diff --git a/odoo/addons/base/tests/test_translate.py b/odoo/addons/base/tests/test_translate.py
index 1cd833c9df6..de607798a82 100644
--- a/odoo/addons/base/tests/test_translate.py
+++ b/odoo/addons/base/tests/test_translate.py
@@ -205,10 +205,16 @@ class TranslationToolsTestCase(unittest.TestCase):
         self.assertEquals(result, expect)
 
     def test_translate_html(self):
+        """ Test html_translate(). """
+        source = """<blockquote>A <h2>B</h2> C</blockquote>"""
+        result = html_translate(lambda term: term, source)
+        self.assertEquals(result, source)
+
+    def test_translate_html_i(self):
         """ Test xml_translate() and html_translate() with <i> elements. """
-        source = """<i class="fa-check"></i>"""
+        source = """<p>A <i class="fa-check"></i> B</p>"""
         result = xml_translate(lambda term: term, source)
-        self.assertEquals(result, """<i class="fa-check"/>""")
+        self.assertEquals(result, """<p>A <i class="fa-check"/> B</p>""")
         result = html_translate(lambda term: term, source)
         self.assertEquals(result, source)
 
diff --git a/odoo/tools/translate.py b/odoo/tools/translate.py
index c947ae5e75d..fbef3443929 100644
--- a/odoo/tools/translate.py
+++ b/odoo/tools/translate.py
@@ -17,7 +17,7 @@ from datetime import datetime
 from os.path import join
 
 from babel.messages import extract
-from lxml import etree
+from lxml import etree, html
 
 import odoo
 from odoo.tools import config
@@ -153,8 +153,13 @@ avoid_pattern = re.compile(r"\s*<!DOCTYPE", re.IGNORECASE | re.MULTILINE | re.UN
 node_pattern = re.compile(r"<[^>]*>(.*)</[^<]*>", re.DOTALL | re.MULTILINE | re.UNICODE)
 
 
-def translate_xml_node(node, callback, method, parser=None):
-    """ Return the translation of the given XML/HTML node. """
+def translate_xml_node(node, callback, parse, serialize):
+    """ Return the translation of the given XML/HTML node.
+
+        :param callback: callback(text) returns translated text or None
+        :param parse: parse(text) returns a node (text is unicode)
+        :param serialize: serialize(node) returns unicode text
+    """
 
     def nonspace(text):
         return bool(text) and not text.isspace()
@@ -182,7 +187,7 @@ def translate_xml_node(node, callback, method, parser=None):
     def translate_content(node):
         """ Return ``node`` with its content translated inline. """
         # serialize the node that contains the stuff to translate
-        text = etree.tostring(node, method=method, encoding='utf8').decode('utf8')
+        text = serialize(node)
         # retrieve the node's content and translate it
         match = node_pattern.match(text)
         trans = translate_text(match.group(1))
@@ -190,7 +195,7 @@ def translate_xml_node(node, callback, method, parser=None):
             # replace the content, and convert it back to an XML node
             text = text[:match.start(1)] + trans + text[match.end(1):]
             try:
-                node = etree.fromstring(encode(text), parser=parser)
+                node = parse(text)
             except etree.ParseError:
                 # fallback: escape the translation as text
                 node = etree.Element(node.tag, node.attrib, node.nsmap)
@@ -275,6 +280,21 @@ def translate_xml_node(node, callback, method, parser=None):
     return node
 
 
+def parse_xml(text):
+    return etree.fromstring(encode(text))
+
+def serialize_xml(node):
+    return etree.tostring(node, method='xml', encoding='utf8').decode('utf8')
+
+_HTML_PARSER = etree.HTMLParser(encoding='utf8')
+
+def parse_html(text):
+    return html.fragment_fromstring(encode(text), parser=_HTML_PARSER)
+
+def serialize_html(node):
+    return etree.tostring(node, method='html', encoding='utf8').decode('utf8')
+
+
 def xml_translate(callback, value):
     """ Translate an XML value (string), using `callback` for translating text
         appearing in `value`.
@@ -283,17 +303,15 @@ def xml_translate(callback, value):
         return value
 
     try:
-        root = etree.fromstring(encode(value))
-        result = translate_xml_node(root, callback, 'xml')
-        return etree.tostring(result, method='xml', encoding='utf8').decode('utf8')
+        root = parse_xml(value)
+        result = translate_xml_node(root, callback, parse_xml, serialize_xml)
+        return serialize_xml(result)
     except etree.ParseError:
         # fallback for translated terms: use an HTML parser and wrap the term
-        wrapped = "<div>%s</div>" % encode(value)
-        root = etree.fromstring(wrapped, etree.HTMLParser(encoding='utf-8'))
-        # root is html > body > div; translate the div only
-        result = translate_xml_node(root[0][0], callback, 'xml')
+        root = parse_html("<div>%s</div>" % value)
+        result = translate_xml_node(root, callback, parse_xml, serialize_xml)
         # remove tags <div> and </div> from result
-        return etree.tostring(result, method='xml', encoding='utf8').decode('utf8')[5:-6]
+        return serialize_xml(result)[5:-6]
 
 def html_translate(callback, value):
     """ Translate an HTML value (string), using `callback` for translating text
@@ -303,14 +321,11 @@ def html_translate(callback, value):
         return value
 
     try:
-        parser = etree.HTMLParser(encoding='utf-8')
         # value may be some HTML fragment, wrap it into a div
-        wrapped = "<div>%s</div>" % encode(value)
-        root = etree.fromstring(wrapped, parser)
-        # root is html > body > div; translate the div only
-        result = translate_xml_node(root[0][0], callback, 'html', parser)
+        root = parse_html("<div>%s</div>" % value)
+        result = translate_xml_node(root, callback, parse_html, serialize_html)
         # remove tags <div> and </div> from result
-        value = etree.tostring(result, method='html', encoding='utf8').decode('utf8')[5:-6]
+        value = serialize_html(result)[5:-6]
     except ValueError:
         _logger.exception("Cannot translate malformed HTML, using source value instead")