diff --git a/CHANGES.md b/CHANGES.md
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -1,14 +1,104 @@
 Seonbi changelog
 ================
 
+Version 0.3.0
+-------------
+
+Released November 18, 2021.
+
+ -  Since this version, it requires GHC 8.8.* at least, and supports GHC 9.0.*
+    at most.
+
+ -  Now supports several content types besides HTML/XHTML.  The below list
+    shows all supported content types:  [[#18]]
+
+     -  `text/html` (previously non-XHTML mode)
+     -  `application/xhtml+xml` (previously XHTML mode)
+     -  `text/plain` (added)
+     -  `text/markdown` (added)
+
+    The below Haskell APIs changed:
+
+     -  Added `Text.Seonbi.ContentTypes` module.
+     -  Added `contentType` field for `Configuration m a`.
+     -  Removed `xhtml` field for `Configuration m a` in favor of
+        new `contentType` field for the same type.
+
+    The below CLI options changed:
+
+     -  Added `-t`/`--content-type` option with the default value `text/html`.
+     -  Removed Removed `-x`/`--xhtml` option in favor of new
+        `-t`/`--content-type` option.  In order to use XHTML mode, give it
+        `-t application/xhtml+xml` option.
+
+    The below HTTP APIs changed:
+
+     -  Added a mandatory field `"content"` to requests.
+     -  Deprecated the `"sourceHtml"` field of requests in favor of the new
+        `"content"` field.
+     -  Added an optional field `"contentType"` with the default value
+        `"text/html"` to requests.
+     -  Deprecated the `"xhtml"` field of requests in favor of the new
+        `"contentType"` field.  The legacy field will be gone in the next
+        minor release.
+        In order to use XHTML mode, configure `"contentType"` field with
+        `"application/xhtml+xml"`.
+     -  Added `"content"` field to responses.
+     -  Deprecated the `"resultHtml"` field of responses in favor of the new
+        `"content"` field.  The legacy field is not provided for non-HTML
+        types, and will be gone in the next minor release.
+     -  Added `"contentType"` field to responses.
+     -  Added `"warnings"` field to responses.
+
+ -  Added `Text.Seonbi.Html.Lang` module.
+
+ -  Some transformations inappropriate for non-Korean contents are no more
+    applied to elements written in other languages than Korean.  The below
+    functions respect elements `lang` attributes:  [[#10]]
+
+     -  `Text.Seonbi.Hanja.phoneticizeHanja`
+     -  `Text.Seonbi.Punctuation.normalizeStops`
+
+ -  Removed several functions from `Text.Seonbi.Trie` module:
+
+     -  `toListBy`
+     -  `lookupBy`
+     -  `submap`
+     -  `match`
+     -  `matches`
+     -  `alterBy`
+     -  `adjust`
+     -  `delete`
+     -  `mapBy`
+     -  `filterMap`
+
+ -  `Text.Seonbi.Trie.Trie` type is not an instance of the following typeclasses
+    anymore:
+
+     -  `Generic a => Generic (Trie a)`
+     -  `Binary a => Binary (Trie a)`
+     -  `Generic1 Trie`
+     -  `type Rep (Trie a)`
+     -  `type Rep1 Trie`
+
+ -  Added `Text.Seonbi.Html.Printer.printText` function.
+ -  Added `Text.Seonbi.Html.Tag.headingLevel` function.
+ -  Added `Text.Seonbi.Html.Tag.headingTag` function.
+ -  Added `Text.Seonbi.Html.Tag.headingTag'` function.
+ -  Added `Text.Seonbi.Html.TagStack.last` function.
+
+[#10]: https://github.com/dahlia/seonbi/issues/10
+[#18]: https://github.com/dahlia/seonbi/issues/18
+
+
 Version 0.2.3
 -------------
 
 Released on September 26, 2021.
 
-  - Fixed stops normalizer's bug where trailing spaces following stops had been
+ -  Fixed stops normalizer's bug where trailing spaces following stops had been
     trimmed after normalized.
-  - Fixed stops normalizer's buf where unnecessary trailing spaces following
+ -  Fixed stops normalizer's buf where unnecessary trailing spaces following
     stops had been inserted after normalized.  In particular, unnecessary
     spaces between stops and closing parentheses/brackets are no more inserted.
 
diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -38,11 +38,20 @@
  -  A hyphen or an equality symbol wrapped by inequality symbols (`<->`, `<=>`)
     into bi-directional arrows (`↔`, `⇔`)
 
-Since its transformations work in HTML-level, it also plays well with web
-markup languages like CommonMark, Markdown, and Textile.  In a similar way to
-SmartyPants, it does not modify characters within several sensitive
-HTML elements like `<pre>`/`<code>`/`<script>`/`<kbd>`.
+Each transformations can be partially turned on and off, and some
+transformations have many options.
 
+All transformations work with both plain texts and rich text tree.
+In a similar way to SmartyPants, it does not modify characters within
+several sensitive HTML elements like `<pre>`/`<code>`/`<script>`/`<kbd>`.
+Chinese/Japanese stops or hanzi/kanji characters inside elements with
+`lang="zh"`/`lang="ja"`[^1] are never transformed.
+
+[^1]: Technically, only Korean contents and language-unspecified elements
+      are transformed.  Elements having `lang` attribute with language tags
+      referring to any Korean language are treated as Korean contents,
+      e.g., `ko`, `ko-Hang`, `kor-KP`, `kor-Kore`.
+
 [releases]: https://github.com/dahlia/seonbi/releases
 [releases-badge]: https://img.shields.io/github/v/release/dahlia/seonbi
 [hackage]: https://hackage.haskell.org/package/seonbi
@@ -56,6 +65,33 @@
 [Korean mixed script]: https://en.wikipedia.org/wiki/Korean_mixed_script
 
 
+End-user apps
+-------------
+
+Technically, Seonbi is basically exposed as a software component,
+which is also known as API (application programming interface), to be used
+as a module of other softwares.
+
+However, as these official interfaces are not for humans but machines, it's
+not easy to use for end-users whom haven't experienced software
+programming.  For such end-users, here's the list of end-user apps:
+
+ -  [Demo web app]: The official web app.  Not that mobile-friendly
+    (usable though).
+ -  [Seonbi Translator] (선비 번역기): An iOS shortcut made by
+    [Heechan Bak] ([@channprj]).
+ -  [seonbi-webapp]: Probably aims to replace the official web app.
+    Made by Lee Dogeon ([@moreal]).
+ -  [seonbi-ext]: A browser extension made by Lee Dogeon ([@moreal]).
+
+[Seonbi Translator]: https://www.icloud.com/shortcuts/1e468a5dd0104fc5b69fe803952fc154
+[Heechan Bak]: https://chann.dev/
+[@channprj]: https://github.com/channprj
+[seonbi-webapp]: https://github.com/moreal/seonbi-webapp
+[@moreal]: https://github.com/moreal
+[seonbi-ext]: https://github.com/moreal/seonbi-ext
+
+
 Installation
 ------------
 
@@ -121,6 +157,7 @@
 
     {
       "preset": "ko-kr",
+      "contentType": "text/html",
       "sourceHtml": "<p>하늘과 바람과 별과 詩</p>"
     }
 
@@ -128,10 +165,11 @@
 
     HTTP/1.1 200 OK
     Content-Type: application/json
-    Server: Seonbi/0.2.3
+    Server: Seonbi/0.3.0
 
     {
       "success": true,
+      "contentType": "text/html",
       "resultHtml": "<p>하늘과 바람과 별과 시</p>"
     }
 
diff --git a/app/seonbi-api.hs b/app/seonbi-api.hs
--- a/app/seonbi-api.hs
+++ b/app/seonbi-api.hs
@@ -7,6 +7,7 @@
 
 import Control.Concurrent (threadDelay)
 import Control.Monad
+import Data.Maybe (catMaybes)
 import Data.String
 import Data.Version
 import GHC.Exts (IsList (..))
@@ -16,6 +17,7 @@
 import qualified Data.Aeson.Types
 import qualified Data.ByteString as B
 import qualified Data.Map.Strict as M
+import qualified Data.Set as S
 import Data.Text
 import Data.Text.Encoding
 import Network.Wai
@@ -28,14 +30,38 @@
 import Text.Seonbi.Trie as Trie
 
 data Input = Input
-    { sourceHtml :: Text
+    { source :: Text
     , configuration :: Configuration IO ()
+    , warnings :: [Text]
     } deriving (Show)
 
 instance FromJSON Input where
     parseJSON = withObject "Input" $ \ v -> do
-        sourceHtml' <- v .: "sourceHtml"
+        sourceMaybe <- v .:? "content"
+        (source', w1) <- case sourceMaybe of
+            Just s -> return (s, Nothing)
+            Nothing -> do
+                sourceHtml' <- v .:? "sourceHtml"
+                case sourceHtml' of
+                    Just h -> return
+                        ( h
+                        , Just $ "key \"sourceHtml\" is deprecated in " <>
+                                 "favour of \"content\""
+                        )
+                    Nothing -> fail "key \"content\" not present"
         preset <- v .:? "preset"
+        contentTypeMaybe <- v .:? "contentType"
+        (contentType', w2) <- case contentTypeMaybe of
+            Just t -> return (t, Nothing)
+            Nothing -> do
+                xhtml <- v .:? "xhtml"
+                case xhtml of
+                    Just x -> return
+                        ( if x then "application/xhtml+xml" else "text/html"
+                        , Just $ "key \"xhtml\" is deprecated in favour of " <>
+                                 "\"contentType\""
+                        )
+                    Nothing -> return ("text/html", Nothing)
         config <- case preset of
             Just locale ->
                 let presets' = presets :: M.Map Text (Configuration IO ())
@@ -50,7 +76,6 @@
                             , intercalate ", " (M.keys presets')
                             ]
             Nothing -> do
-                xhtml' <- v .:? "xhtml" .!= False
                 quote' <- v .:? "quote"
                 cite' <- v .:? "cite"
                 arrow' <- v .:? "arrow"
@@ -60,7 +85,7 @@
                 hanja' <- v .:? "hanja" .!= Nothing
                 return Configuration
                     { debugLogger = Nothing
-                    , xhtml = xhtml'
+                    , contentType = contentType'
                     , quote = quote'
                     , cite = cite'
                     , arrow = arrow'
@@ -69,8 +94,23 @@
                     , stop = stop'
                     , hanja = hanja'
                     }
-        return $ Input sourceHtml' config
+        return $ Input
+            { source = source'
+            , configuration = config { contentType = contentType' }
+            , warnings = catMaybes [w1, w2]
+            }
 
