []
+    for word in words:
+        results.append(len(re.findall(word, text)))
+    return results
+
+def char_runs(text):
+    chars = [r'[)]+',r'[}]+', r'[\]]+', r'[=]+']
+    results = []
+    for char in chars:
+        found = sorted(re.findall(char, text), key=len)
+        if found:
+            results.append(len(found[-1]))
+        else:
+            results.append(0)
+    return results
+
+def percent_characters(text):
+    chars = ';!=.<>/\[]{}:_#%$&*'
+    results = []
+    for char in chars:
+        total = max(1, len(text))
+        found = text.count(char)
+        if found:
+            results.append(found / total)
+        else:
+            results.append(0)
+    return results
+
+def endings(text):
+    ends = [r'[)]$', r';$', r'}$', r']$', r'\):$']
+    results = []
+    for end in ends:
+        results.append(len(re.findall(end, text, re.MULTILINE)))
+    return results
+
+
+class FunctionFeaturizer(TransformerMixin):
+    def __init__(self, *featurizers):
+        self.featurizers = featurizers
+
+    def fit(self, X, y=None):
+        return self
+
+    def transform(self, X):
+        fvs = []
+        for datum in X:
+            vec = list(itertools.chain.from_iterable([function(datum) for function in self.featurizers]))
+            fvs.append(vec)
+        return fvs
+
+class PipelineDebugger(TransformerMixin):
+    def __init__(self, name):
+        self.name = name
+
+    def fit(self, X, y=None):
+        return self
+
+    def transform(self, X):
+        print(self.name)
+        print("=" * 40)
+        x = X[random.randrange(0, len(X))]
+        print("len:", len(x))
+        print(x)
+        return X
+
+if __name__ == '__main__':
+    texts = []
+    tags = []
+
+    languages = ['c', 'perl', 'clojure', 'haskell', 'java', 'javascript', 'ruby', 'ocaml', 'lisp', 'scala', 'csharp', 'php', 'python', 'scheme', 'tcl']
+    for language in languages:
+        texts.extend(gd.get_code_from_html(language)[0])
+        tags.extend(gd.get_code_from_html(language)[1])
+
+    folders = ['clojure', 'csharp', 'gcc', 'hack', 'hs', 'java', 'javascript', 'jruby', 'ocaml', 'perl', 'php', 'python3', 'racket', 'sbcl', 'scala', 'yarv']
+    for folder in folders:
+        tags.extend(gd.get_benchmark_code(folder)[1])
+        texts.extend(gd.get_benchmark_code(folder)[0])
+
+    df_texts = pd.DataFrame(texts)
+    print(df_texts.head())
+    df_tags = pd.DataFrame(tags)
+    merged = pd.merge(df_texts, df_tags, left_index=True, right_index=True)
+    merged.columns = ['Snippet', 'Language']
+
+    train_X, test_X, train_y, test_y = train_test_split(merged['Snippet'], merged['Language'], test_size=0.33)
+
+    classifier = Pipeline([('features', FunctionFeaturizer(count_words, percent_characters, char_runs, endings)),
+                            ('bayes', MultinomialNB())])
+    classifier.fit(train_X, train_y)
+
+    with open("./classifier", "wb") as file:
+        pickle.dump(classifier, file)
\ No newline at end of file
diff --git a/language_classifier/predict.py b/language_classifier/predict.py
new file mode 100644
index 0000000..682dd5f
--- /dev/null
+++ b/language_classifier/predict.py
@@ -0,0 +1,17 @@
+import gather_data as gd
+import pandas as pd
+import sys
+import pickle
+
+def predict(classifier, data):
+    prediction = classifier.predict(data)
+    print(prediction)
+
+
+if __name__ == '__main__':
+    content = gd.get_snippet(sys.argv[1])
+    df = pd.DataFrame(content)
+    with open("./classifier", "rb") as file:
+        predictor = pickle.load(file)
+        predict(predictor, df)
+
diff --git a/language_classifier/tests/test_lang_classifier.py b/language_classifier/tests/test_lang_classifier.py
new file mode 100644
index 0000000..4306da8
--- /dev/null
+++ b/language_classifier/tests/test_lang_classifier.py
@@ -0,0 +1,21 @@
+from language_classifier.lang_classifier import *
+
+test_data = []
+data_lang = 'python'
+
+with open("language_classifier/tests/feature_test.txt") as file:
+    test_file = file.read()
+
+def test_total_characters():
+    assert count_characters(test_file) == 32
+
+def test_percent_char():
+    assert percent_character(test_file, '.') == 6/32
+    assert percent_character(test_file, ';') == 7/32
+    assert percent_character(test_file, '\t') == 4/32
+
+def test_count_vars():
+    assert count_vars(test_file) == 2
+
+def test_percent_word_chars():
+    assert count_word_chars(test_file) == 6/32
\ No newline at end of file