diff --git a/requirements.colab.txt b/requirements.colab.txt
new file mode 100644
index 0000000..a876f56
--- /dev/null
+++ b/requirements.colab.txt
@@ -0,0 +1,12 @@
+transformers==2.1.1
+
+gensim
+requests
+torch
+tqdm
+pandas
+sklearn
+numpy
+matplotlib
+pyyaml
+nltk
diff --git a/wiki/utils.py b/wiki/utils.py
new file mode 100644
index 0000000..2e92e63
--- /dev/null
+++ b/wiki/utils.py
@@ -0,0 +1,35 @@
+import requests
+
+
+def get_text_from_wikipedia(title, wiki_domain='en.wikipedia.org'):
+    """
+    Retrieve Wikipedia article content as plain-text
+
+    :param title: Title of Wikipedia article
+    :param wiki_domain: API domain (e.g., en.wikipedia.org)
+    :return: Article content as plain-text
+    """
+
+    res = requests.get(f'https://{wiki_domain}/w/api.php', params={
+        "action":"query",
+        "prop":"revisions",
+        "rvprop":"content",
+        "format":"json",
+        "titles":title,
+        "rvslots":"main"
+    })
+
+    from gensim.corpora.wikicorpus import filter_wiki
+    pages = [p for p_id, p in res.json()['query']['pages'].items()]
+
+    if len(pages) == 0:
+        raise ValueError(f'Cannot find Wikipedia article: {title}')
+    elif len(pages) > 1:
+        raise ValueError(f'Wikipedia article title is unambigious. Multiple articles returned from API: {title}')
+    else:
+        p = pages[0]
+
+    wikitext = p['revisions'][0]['slots']['main']['*']  #
+    text = filter_wiki(wikitext).strip()
+
+    return text