+instance FromJSON ContentType where
+    parseJSON = withText "ContentType" $ \ t ->
+        if contentTypeFromText t `S.member` contentTypes
+        then return (contentTypeFromText t)
+        else fail $ unpack $ Data.Text.concat
+            [ "Unknown content type: "
+            , t
+            , "; available content types: "
+            , intercalate ", " $ contentTypeText <$> S.elems contentTypes
+            ]
+
 instance FromJSON QuoteOption
 instance FromJSON CiteOption
 instance FromJSON ArrowOption
@@ -117,12 +157,29 @@
             inputJson <- lazyRequestBody request
             threadDelay (debugDelayMs * 1000)
             case eitherDecode' inputJson of
-                Right (Input source config) -> do
+                Right (Input source config warnings) -> do
                     result <- transformHtmlText config source
-                    respond' status200 $ object
+                    let type' = contentType config
+                    let warningComments =
+                            if Prelude.null warnings
+                            then Data.Text.empty
+                            else Data.Text.concat
+                                [ "<!--\n"
+                                , Data.Text.intercalate "\n" warnings
+                                , "\n-->"
+                                ]
+                    respond' status200 $ object $
                         [ "success" .= Bool True
-                        , "resultHtml" .= String result
+                        , "content" .= String result
+                        , "warnings" .= Array
+                            (GHC.Exts.fromList $ String <$> warnings)
+                        , "contentType" .= String (contentTypeText type')
                         ]
+                        ++
+                        [ "resultHtml" .= String (warningComments <> result)
+                        | type' == "text/html" ||
+                          type' == "application/xhtml+xml"
+                        ]
                 Left msg -> respond' status400 $ object
                     [ "success" .= Bool False
                     , "message" .= String (pack msg)
@@ -194,13 +251,14 @@
         )
     <*> ( AppOptions
         <$> (
-                ( Just <$> strOption
-                    ( long "allow-origin"
-                    <> short 'o'
-                    <> metavar "ORIGIN"
-                    <> help "Allow cross-origin (i.e., CORS)"
+                optional
+                    ( strOption
+                        ( long "allow-origin"
+                        <> short 'o'
+                        <> metavar "ORIGIN"
+                        <> help "Allow cross-origin (i.e., CORS)"
+                        )
                     )
-                )
                 <|> pure Nothing
             )
         <*> option auto
@@ -233,5 +291,6 @@
     let netloc = showHostPreference (getHost serverSettings') ++ ":" ++
             show (getPort serverSettings')
     let url = "http://" ++ netloc ++ "/"
+    hPutStrLn stderr $ "seonbi-api v" ++ showVersion Meta.version
     hPutStrLn stderr url
     runSettings serverSettings' $ app appOptions
diff --git a/app/seonbi.hs b/app/seonbi.hs
--- a/app/seonbi.hs
+++ b/app/seonbi.hs
@@ -25,6 +25,7 @@
 #endif
 import Data.ByteString.Lazy
 import Data.Map.Strict
+import qualified Data.Set as S
 import qualified Data.Text as T
 import Data.Text.Lazy
 import Data.Text.Lazy.Encoding
@@ -92,7 +93,7 @@
     , config :: Configuration IO ()
     , dictionaries :: [FilePath]
     , noKrStdict :: Bool
-    , xhtml :: Bool
+    , contentType' :: ContentType
     , debug :: Bool
     , version :: Bool
     , input :: FilePath
@@ -137,9 +138,11 @@
 enumKeyword = T.unpack . enumKeyword'
 
 enumKeywords :: forall a . (Enum a, Show a) => Proxy a -> String
-enumKeywords _ = T.unpack $ T.intercalate ", " $
-    fmap enumKeyword' [(toEnum 0 :: a) ..]
+enumKeywords _ = commas $ enumKeyword' <$> [(toEnum 0 :: a) ..]
 
+commas :: [T.Text] -> String
+commas = T.unpack . T.intercalate ", "
+
 parser :: Parser Seonbi
 parser = Seonbi
     <$> strOption
@@ -166,7 +169,7 @@
                      "Available presets: " ++
                      Data.List.intercalate ", " (Data.Map.Strict.keys presets'))
             )
-        <|> ( Configuration Nothing False
+        <|> ( Configuration Nothing "text/html"
             <$> ( flag' Nothing
                     ( long "no-quote"
                     <> short 'Q'
@@ -297,10 +300,14 @@
         <> help ("Do not use Standard Korean Language Dictionary " ++
                  "(標準國語大辭典) by South Korean NIKL (國立國語院)")
         )
-    <*> switch
-        ( long "xhtml"
-        <> short 'x'
-        <> help "XHTML mode"
+    <*> strOption
+        ( long "content-type"
+        <> short 't'
+        <> metavar "TYPE"
+        <> value "text/html"
+        <> help ("Content type.  Available types: " ++ commas
+                (contentTypeText <$> S.elems contentTypes) ++
+                "  [default: text/html]")
         )
     <*> switch
         ( long "debug"
@@ -347,6 +354,7 @@
         , config
         , dictionaries
         , noKrStdict
+        , contentType'
         , debug
         , version
         , input
@@ -371,6 +379,7 @@
                 { debugLogger = debugLogger'
                 , hanja = Just hanja' { reading = reading' }
                 }
+    let configWithContentType = config' { contentType = contentType' }
     when version $ do
         Prelude.putStrLn $ showVersion Meta.version
         exitSuccess
@@ -390,7 +399,8 @@
             enc -> enc
     debugPrint ("encoding: " ++ encodingName)
     result <- catchIOError
-        (transformHtmlLazyText config' $ toUnicode encodingName contents)
+        (transformHtmlLazyText configWithContentType $
+            toUnicode encodingName contents)
         (\ e -> hPutStrLn stderr (ioeGetErrorString e) >> exitFailure)
     let resultBytes = fromUnicode encodingName result
     if output == "-"
diff --git a/seonbi.cabal b/seonbi.cabal
--- a/seonbi.cabal
+++ b/seonbi.cabal
@@ -5,7 +5,7 @@
 -- see: https://github.com/sol/hpack
 
 name:           seonbi
-version:        0.2.3
+version:        0.3.0
 synopsis:       SmartyPants for Korean language
 description:    Please see the README.md on GitHub at <https://github.com/dahlia/seonbi>.
 category:       Text
@@ -36,10 +36,10 @@
     , bytestring
     , directory ==1.*
     , filepath
-    , http-client >=0.5 && <0.7
+    , http-client >=0.5 && <0.8
     , temporary >=1.2 && <1.4
     , text
-    , zip >=1.1 && <1.3
+    , zip >=1.1 && <1.8
 
 flag embed-dictionary
   description: Embed dictionary rather than load from file
@@ -58,12 +58,14 @@
 
 library
   exposed-modules:
+      Text.Seonbi.ContentTypes
       Text.Seonbi.Facade
       Text.Seonbi.Hangul
       Text.Seonbi.Hanja
       Text.Seonbi.Html
       Text.Seonbi.Html.Clipper
       Text.Seonbi.Html.Entity
+      Text.Seonbi.Html.Lang
       Text.Seonbi.Html.Preservation
       Text.Seonbi.Html.Printer
       Text.Seonbi.Html.Scanner
@@ -80,17 +82,20 @@
   hs-source-dirs:
       src
   build-depends:
-      aeson >=1.3.1 && <1.5
+      aeson >=1.3.1 && <1.6
     , attoparsec >=0.12 && <1
-    , base >=4.7 && <5
+    , base >=4.12 && <5
     , bytestring
+    , bytestring-trie >=0.2.5 && <0.3
+    , case-insensitive ==1.*
     , cassava ==0.5.*
+    , cmark >=0.6 && <1
     , containers
     , data-default >=0.2 && <1
-    , file-embed >=0.0.10 && <0.0.12
+    , file-embed >=0.0.10 && <0.0.16
     , filepath ==1.*
+    , html-entities ==1.*
     , text
-    , text-trie >=0.2.5 && <0.3
   if os(darwin)
     ghc-options: -Wall -fprint-explicit-kinds -optP-Wno-nonportable-include-path
   else
@@ -108,10 +113,11 @@
   hs-source-dirs:
       app
   build-depends:
-      aeson >=1.3.1 && <1.5
-    , base >=4.7 && <5
+      aeson >=1.3.1 && <1.6
+    , base >=4.12 && <5
     , bytestring
     , containers
+    , html-entities ==1.*
     , text
   if os(darwin)
     ghc-options: -Wall -fprint-explicit-kinds -optP-Wno-nonportable-include-path
@@ -120,16 +126,16 @@
   if flag(iconv)
     cpp-options: -DICONV
     build-depends:
-        cases ==0.1.3.2
+        cases >=0.1.3.2 && <0.1.5
       , html-charset ==0.1.*
       , iconv ==0.4.*
-      , optparse-applicative ==0.14.*
+      , optparse-applicative >=0.14 && <0.17
       , seonbi
   else
     build-depends:
-        cases ==0.1.3.2
+        cases >=0.1.3.2 && <0.1.5
       , html-charset ==0.1.*
-      , optparse-applicative ==0.14.*
+      , optparse-applicative >=0.14 && <0.17
       , seonbi
   if flag(static)
     ld-options: -static
@@ -148,12 +154,13 @@
   hs-source-dirs:
       app
   build-depends:
-      aeson >=1.3.1 && <1.5
-    , base >=4.7 && <5
+      aeson >=1.3.1 && <1.6
+    , base >=4.12 && <5
     , bytestring
     , containers
+    , html-entities ==1.*
     , http-types ==0.12.*
-    , optparse-applicative ==0.14.*
+    , optparse-applicative >=0.14 && <0.17
     , seonbi
     , text
     , wai >=3.2 && <3.4
@@ -180,12 +187,13 @@
   ghc-options: -threaded
   build-depends:
       QuickCheck
-    , aeson >=1.3.1 && <1.5
-    , base >=4.7 && <5
+    , aeson >=1.3.1 && <1.6
+    , base >=4.12 && <5
     , bytestring
     , containers
     , doctest
     , doctest-discover
+    , html-entities ==1.*
     , seonbi
     , text
     , unicode-show
@@ -202,11 +210,12 @@
       test
   ghc-options: -threaded
   build-depends:
-      aeson >=1.3.1 && <1.5
-    , base >=4.7 && <5
+      aeson >=1.3.1 && <1.6
+    , base >=4.12 && <5
     , bytestring
     , containers
-    , hlint >=2.1.7 && <2.2
+    , hlint >=2.1.7 && <3.3
+    , html-entities ==1.*
     , text
   if os(darwin)
     ghc-options: -Wall -fprint-explicit-kinds -optP-Wno-nonportable-include-path
@@ -218,10 +227,12 @@
   type: exitcode-stdio-1.0
   main-is: Spec.hs
   other-modules:
+      Text.Seonbi.ContentTypesSpec
       Text.Seonbi.FacadeSpec
       Text.Seonbi.HangulSpec
       Text.Seonbi.HanjaSpec
       Text.Seonbi.Html.ClipperSpec
+      Text.Seonbi.Html.LangSpec
       Text.Seonbi.Html.PreservationSpec
       Text.Seonbi.Html.PrinterSpec
       Text.Seonbi.Html.ScannerSpec
@@ -234,17 +245,18 @@
       test
   ghc-options: -Wno-incomplete-uni-patterns -threaded -rtsopts -with-rtsopts=-N
   build-depends:
-      Diff >=0.3.4 && <0.4
-    , aeson >=1.3.1 && <1.5
-    , base >=4.7 && <5
+      Diff >=0.3.4 && <0.5
+    , aeson >=1.3.1 && <1.6
+    , base >=4.12 && <5
     , bytestring
     , containers
     , directory ==1.*
     , filepath ==1.*
     , hspec >=2.4.8 && <3
     , hspec-discover >=2.4.8 && <3
+    , html-entities ==1.*
     , interpolatedstring-perl6 >=1.0.1 && <2
-    , random ==1.1.*
+    , random >=1.1 && <1.3
     , seonbi
     , text
   if os(darwin)
diff --git a/src/Text/Seonbi/ContentTypes.hs b/src/Text/Seonbi/ContentTypes.hs
new file mode 100644
--- /dev/null
+++ b/src/Text/Seonbi/ContentTypes.hs
@@ -0,0 +1,348 @@
+{-# LANGUAGE CPP #-}
+{-# LANGUAGE OverloadedLists #-}
+{-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE RankNTypes #-}
+module Text.Seonbi.ContentTypes
+    ( ContentType
+    , HtmlTransformer
+    , TextTransformer
+    , asCommonMarkTransformer
+    , asHtmlTransformer
+    , asHtmlTransformer'
+    , asPlainTextTransformer
+    , asXhtmlTransformer
+    , contentTypeFromText
+    , contentTypes
+    , contentTypeText
+    , transformWithContentType
+    ) where
+
+#if MIN_VERSION_base(4,13,0)
+import Prelude hiding (MonadFail)
+#endif
+
+import Control.Monad (forM)
+import Control.Monad.Fail (MonadFail)
+import Data.Maybe (fromMaybe, isNothing)
+import Data.List
+import Text.Read (readMaybe)
+
+import CMark
+import Data.CaseInsensitive
+import Data.Set
+import Data.Text as ST
+import Data.Text.Lazy as LT
+import Data.Text.Lazy.Builder
+import HTMLEntities.Builder
+import HTMLEntities.Decoder
+
+import Text.Seonbi.Html
+import Text.Seonbi.Html.Tag (headingLevel, headingTag')
+import qualified Text.Seonbi.Html.TagStack as TagStack
+
+-- | Represents a function that transforms an 'HtmlEntity' list.
+type HtmlTransformer m
+    = (Monad m, MonadFail m) => [HtmlEntity] -> m [HtmlEntity]
+
+-- | Represents a function that transforms a text.
+type TextTransformer m
+    = (Monad m, MonadFail m) => LT.Text -> m LT.Text
+
+-- | Represents a function that transforms an 'HtmlTransformer' into
+-- a 'TextTransformer'.
+type TransformerTransformer m
+    = (Monad m, MonadFail m) => HtmlTransformer m -> TextTransformer m
+
+-- | Gets a 'TransformerTransformer' that transforms 'HtmlTransformer' into
+-- a 'TextTransformer' which transforms an HTML/XHTML text.
+asHtmlTransformer'
+    :: (Monad m, MonadFail m)
+    => Bool
+    -- ^ 'True' for XHTML, and 'False' for HTML.
+    -> TransformerTransformer m
+    -- ^ A 'TransformerTransformer' that transforms an 'HtmlTransformer' into
+    -- a 'TextTransformer' which transforms an HTML/XHTML text.
+asHtmlTransformer' xhtml transformer htmlText = do
+    case scanHtml htmlText of
+        Done "" input -> do
+            output <- transformer input
+            return $ printHtml' output
+        _ ->
+            fail "failed to parse input"
+  where
+    printHtml' :: [HtmlEntity] -> LT.Text
+    printHtml'
+      | xhtml = printXhtml
+      | otherwise = printHtml
+
+-- | Transforms an 'HtmlTransformer' into a 'TextTransformer' which transforms
+-- an HTML text.
+asHtmlTransformer :: (Monad m, MonadFail m) => TransformerTransformer m
+asHtmlTransformer = asHtmlTransformer' False
+
+-- | Transforms an 'HtmlTransformer' into a 'TextTransformer' which transforms
+-- an XHTML text.
+asXhtmlTransformer :: (Monad m, MonadFail m) => TransformerTransformer m
+asXhtmlTransformer = asHtmlTransformer' True
+
+-- | Transforms an 'HtmlTransformer' into a 'TextTransformer' which transforms
+-- a plain text.
+asPlainTextTransformer :: (Monad m, MonadFail m) => TransformerTransformer m
+asPlainTextTransformer transformer text' = do
+    let escaped = toLazyText $ HTMLEntities.Builder.text $ LT.toStrict text'
+    let entities = [HtmlText TagStack.empty $ LT.toStrict escaped]
+    output <- transformer entities
+    return $ printText output
+
+-- | Transforms an 'HtmlTransformer' into a 'TextTransformer' which transforms
+-- a CommonMark (Markdown) text.
+asCommonMarkTransformer :: (Monad m, MonadFail m) => TransformerTransformer m
+asCommonMarkTransformer transformer input = do
+    let inputNode = commonmarkToNode [optSourcePos, optUnsafe] $
+            LT.toStrict input
+    inputEntities <- fromNode [] inputNode
+    outputEntities <- transformer $ normalizeText inputEntities
+    let outputNodes = toNode outputEntities
+    let outputNode = case outputNodes of
+            [node@(Node _ DOCUMENT _)] -> node
+            nodes -> Node Nothing DOCUMENT nodes
+    return $ LT.fromStrict $
+        nodeToCommonmark [optSourcePos , optUnsafe] Nothing outputNode
+  where
+    fromNode :: (Monad m, MonadFail m) => HtmlTagStack -> Node -> m [HtmlEntity]
+    fromNode stack (Node posInfo nodeType children) = case nodeType of
+        DOCUMENT ->
+            nodeWithChildren Article
+        THEMATIC_BREAK -> return
+            [ HtmlStartTag stack HR (posAttr posInfo)
+            , HtmlEndTag stack HR
+            ]
+        PARAGRAPH ->
+            nodeWithChildren P
+        BLOCK_QUOTE ->
+            nodeWithChildren BlockQuote
+        HTML_BLOCK html ->
+            case scanHtml $ LT.fromStrict html of
+                Done "" input' -> return $ rebaseStack input'
+                _ -> return [HtmlCdata stack html]
+        CUSTOM_BLOCK _ _ ->
+            return []
+        CODE_BLOCK info text' -> return
+            [ HtmlStartTag stack Pre (posAttr posInfo <> attr' "info" info)
+            , HtmlCdata (nextStack Pre) text'
+            , HtmlEndTag stack Pre
+            ]
+        HEADING level ->
+            nodeWithChildren $ headingTag' level
+        LIST listAttrs ->
+            nodeWithChildren' UL $ attr' "list-attrs" listAttrs
+        ITEM ->
+            nodeWithChildren LI
+        TEXT text' ->
+            return [HtmlCdata stack text']
+        SOFTBREAK -> return
+            [ HtmlStartTag stack BR (posAttr posInfo <> attr' "softbreak" True)
+            , HtmlEndTag stack BR
+            ]
+        LINEBREAK -> return
+            [ HtmlStartTag stack BR (posAttr posInfo)
+            , HtmlEndTag stack BR
+            ]
+        HTML_INLINE html ->
+            case scanHtml $ LT.fromStrict html of
+                Done "" input' -> return $ rebaseStack input'
+                _ -> return [HtmlCdata stack html]
+        CUSTOM_INLINE _ _ ->
+            return []
+        CODE text' -> return
+            [ HtmlStartTag stack Code (posAttr posInfo)
+            , HtmlCdata (nextStack Code) text'
+            , HtmlEndTag stack Code
+            ]
+        EMPH ->
+            nodeWithChildren Em
+        STRONG ->
+            nodeWithChildren Strong
+        LINK href title ->
+            nodeWithChildren' A $ rawAttr " href" href <> rawAttr " title" title
+        IMAGE src title ->
+            nodeWithChildren' Img $ rawAttr " src" src <> rawAttr " title" title
+      where
+        nextStack :: HtmlTag -> HtmlTagStack
+        nextStack = (`TagStack.push` stack)
+        nodeWithChildren :: (Monad m, MonadFail m) => HtmlTag -> m [HtmlEntity]
+        nodeWithChildren tag' = nodeWithChildren' tag' ""
+        nodeWithChildren' :: (Monad m, MonadFail m)
+                          => HtmlTag -> ST.Text -> m [HtmlEntity]
+        nodeWithChildren' tag' extraAttrs = do
+            mid <- forM children $ do
+                fromNode (nextStack tag')
+            let middle = Data.List.concat mid
+            return $ HtmlStartTag stack tag' (posAttr posInfo <> extraAttrs) :
+                middle ++ [HtmlEndTag stack tag']
+        rebase' :: HtmlTagStack -> HtmlTagStack
+        rebase' = TagStack.rebase [] stack
+        rebaseStack :: [HtmlEntity] -> [HtmlEntity]
+        rebaseStack = Prelude.map (\e -> e { tagStack = rebase' $ tagStack e })
+    toNode :: [HtmlEntity] -> [Node]
+    toNode [] = []
+    toNode (x:xs) = case x of
+        HtmlComment _ comment' ->
+            Node Nothing (htmlNode comment') [] : toNode xs
+        HtmlCdata _ cdata ->
+            Node Nothing (TEXT cdata) [] : toNode xs
+        HtmlText _ rawText' ->
+            Node Nothing (TEXT $ unescape rawText') [] : toNode xs
+        HtmlEndTag _ _ ->
+            toNode xs
+        start@(HtmlStartTag stack tag' attrs) ->
+            let (children', rest) = Data.List.break (endOf stack tag') xs
+                (end, rest') = case rest of
+                    end'@(HtmlEndTag endStack endTag):afterEnd ->
+                        if endStack == stack && endTag == tag'
+                            then ([end'], afterEnd)
+                            else ([], rest)
+                    _ -> ([], rest)
+                posInfo = fromMaybe Nothing
+                    (getAttr attrs "posinfo" :: Maybe (Maybe PosInfo))
+                softbreak = getAttr attrs "softbreak" :: Maybe Bool
+                childrenHtmlNode = htmlNode $ LT.toStrict $ printHtml $
+                    start : children' ++ end
+                nodeType = case tag' of
+                    Article -> DOCUMENT
+                    BlockQuote -> BLOCK_QUOTE
+                    HR -> THEMATIC_BREAK
+                    P -> PARAGRAPH
+                    Pre -> case getAttr "info" attrs of
+                        Just info -> CODE_BLOCK info $ printText' children'
+                        Nothing -> childrenHtmlNode
+                    UL ->
+                        maybe childrenHtmlNode LIST (getAttr attrs "list-attrs")
+                    LI -> ITEM
+                    BR ->
+                        if softbreak == Just True then SOFTBREAK else LINEBREAK
+                    Code -> CODE $ printText' children'
+                    Em -> EMPH
+                    Strong -> STRONG
+                    A -> LINK
+                        (fromMaybe ST.empty $ getRawAttr attrs "href")
+                        (fromMaybe ST.empty $ getRawAttr attrs "title")
+                    Img -> IMAGE
+                        (fromMaybe ST.empty $ getRawAttr attrs "src")
+                        (fromMaybe ST.empty $ getRawAttr attrs "title")
+                    _ ->
+                        maybe childrenHtmlNode HEADING (headingLevel tag')
+                (nodeType', nodeChildren) =
+                    if isNothing posInfo && isNothing softbreak
+                        then (childrenHtmlNode, [])
+                        else (nodeType, toNode children')
+            in
+                Node posInfo nodeType' nodeChildren : toNode rest'
+      where
+        block :: Bool
+        block = case TagStack.last (tagStack x) of
+            Just A -> False
+            Just Em -> False
+            Just H1 -> False
+            Just H2 -> False
+            Just H3 -> False
+            Just H4 -> False
+            Just H5 -> False
+            Just H6 -> False
+            Just P -> False
+            Just Strong -> False
+            Just tag' -> isNothing (headingLevel tag')
+            _ -> True
+        htmlNode :: ST.Text -> NodeType
+        htmlNode
+          | block = HTML_BLOCK
+          | otherwise = HTML_INLINE
+    unescape :: ST.Text -> ST.Text
+    unescape = toStrict . toLazyText . htmlEncodedText
+    rawAttr :: ST.Text -> ST.Text -> ST.Text
+    rawAttr name value = ST.append name $ toStrict $
+        "=\"" <> toLazyText (HTMLEntities.Builder.text value) <> "\""
+    attr :: Show a => ST.Text -> a -> ST.Text
+    attr name value =
+        rawAttr ("data-seonbi-cmark-" <> name) $ ST.pack (show value)
+    attr' :: Show a => ST.Text -> a -> ST.Text
+    attr' name = ST.cons ' ' . attr name
+    posAttr :: Maybe PosInfo -> ST.Text
+    posAttr = attr "posinfo"
+    getRawAttr :: HtmlRawAttrs -> ST.Text -> Maybe ST.Text
+    getRawAttr attrs name =
+        case ST.breakOn prefix attrs of
+            (_, "") -> Nothing
+            (_, head') ->
+                case ST.break (== '"') (ST.drop (ST.length prefix) head') of
+                    (_, "") -> Nothing
+                    (value, _) ->
+                        Just $ toStrict $ toLazyText $ htmlEncodedText value
+      where
+        prefix :: ST.Text
+        prefix = name <> "=\""
+    getAttr :: Read a => HtmlRawAttrs -> ST.Text -> Maybe a
+    getAttr attrs name =
+        case getRawAttr attrs ("data-seonbi-cmark-" <> name) of
+            Nothing -> Nothing
+            Just value -> readMaybe $ ST.unpack value
+    endOf :: HtmlTagStack -> HtmlTag -> HtmlEntity -> Bool
+    endOf stack tag' (HtmlEndTag endStack endTag) =
+        not (endStack `TagStack.descendsFrom` stack) || endTag == tag'
+    endOf _ _ _ = False
+    printText' :: [HtmlEntity] -> ST.Text
+    printText' = toStrict . printText
+
+
+-- | Represents a case-insensitive content type.
+type ContentType = CI ST.Text
+
+-- | Converts a 'Text' to a 'ContentType'.
+contentTypeFromText :: ST.Text -> ContentType
+contentTypeFromText = mk
+
+-- | Converts a 'ContentType' to a 'Text'.
+contentTypeText :: ContentType -> ST.Text
+contentTypeText = original
+
+newtype TransformerTransformer' m =
+    TransformerTransformer' (TransformerTransformer m)
+transformers :: (Monad m, MonadFail m)
+             => [(ContentType, TransformerTransformer' m)]
+transformers =
+    [ ("text/html", TransformerTransformer' asHtmlTransformer)
+    , ("application/xhtml+xml", TransformerTransformer' asXhtmlTransformer)
+    , ("text/plain", TransformerTransformer' asPlainTextTransformer)
+    , ("text/markdown", TransformerTransformer' asCommonMarkTransformer)
+    ]
+
+-- | Supported content types.
+contentTypes :: Set ContentType
+contentTypes = (Data.Set.fromList . Prelude.map fst)
+    (transformers :: [(ContentType, TransformerTransformer' IO)])
+
+getTransformerTransformer :: (Monad m, MonadFail m)
+                          => ContentType
+                          -> Maybe (TransformerTransformer' m)
+getTransformerTransformer contentType =
+    snd <$> Data.List.find ((== contentType) . fst) transformers
+
+-- | Applies an 'HtmlTransformer' to the given text with respect to the
+-- given content type.
+transformWithContentType
+    :: (Monad m, MonadFail m)
+    => ContentType
+    -- ^ A content type.  If the content type is unsupported (i.e. not in
+    -- 'contentTypes'), this function fails.
+    -> HtmlTransformer m
+    -- ^ An 'HtmlTransformer' to apply.
+    -> LT.Text
+    -- ^ A input text to transform.
+    -> m LT.Text
+    -- ^ A transformed text.
+transformWithContentType contentType transformer inputText =
+    case getTransformerTransformer contentType of
+        Nothing -> fail $ ST.unpack $
+            "unknown content type: " <> contentTypeText contentType
+        Just (TransformerTransformer' transformTransformer) ->
+            transformTransformer transformer inputText
diff --git a/src/Text/Seonbi/Facade.hs b/src/Text/Seonbi/Facade.hs
--- a/src/Text/Seonbi/Facade.hs
+++ b/src/Text/Seonbi/Facade.hs
@@ -1,8 +1,10 @@
 {-# LANGUAGE CPP #-}
 {-# LANGUAGE DeriveGeneric #-}
+{-# LANGUAGE KindSignatures #-}
 {-# LANGUAGE NamedFieldPuns #-}
 {-# LANGUAGE OverloadedLists #-}
 {-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE RankNTypes #-}
 #ifdef EMBED_DICTIONARY
 {-# LANGUAGE TemplateHaskell #-}
 #endif
@@ -17,6 +19,11 @@
     , ko_KP
     , ko_KR
     , presets
+      -- * Content types
+    , ContentType
+    , contentTypeFromText
+    , contentTypes
+    , contentTypeText
       -- * Dictionaries
     , HanjaDictionary
     , readDictionaryFile
@@ -31,7 +38,13 @@
     , StopOption (..)
     ) where
 
+#if MIN_VERSION_base(4,13,0)
+import Prelude hiding (MonadFail)
+#endif
+
+import Control.Monad.Fail (MonadFail)
 import Data.Char
+import Data.Kind (Type)
 import Data.Maybe
 import Data.String (IsString)
 import GHC.Exts (IsList (toList))
@@ -58,6 +71,7 @@
 #ifndef EMBED_DICTIONARY
 import Paths_seonbi (getDataDir)
 #endif
+import Text.Seonbi.ContentTypes
 import Text.Seonbi.Hanja
 import Text.Seonbi.Html
 import Text.Seonbi.Punctuation
@@ -71,8 +85,9 @@
 data Monad m => Configuration m a = Configuration
     { -- | An optional debugging logger to print its internal AST.
       debugLogger :: Maybe (HtmlEntity -> m a)
-      -- | Whether to take and result in XHTML instead of HTML.
-    , xhtml :: Bool
+      -- | A content type of the input and output.  It has to be a member of
+      -- 'contentTypes'.
+    , contentType :: ContentType
       -- | An option to decide how quotation marks are rendered.
       -- If 'Nothing' no quotes are transformed.
     , quote :: Maybe QuoteOption
@@ -97,7 +112,7 @@
     show c = "Configuration {\n" <>
         "  debugLogger = " <>
             maybe "Nothing" (const "Just ...") (debugLogger c) <> "," <>
-        "  xhtml = " <> show (xhtml c) <> "," <>
+        "  contentType = " <> show (contentType c) <> "," <>
         "  quote = " <> show (quote c) <> "," <>
         "  arrow = " <> show (cite c) <> "," <>
         "  cite = " <> show (arrow c) <> "," <>
@@ -227,28 +242,26 @@
         show initialSoundLaw <>
         " }"
 
--- | Transforms a given HTML text.  'Nothing' if it fails to parse the given
--- HTML text.
-transformHtmlText :: Monad m => Configuration m a -> Text -> m Text
+-- | Transforms a given text.  'Nothing' if it fails to parse the given
+-- text.
+transformHtmlText :: forall (m :: Type -> Type) a. (Monad m, MonadFail m)
+                  => Configuration m a -> Text -> m Text
 transformHtmlText config =
     fmap LT.toStrict . transformHtmlLazyText config . LT.fromStrict
 
 -- | A lazy version of 'transformHtmlText' function.
-transformHtmlLazyText :: Monad m => Configuration m a -> LT.Text -> m LT.Text
-transformHtmlLazyText config@Configuration { xhtml, debugLogger } htmlText =
-    case scanHtml htmlText of
-        Done "" input -> do
-            case debugLogger of
-                Just logger -> mapM_ logger input
-                Nothing -> return ()
-            return $ printHtml' $ toTransformer config input
-        _ ->
-            fail "failed to parse input"
+transformHtmlLazyText :: forall (m :: Type -> Type) a. (Monad m, MonadFail m)
+                      => Configuration m a -> LT.Text -> m LT.Text
+transformHtmlLazyText config@Configuration { contentType, debugLogger } =
+    transformWithContentType contentType transformerM
   where
-    printHtml' :: [HtmlEntity] -> LT.Text
-    printHtml'
-      | xhtml = printXhtml
-      | otherwise = printHtml
+    transformer :: [HtmlEntity] -> [HtmlEntity]
+    transformer = toTransformer config
+    transformerM = case debugLogger of
+        Nothing -> return <$> transformer
+        Just logger -> \ input -> do
+            mapM_ logger input
+            return $ transformer input
 
 toTransformers :: Monad m => Configuration m a -> [[HtmlEntity] -> [HtmlEntity]]
 toTransformers Configuration { quote
@@ -345,7 +358,7 @@
             , initialSoundLaw = True
             }
         }
-    , xhtml = False
+    , contentType = "text/html"
     }
 
 -- | Preset 'Configuration' for North Korean orthography.
diff --git a/src/Text/Seonbi/Hanja.hs b/src/Text/Seonbi/Hanja.hs
--- a/src/Text/Seonbi/Hanja.hs
+++ b/src/Text/Seonbi/Hanja.hs
@@ -46,6 +46,7 @@
 
 import Text.Seonbi.Hangul
 import Text.Seonbi.Html
+import Text.Seonbi.Html.Lang
 import Text.Seonbi.Html.Preservation
 import Text.Seonbi.Html.TagStack (push)
 import qualified Text.Seonbi.Trie as Trie
@@ -184,13 +185,20 @@
                           else Right (stack, hanj, hang)
                         | (hanj, hang) <- Prelude.zip hanjaWords hangulWords
                         ]
-        | e <- concatMap transform entities
+        | e <- concatMap transform (annotateWithLang entities)
         ]
     splitByDigits :: Text -> [Text]
     splitByDigits = Data.Text.groupBy (\ a b -> isDigit a == isDigit b)
-    transform :: HtmlEntity -> [Either HtmlEntity (HtmlTagStack, Text, Text)]
-    transform entity@HtmlText { tagStack = tagStack', rawText = rawText' }
-      | isPreservedTagStack tagStack' =
+    transform :: LangHtmlEntity
+              -> [Either HtmlEntity (HtmlTagStack, Text, Text)]
+    transform LangHtmlEntity
+                { lang = lang
+                , entity = entity@HtmlText
+                    { tagStack = tagStack'
+                    , rawText = rawText'
+                    }
+                }
+      | isPreservedTagStack tagStack' || isNeverKorean lang =
             [Left entity]
       | otherwise =
             case analyzeHanjaText rawText' of
@@ -202,7 +210,7 @@
                     -- Note that htmlText here can have HTML entities.
                     | (trueIfHanja, htmlText) <- pairs
                     ]
-    transform entity =
+    transform LangHtmlEntity { entity } =
         [Left entity]
     -- FIXME: This should be public:
     debugRenderer :: HanjaWordRenderer -> HanjaWordRenderer
diff --git a/src/Text/Seonbi/Html.hs b/src/Text/Seonbi/Html.hs
--- a/src/Text/Seonbi/Html.hs
+++ b/src/Text/Seonbi/Html.hs
@@ -10,6 +10,7 @@
       --
       -- | See more on "Text.Seonbi.Html.Printer" module.
     , printHtml
+    , printText
     , printXhtml
     , -- * HTML entities
       --
diff --git a/src/Text/Seonbi/Html/Entity.hs b/src/Text/Seonbi/Html/Entity.hs
--- a/src/Text/Seonbi/Html/Entity.hs
+++ b/src/Text/Seonbi/Html/Entity.hs
@@ -8,6 +8,7 @@
 import Text.Seonbi.Html.Tag (HtmlTag)
 import Text.Seonbi.Html.TagStack (HtmlTagStack)
 
+-- | All element attributes in a string.
 type HtmlRawAttrs = Text
 
 -- | An event entity emitted by 'scanHtml'.
diff --git a/src/Text/Seonbi/Html/Lang.hs b/src/Text/Seonbi/Html/Lang.hs
new file mode 100644
--- /dev/null
+++ b/src/Text/Seonbi/Html/Lang.hs
@@ -0,0 +1,166 @@
+{-# LANGUAGE OverloadedStrings #-}
+module Text.Seonbi.Html.Lang
+    ( LangHtmlEntity (..)
+    , LanguageTag
+    , annotateWithLang
+    , extractLang
+    , isKorean
+    , isNeverKorean
+    ) where
+
+import Control.Applicative
+import Data.Char (isSpace)
+import Data.Maybe
+
+import Data.Attoparsec.Text
+import Data.Text
+
+import Text.Seonbi.Html.Entity
+import Text.Seonbi.Html.Tag (HtmlTag)
+
+-- | Represents a language tag.  Although it is defined as an alias for 'Text',
+-- it can be structured in the future.  Do not use its contents directly.
+type LanguageTag = Text
+
+-- | Extracts the language tag from the given raw HTML attributes if it has
+-- @lang@ attribute.
+--
+-- >>> extractLang ""
+-- Nothing
+-- >>> extractLang "lang=en"
+-- Just "en"
+-- >>> extractLang "lang=\"ko-KR\""
+-- Just "ko-kr"
+-- >>> extractLang " lang='ko-Hang'"
+-- Just "ko-hang"
+extractLang
+    :: HtmlRawAttrs
+    -- ^ A raw HTML attributes to extract the language tag from.
+    -> Maybe LanguageTag
+    -- ^ A language tag extracted from the given raw HTML attributes.
+    -- If the given raw HTML attributes does not have @lang@ attribute or
+    -- its value is invalid, 'Nothing' is returned.
+extractLang attrs =
+    case parseOnly parser' attrs of
+        Right (Just lang') ->
+            let lt = toLower . strip . normalizeEntities $ lang'
+            in if Data.Text.null lt then Nothing else Just lt
+        _ -> Nothing
+  where
+    parser' :: Parser (Maybe Text)
+    parser' = do
+        skipSpace
+        attrs' <- langAttr `sepBy` space
+        skipSpace
+        return $ listToMaybe $ catMaybes attrs'
+    langAttr :: Parser (Maybe Text)
+    langAttr = do
+        (isLang, cont) <- attrIsLang
+        value <- if cont then attrValue else return ""
+        return (if isLang then Just value else Nothing)
+    attrIsLang :: Parser (Bool, Bool)
+    attrIsLang = choice
+        [ asciiCI "lang=" >> return (True, True)
+        , do { _ <- takeWhile1 (/= '=')
+             ; eq <- optional (char '=')
+             ; return (False, isJust eq)
+             }
+        ]
+    attrValue :: Parser Text
+    attrValue = choice
+        [ do { skip (== '"'); v <- takeTill (== '"'); skip (== '"'); return v }
+        , do { skip (== '\'')
+             ; v <- takeTill (== '\'')
+             ; skip (== '\''); return v
+             }
+        , takeWhile1 (not . isSpace)
+        ]
+    normalizeEntities :: Text -> Text
+    normalizeEntities
+        = Data.Text.replace "&hyphen;" "-"
+        . Data.Text.replace "&dash;" "-"
+        . Data.Text.replace "&#8208;" "-"
+        . Data.Text.replace "&#x2010;" "-"
+        . Data.Text.replace "&#X2010;" "-"
+
+-- | Annotates 'HtmlEntity' with the 'lang' tag extracted from it or its
+-- ancestors.
+data LangHtmlEntity = LangHtmlEntity
+    { -- | The @lang@ tag extracted from the HTML 'entity' or its ancestors.
+      lang :: Maybe LanguageTag
+      -- | The annotated HTML 'entity'.
+    , entity :: HtmlEntity
+    } deriving (Show, Eq)
+
+-- | Annotates the given HTML entities with the language tag extracted from
+-- their @lang@ attributes.  If a parent entity has @lang@ attribute, its
+-- all descendants are annotated with the same language tag.
+annotateWithLang :: [HtmlEntity] -> [LangHtmlEntity]
+annotateWithLang =
+    annotate []
+  where
+    annotate :: [(HtmlTag, Maybe LanguageTag)]
+             -> [HtmlEntity]
+             -> [LangHtmlEntity]
+    annotate _ [] = []
+    annotate stack (x@HtmlStartTag { tag = tag', rawAttributes = attrs } : xs) =
+        LangHtmlEntity thisLang x : annotate nextStack xs
+      where
+        parentLang :: Maybe LanguageTag
+        parentLang = case stack of
+            (_, l):_ -> l
+            _ -> Nothing
+        thisLang :: Maybe LanguageTag
+        thisLang = extractLang attrs <|> parentLang
+        nextStack :: [(HtmlTag, Maybe LanguageTag)]
+        nextStack = (tag', thisLang) : stack
+    annotate stack (x@HtmlEndTag { tag = tag' } : xs) =
+        LangHtmlEntity thisLang x : annotate nextStack xs
+      where
+        (nextStack, thisLang) = case stack of
+            [] -> ([], Nothing)
+            s@((t, lang'):ys) ->
+                (if t == tag' then ys else s, lang')
+    annotate stack (x : xs) =
+        LangHtmlEntity parentLang x : annotate stack xs
+      where
+        parentLang :: Maybe LanguageTag
+        parentLang = case stack of
+            (_, l):_ -> l
+            _ -> Nothing
+
+-- | Determines whether the given language tag refers to any kind of Korean.
+--
+-- >>> isKorean "ko"
+-- True
+-- >>> isKorean "ko-KR"
+-- True
+-- >>> isKorean "kor-Hang"
+-- True
+-- >>> isKorean "en"
+-- False
+-- >>> isKorean "en-KR"
+-- False
+isKorean :: LanguageTag -> Bool
+isKorean lang' =
+    l == "ko" || l == "kor" ||
+    "ko-" `isPrefixOf` l ||
+    "kor-" `isPrefixOf` l
+  where
+    l :: Text
+    l = toLower lang'
+
+-- | Determines whether the given language tag undoubtedly does not refer
+-- to any kind of Korean.
+--
+-- >>> isNeverKorean $ Just "ko"
+-- False
+-- >>> isNeverKorean $ Just "ko-KR"
+-- False
+-- >>> isNeverKorean Nothing
+-- False
+-- >>> isNeverKorean $ Just "en"
+-- True
+isNeverKorean :: Maybe LanguageTag -> Bool
+isNeverKorean Nothing = False
+isNeverKorean (Just lang') = not (isKorean lang')
diff --git a/src/Text/Seonbi/Html/Printer.hs b/src/Text/Seonbi/Html/Printer.hs
--- a/src/Text/Seonbi/Html/Printer.hs
+++ b/src/Text/Seonbi/Html/Printer.hs
@@ -1,6 +1,8 @@
+{-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE OverloadedStrings #-}
 module Text.Seonbi.Html.Printer
     ( printHtml
+    , printText
     , printXhtml
     ) where
 
@@ -9,14 +11,19 @@
 
 import qualified Data.Text
 import Data.Text.Lazy
+import Data.Text.Lazy.Builder
+import HTMLEntities.Decoder
 
 import Text.Seonbi.Html.Entity
 import Text.Seonbi.Html.Tag
 
--- | Print the list of 'HtmlEntity' into a lazy 'Text'.
---
+-- $setup
 -- >>> :set -XOverloadedStrings
 -- >>> import Text.Seonbi.Html.Scanner
+-- >>> :set -interactive-print=Text.Show.Unicode.uprint
+
+-- | Print the list of 'HtmlEntity' into a lazy 'Text'.
+--
 -- >>> let Done "" tokens = scanHtml "<p>Hello,<br>\n<em>world</em>!</p>"
 -- >>> printHtml tokens
 -- "<p>Hello,<br>\n<em>world</em>!</p>"
@@ -26,8 +33,6 @@
 -- | Similar to 'printHtml' except it renders void (self-closing) tags as
 -- like @<br/>@ instead of @<br>@.
 --
--- >>> :set -XOverloadedStrings
--- >>> import Text.Seonbi.Html.Scanner
 -- >>> let Done "" tokens = scanHtml "<p>Hello,<br>\n<em>world</em>!</p>"
 -- >>> printXhtml tokens
 -- "<p>Hello,<br/>\n<em>world</em>!</p>"
@@ -72,3 +77,26 @@
     renderAttrs attrs
       | isSpace (Data.Text.head attrs) = fromStrict attrs
       | otherwise = ' ' `cons` fromStrict attrs
+
+-- | Print only the text contents (including CDATA sections) without tags
+-- into a lazy 'Text'.
+--
+-- >>> let Done "" tokens = scanHtml "<p>Hello,<br>\n<em>world</em>!</p>"
+-- >>> printText tokens
+-- "Hello,\nworld!"
+--
+-- Entities are decoded:
+--
+-- >>> let Done "" tokens = scanHtml "<p><code>&lt;&gt;&quot;&amp;</code></p>"
+-- >>> printText tokens
+-- "<>\"&"
+printText :: [HtmlEntity] -> Text
+printText [] = Data.Text.Lazy.empty
+printText (x:xs) =
+    render x <> printText xs
+  where
+    render :: HtmlEntity -> Text
+    render = \ case
+        HtmlText { rawText = t } -> toLazyText $ htmlEncodedText t
+        HtmlCdata { text = t } -> fromStrict t
+        _ -> Data.Text.Lazy.empty
diff --git a/src/Text/Seonbi/Html/Tag.hs b/src/Text/Seonbi/Html/Tag.hs
--- a/src/Text/Seonbi/Html/Tag.hs
+++ b/src/Text/Seonbi/Html/Tag.hs
@@ -2,12 +2,16 @@
 module Text.Seonbi.Html.Tag
     ( HtmlTag (..)
     , HtmlTagKind (..)
+    , headingLevel
+    , headingTag
+    , headingTag'
     , htmlTagKind
     , htmlTagName
     , htmlTagNames
     , htmlTags
     ) where
 
+import Data.Maybe
 import Data.Map.Strict
 import Data.Set
 import Data.Text
@@ -206,7 +210,7 @@
         [(htmlTagName t, t) | t <- Data.Set.toList htmlTags]
 
 -- | The kind of an 'HtmlTag'.
--- 
+--
 -- >>> Data.Set.filter ((== EscapableRawText) . htmlTagKind) htmlTags
 -- fromList [TextArea,Title]
 htmlTagKind :: HtmlTag -> HtmlTagKind
@@ -325,3 +329,57 @@
     Var -> Normal
     Video -> Normal
     WBR -> Void
+
+-- | Get the heading level of an 'HtmlTag', if it is a heading tag
+-- ('H1' to 'H6').
+--
+-- >>> headingLevel H1
+-- Just 1
+-- >>> headingLevel H6
+-- Just 6
+-- >>> headingLevel P
+-- Nothing
+headingLevel :: HtmlTag -> Maybe Int
+headingLevel = \ case
+    H1 -> Just 1
+    H2 -> Just 2
+    H3 -> Just 3
+    H4 -> Just 4
+    H5 -> Just 5
+    H6 -> Just 6
+    _ -> Nothing
+
+-- | Get the heading tag with the given heading level.  If the level is
+-- invalid, then 'Nothing' is returned.
+--
+-- >>> headingTag 1
+-- Just H1
+-- >>> headingTag 6
+-- Just H6
+-- >>> headingTag 7
+-- Nothing
+headingTag :: Int -> Maybe HtmlTag
+headingTag = \ case
+    1 -> Just H1
+    2 -> Just H2
+    3 -> Just H3
+    4 -> Just H4
+    5 -> Just H5
+    6 -> Just H6
+    _ -> Nothing
+
+-- | Get the heading tag with the given heading level.  If the level is
+-- greater than 6, then 'H6' is returned.  If the level is less than 1,
+-- then 'H1' is returned.
+--
+-- >>> headingTag' 1
+-- H1
+-- >>> headingTag' 6
+-- H6
+-- >>> headingTag' 0
+-- H1
+-- >>> headingTag' 7
+-- H6
+headingTag' :: Int -> HtmlTag
+headingTag' level =
+    fromMaybe (if level > 6 then H6 else H1) $ headingTag level
diff --git a/src/Text/Seonbi/Html/TagStack.hs b/src/Text/Seonbi/Html/TagStack.hs
--- a/src/Text/Seonbi/Html/TagStack.hs
+++ b/src/Text/Seonbi/Html/TagStack.hs
@@ -7,13 +7,16 @@
     , depth
     , empty
     , fromList
+    , last
     , pop
     , push
     , rebase
     , toList
     ) where
 
-import Data.List
+import Prelude hiding (last)
+
+import Data.List hiding (last)
 import GHC.Exts (IsList (..))
 
 import Text.Seonbi.Html.Tag
@@ -50,6 +53,18 @@
 depth :: HtmlTagStack -> Int
 depth (HtmlTagStack stack) = Data.List.length stack
 
+-- | Get the deepest tag from a 'HtmlTagStack'.
+--
+-- >>> :set -XOverloadedLists
+-- >>> let stack = [Div, Article, P, Em] :: HtmlTagStack
+-- >>> last stack
+-- Just Em
+-- >>> last []
+-- Nothing
+last :: HtmlTagStack -> Maybe HtmlTag
+last (HtmlTagStack []) = Nothing
+last (HtmlTagStack (tag:_)) = Just tag
+
 -- | Build a new stack from a stack by replacing its bottom with a new base.
 --
 -- >>> :set -XOverloadedLists
@@ -107,7 +122,7 @@
                 HtmlTagStack (head' ++ tail')
 pop _ (HtmlTagStack []) = empty
 
--- | Check if a node ('HtmlEntity') that a 'HtmlTagStack' (the first argument) 
+-- | Check if a node ('HtmlEntity') that a 'HtmlTagStack' (the first argument)
 -- refers is contained by a node that another 'HtmlTagStack' (the second
 -- argument), or they are sibling at least.
 --
diff --git a/src/Text/Seonbi/Punctuation.hs b/src/Text/Seonbi/Punctuation.hs
--- a/src/Text/Seonbi/Punctuation.hs
+++ b/src/Text/Seonbi/Punctuation.hs
@@ -51,6 +51,7 @@
 
 import Text.Seonbi.Html
 import Text.Seonbi.Html.Clipper
+import Text.Seonbi.Html.Lang
 import Text.Seonbi.Html.Preservation
 import Text.Seonbi.Html.Wrapper
 import Text.Seonbi.PairedTransformer
@@ -295,14 +296,18 @@
 
 -- | Normalizes sentence stops (periods, commas, and interpuncts).
 normalizeStops :: Stops -> [HtmlEntity] -> [HtmlEntity]
-normalizeStops stops input = (`fmap` normalizeText input) $ \ case
-    e@HtmlText { tagStack = stack, rawText = txt } ->
-        if isPreservedTagStack stack
+normalizeStops stops input = (`fmap` annotatedEntities) $ \ case
+    LangHtmlEntity { lang = l
+                   , entity = e@HtmlText { tagStack = stack, rawText = txt }
+                   } ->
+        if isPreservedTagStack stack || isNeverKorean l
         then e
         else e { rawText = replaceText txt }
-    e ->
+    LangHtmlEntity { entity = e } ->
         e
   where
+    annotatedEntities :: [LangHtmlEntity]
+    annotatedEntities = (annotateWithLang . normalizeText) input
     replaceText :: Text -> Text
     replaceText txt =
         case parseOnly parser txt of
diff --git a/src/Text/Seonbi/Trie.hs b/src/Text/Seonbi/Trie.hs
--- a/src/Text/Seonbi/Trie.hs
+++ b/src/Text/Seonbi/Trie.hs
@@ -1,17 +1,148 @@
 {-# OPTIONS_GHC -fno-warn-orphans #-}
+{-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE RankNTypes #-}
 {-# LANGUAGE TypeFamilies #-}
--- | This re-exports 'Data.Trie.Text'.
+-- | A trie from 'Text' keys to values.
 module Text.Seonbi.Trie
-    ( module Data.Trie.Text
+    ( Trie
+    , elems
+    , empty
+    , fromList
+    , insert
+    , keys
+    , lookup
+    , member
+    , mergeBy
+    , null
+    , singleton
+    , size
+    , toList
+    , unionL
+    , unionR
     ) where
 
-import GHC.Exts
+import Prelude hiding (lookup, null)
 
-import Data.Text
-import Data.Text.Lazy (toStrict)
-import Data.Trie.Text
+import Control.Monad (ap)
+import qualified GHC.Exts
 
-instance IsList (Trie a) where
+import Data.ByteString (ByteString)
+import Data.Text hiding (empty, null, singleton)
+import Data.Text.Encoding (encodeUtf8, decodeUtf8)
+import qualified Data.Trie as BTrie
+
+-- | A trie from 'Text' keys to 'a' values.
+newtype Trie a
+  = Trie (BTrie.Trie a)
+  deriving (Eq, Show)
+
+encodeKey :: Text -> ByteString
+encodeKey = encodeUtf8
+
+decodeKey :: ByteString -> Text
+decodeKey = decodeUtf8
+
+-- | The empty trie.
+empty :: Trie a
+empty = Trie BTrie.empty
+
+-- | Checks if the trie is empty.
+null :: Trie a -> Bool
+null (Trie btrie) = BTrie.null btrie
+
+-- | Constructs a singleton trie.
+singleton :: Text -> a -> Trie a
+singleton key value = Trie $ BTrie.singleton (encodeKey key) value
+
+-- | Gets the number of elements in the trie.
+size :: Trie a -> Int
+size (Trie btrie) = BTrie.size btrie
+
+fromList' :: [(Text, a)] -> Trie a
+fromList' list = Trie $ BTrie.fromList [(encodeKey k, v) | (k, v) <- list]
+
+toList' :: Trie a -> [(Text, a)]
+toList' (Trie btrie) = [(decodeKey k, v) | (k, v) <- BTrie.toList btrie]
+
+-- | Converts a list of associated pairs into a trie.  For duplicate keys,
+-- values earlier in the list shadow later ones.
+fromList :: [(Text, a)] -> Trie a
+fromList = fromList'
+
+-- | Converts a trie into a list of associated pairs.  Keys will be ordered.
+toList :: Trie a -> [(Text, a)]
+toList = toList'
+
+-- | Lists all keys in the trie.  Keys will be ordered.
+keys :: Trie a -> [Text]
+keys (Trie btrie) = Prelude.map decodeKey $ BTrie.keys btrie
+
+-- | Lists all values in the trie.  Values are ordered by their associated keys.
+elems :: Trie a -> [a]
+elems (Trie btrie) = BTrie.elems btrie
+
+-- | Gets the value associated with a key if it exists.
+lookup :: Text -> Trie a -> Maybe a
+lookup key (Trie btrie) = BTrie.lookup (encodeKey key) btrie
+
+-- | Checks if a key has a value in the trie.
+member :: Text -> Trie a -> Bool
+member key (Trie btrie) = BTrie.member (encodeKey key) btrie
+
+-- | Inserts a new key into the trie.
+insert
+    :: Text
+    -- ^ A new key to insert.  If there is already the same key in the trie,
+    -- the existing value is overwritten by the new value.
+    -> a
+    -- ^ A value associated to the key.
+    -> Trie a
+    -- ^ An existing trie.
+    -> Trie a
+    -- ^ The new trie with the inserted key.
+insert key value (Trie btrie) = Trie $ BTrie.insert (encodeKey key) value btrie
+
+-- | Combines two tries, using a function to resolve collisions.  This can only
+-- define the space of functions between union and symmetric difference but,
+-- with those two, all set operations can be defined (albeit inefficiently).
+mergeBy :: (a -> a -> Maybe a) -> Trie a -> Trie a -> Trie a
+mergeBy f (Trie a) (Trie b) = Trie $ BTrie.mergeBy f a b
+
+-- | Combines two tries, resolving conflicts by choosing the value from the
+-- left (former) trie.
+unionL :: Trie a -> Trie a -> Trie a
+unionL (Trie left) (Trie right) = Trie $ BTrie.unionL left right
+
+-- | Combines two tries, resolving conflicts by choosing the value from the
+-- right (latter) trie.
+unionR :: Trie a -> Trie a -> Trie a
+unionR (Trie left) (Trie right) = Trie $ BTrie.unionR left right
+
+instance Functor Trie where
+    fmap f (Trie btrie) = Trie $ fmap f btrie
+
+instance Foldable Trie where
+    foldMap f (Trie btrie) = foldMap f btrie
+
+instance Traversable Trie where
+    traverse f (Trie btrie) = Trie <$> traverse f btrie
+
+instance Applicative Trie where
+    pure = return
+    (<*>) = ap
+
+instance Monad Trie where
+    return = singleton ""
+    Trie btrie >>= f = Trie $ btrie >>= (\ v -> case f v of { Trie b -> b })
+
+instance (Semigroup a) => Semigroup (Trie a) where
+    (Trie a) <> (Trie b) = Trie (a <> b)
+
+instance (Monoid a) => Monoid (Trie a) where
+    mempty = Trie mempty
+    mappend (Trie a) (Trie b) = Trie $ mappend a b
+
+instance GHC.Exts.IsList (Trie a) where
     type Item (Trie a) = (Text, a)
-    fromList = Data.Trie.Text.fromList
-    toList trie = [(toStrict k, v) | (k, v) <- Data.Trie.Text.toList trie]
+    fromList = fromList'
+    toList = toList'
diff --git a/test/Text/Seonbi/ContentTypesSpec.hs b/test/Text/Seonbi/ContentTypesSpec.hs
new file mode 100644
--- /dev/null
+++ b/test/Text/Seonbi/ContentTypesSpec.hs
@@ -0,0 +1,56 @@
+{-# LANGUAGE OverloadedStrings #-}
+module Text.Seonbi.ContentTypesSpec (spec) where
+
+import Data.Text
+import qualified Data.Text.Lazy as LT
+import Data.Text.Lazy.Builder
+
+import Test.Hspec
+
+import Text.Seonbi.Html
+import Text.Seonbi.ContentTypes
+import qualified HTMLEntities.Builder
+import HTMLEntities.Decoder
+
+textReverser :: (Monad m, MonadFail m) => HtmlTransformer m
+textReverser entities =
+    return $ reverseText <$> entities
+  where
+    reverseText :: HtmlEntity -> HtmlEntity
+    reverseText e@HtmlText { rawText = t } =
+        e { rawText = encode $ Data.Text.reverse $ decode t }
+    reverseText e@HtmlCdata { text = t } =
+        e { text = Data.Text.reverse t }
+    reverseText e =
+        e
+    decode :: Text -> Text
+    decode = LT.toStrict . toLazyText . htmlEncodedText
+    encode :: Text -> Text
+    encode = LT.toStrict . toLazyText . HTMLEntities.Builder.text
+
+spec :: Spec
+spec = do
+    specify "asHtmlTransformer" $ do
+        r <- asHtmlTransformer textReverser "<p>foo <em>bar</em><br> baz</p>"
+        r `shouldBe` "<p> oof<em>rab</em><br>zab </p>"
+    specify "asXhtmlTransformer" $ do
+        r <- asXhtmlTransformer textReverser "<p>foo <em>bar</em><br> baz</p>"
+        r `shouldBe` "<p> oof<em>rab</em><br/>zab </p>"
+    specify "asPlainTextTransformer" $ do
+        r <- asPlainTextTransformer textReverser
+                "<p>foo <em>bar</em><br> baz</p>"
+        r `shouldBe` ">p/<zab >rb<>me/<rab>me< oof>p<"
+    specify "asCommonMarkTransformer" $ do
+        r <- asCommonMarkTransformer textReverser
+                "# Foo\n\nBar *Baz*\nQux\n\n> Quote <em>tag</em>\n"
+        r `shouldBe` "# ooF\n\n raB*zaB*\nxuQ\n\n>  etouQ<em>gat</em>\n"
+    specify "transformWithContentType" $ do
+        let input = "*foo* <em>bar</em><br>"
+        h <- transformWithContentType "text/html" textReverser input
+        h `shouldBe` " *oof*<em>rab</em><br>"
+        x <- transformWithContentType "application/xhtml+xml" textReverser input
+        x `shouldBe` " *oof*<em>rab</em><br/>"
+        p <- transformWithContentType "text/plain" textReverser input
+        p `shouldBe` ">rb<>me/<rab>me< *oof*"
+        m <- transformWithContentType "text/markdown" textReverser input
+        m `shouldBe` "*oof* <em>rab</em><br>\n"
diff --git a/test/Text/Seonbi/FacadeSpec.hs b/test/Text/Seonbi/FacadeSpec.hs
--- a/test/Text/Seonbi/FacadeSpec.hs
+++ b/test/Text/Seonbi/FacadeSpec.hs
@@ -25,7 +25,7 @@
     , (".ko-KP.html", ko_KP)
     ]
 
-shouldHaveSameText :: (HasCallStack) => Text -> Text -> Expectation
+shouldHaveSameText :: HasCallStack => Text -> Text -> Expectation
 actual `shouldHaveSameText` expected =
     unless (actual == expected) (expectationFailure msg)
   where
@@ -79,6 +79,6 @@
         , emDash = False
         , stop = Nothing
         , hanja = Nothing
-        , xhtml = False
+        , contentType = "text/html"
         , debugLogger = Nothing
         }
diff --git a/test/Text/Seonbi/HangulSpec.hs b/test/Text/Seonbi/HangulSpec.hs
--- a/test/Text/Seonbi/HangulSpec.hs
+++ b/test/Text/Seonbi/HangulSpec.hs
@@ -1,4 +1,3 @@
-{-# LANGUAGE OverloadedLists #-}
 {-# LANGUAGE OverloadedStrings #-}
 module Text.Seonbi.HangulSpec (spec) where
 
diff --git a/test/Text/Seonbi/HanjaSpec.hs b/test/Text/Seonbi/HanjaSpec.hs
--- a/test/Text/Seonbi/HanjaSpec.hs
+++ b/test/Text/Seonbi/HanjaSpec.hs
@@ -302,6 +302,33 @@
             let phone = normalizeText . phoneticizeHanja def
             phone [HtmlText [Pre] "1996年 그들이 地球를 支配했을 때"] `shouldBe`
                 [HtmlText [Pre] "1996年 그들이 地球를 支配했을 때"]
+        it "transforms nothing in non-Korean elements" $ do
+            let phone = normalizeText . phoneticizeHanja def
+            let input =
+                    [ HtmlStartTag [] P ""
+                    , HtmlText [P] "2004年 開封한 日本 映畫 "
+                    , HtmlStartTag [P] Span "lang=\"ja\""
+                    , HtmlText [P, Span] "誰も知らない"
+                    , HtmlEndTag [P] Span
+                    , HtmlText [P] "는 이듬해 韓國에서도 "
+                    , HtmlStartTag [P] Span "lang=ko-Hang"
+                    , HtmlText [P, Span] "아무도 모른다"
+                    , HtmlEndTag [P] Span
+                    , HtmlText [P] "라는 題目으로 開封했다."
+                    ]
+            let output =
+                    [ HtmlStartTag [] P ""
+                    , HtmlText [P] "2004년 개봉한 일본 영화 "
+                    , HtmlStartTag [P] Span "lang=\"ja\""
+                    , HtmlText [P, Span] "誰も知らない"
+                    , HtmlEndTag [P] Span
+                    , HtmlText [P] "는 이듬해 한국에서도 "
+                    , HtmlStartTag [P] Span "lang=ko-Hang"
+                    , HtmlText [P, Span] "아무도 모른다"
+                    , HtmlEndTag [P] Span
+                    , HtmlText [P] "라는 제목으로 개봉했다."
+                    ]
+            phone input `shouldBe` output
     describe "convertInitialSoundLaw" $ do
         specify "녀, 뇨, 뉴, 니 should be 여, 요, 유, 이" $ do
             convertInitialSoundLaw '녀' `shouldBe` '여'
diff --git a/test/Text/Seonbi/Html/LangSpec.hs b/test/Text/Seonbi/Html/LangSpec.hs
new file mode 100644
--- /dev/null
+++ b/test/Text/Seonbi/Html/LangSpec.hs
@@ -0,0 +1,111 @@
+{-# LANGUAGE OverloadedLists #-}
+{-# LANGUAGE OverloadedStrings #-}
+module Text.Seonbi.Html.LangSpec (spec) where
+
+import Test.Hspec
+
+import Text.Seonbi.Html
+import Text.Seonbi.Html.Lang
+
+source :: [HtmlEntity]
+source =
+    [ HtmlStartTag
+        { tagStack = []
+        , tag = P
+        , rawAttributes = "id=\"foo\" lang=\"en\""
+        }
+    , HtmlText { tagStack = [P], rawText = "English" }
+    , HtmlEndTag { tagStack = [], tag = P }
+    , HtmlStartTag { tagStack = [], tag = Div, rawAttributes = "" }
+    , HtmlStartTag
+        { tagStack = [Div]
+        , tag = P
+        , rawAttributes = "class=bar lang=ja"
+        }
+    , HtmlStartTag { tagStack = [Div, P], tag = B, rawAttributes = "" }
+    , HtmlText { tagStack = [Div, P, B], rawText = "日本語" }
+    , HtmlEndTag { tagStack = [Div, P], tag = B }
+    , HtmlStartTag
+        { tagStack = [Div, P]
+        , tag = Span
+        , rawAttributes = "lang='yue-Hant'"
+        }
+    , HtmlText { tagStack = [Div, P, Span], rawText = "與" }
+    , HtmlStartTag { tagStack = [Div, P, Span], tag = B, rawAttributes = "" }
+    , HtmlText { tagStack = [Div, P, Span, B], rawText = "與粵語" }
+    , HtmlEndTag { tagStack = [Div, P, Span], tag = B }
+    , HtmlEndTag { tagStack = [Div, P], tag = Span }
+    , HtmlEndTag { tagStack = [Div], tag = P }
+    , HtmlEndTag { tagStack = [], tag = Div }
+    ]
+
+annotated :: [LangHtmlEntity]
+annotated =
+    [ LangHtmlEntity
+        (Just "en")
+        HtmlStartTag
+            { tagStack = []
+            , tag = P
+            , rawAttributes = "id=\"foo\" lang=\"en\""
+            }
+    , LangHtmlEntity
+        (Just "en")
+        HtmlText { tagStack = [P], rawText = "English" }
+    , LangHtmlEntity (Just "en") HtmlEndTag { tagStack = [], tag = P }
+    , LangHtmlEntity
+        Nothing
+        HtmlStartTag { tagStack = [], tag = Div, rawAttributes = "" }
+    , LangHtmlEntity
+        (Just "ja")
+        HtmlStartTag
+            { tagStack = [Div]
+            , tag = P
+            , rawAttributes = "class=bar lang=ja"
+            }
+    , LangHtmlEntity
+        (Just "ja")
+        HtmlStartTag { tagStack = [Div, P], tag = B, rawAttributes = "" }
+    , LangHtmlEntity
+        (Just "ja")
+        HtmlText { tagStack = [Div, P, B], rawText = "日本語" }
+    , LangHtmlEntity (Just "ja") HtmlEndTag { tagStack = [Div, P], tag = B }
+    , LangHtmlEntity
+        (Just "yue-hant")
+        HtmlStartTag
+            { tagStack = [Div, P]
+            , tag = Span
+            , rawAttributes = "lang='yue-Hant'"
+            }
+    , LangHtmlEntity
+        (Just "yue-hant")
+        HtmlText { tagStack = [Div, P, Span], rawText = "與" }
+    , LangHtmlEntity
+        (Just "yue-hant")
+        HtmlStartTag { tagStack = [Div, P, Span], tag = B, rawAttributes = "" }
+    , LangHtmlEntity
+        (Just "yue-hant")
+        HtmlText { tagStack = [Div, P, Span, B], rawText = "與粵語" }
+    , LangHtmlEntity
+        (Just "yue-hant")
+        HtmlEndTag { tagStack = [Div, P, Span], tag = B }
+    , LangHtmlEntity
+        (Just "yue-hant")
+        HtmlEndTag { tagStack = [Div, P], tag = Span }
+    , LangHtmlEntity (Just "ja") HtmlEndTag { tagStack = [Div] , tag = P }
+    , LangHtmlEntity Nothing HtmlEndTag { tagStack = [], tag = Div }
+    ]
+
+spec :: Spec
+spec = do
+    specify "extractLang" $ do
+        extractLang "" `shouldBe` Nothing
+        extractLang "lang=en" `shouldBe` Just "en"
+        extractLang "lang=en-US" `shouldBe` Just "en-us"
+        extractLang "lang='ko-KR'" `shouldBe` Just "ko-kr"
+        extractLang "lang=\"zh-Hant\"" `shouldBe` Just "zh-hant"
+        extractLang "lang=\"yue-Hans-HK\"" `shouldBe` Just "yue-hans-hk"
+        extractLang "id=\"foo\" lang=\"en\"" `shouldBe` Just "en"
+        extractLang "id=\"foo\" lang=zh-CN class=bar" `shouldBe` Just "zh-cn"
+    specify "annotateWithLang" $ do
+        annotateWithLang [] `shouldBe` []
+        annotateWithLang source `shouldBe` annotated
diff --git a/test/Text/Seonbi/PunctuationSpec.hs b/test/Text/Seonbi/PunctuationSpec.hs
--- a/test/Text/Seonbi/PunctuationSpec.hs
+++ b/test/Text/Seonbi/PunctuationSpec.hs
@@ -535,3 +535,20 @@
                     , HtmlEndTag [P] Span
                     , HtmlEndTag [] P
                     ]
+        it "does not touch inside non-Korean contents" $ do
+            let input =
+                    [ HtmlStartTag [] P ""
+                    , HtmlText [P] "《大學》의 「三綱領」에 다음과 같은 말이 나온다。"
+                    , HtmlEndTag [] P
+                    , HtmlStartTag [] BlockQuote "lang=lzh"
+                    , HtmlText [BlockQuote] "大學之道，在明明德，在親民，在止於至善。"
+                    , HtmlEndTag [] BlockQuote
+                    ] :: [HtmlEntity]
+            normalizeStops horizontalStops input `shouldBe`
+                    [ HtmlStartTag [] P ""
+                    , HtmlText [P] "《大學》의 「三綱領」에 다음과 같은 말이 나온다."
+                    , HtmlEndTag [] P
+                    , HtmlStartTag [] BlockQuote "lang=lzh"
+                    , HtmlText [BlockQuote] "大學之道，在明明德，在親民，在止於至善。"
+                    , HtmlEndTag [] BlockQuote
+                    ]
diff --git a/test/doctest.hs b/test/doctest.hs
--- a/test/doctest.hs
+++ b/test/doctest.hs
@@ -1,1 +1,15 @@
+{-# LANGUAGE CPP #-}
+#if __GLASGOW_HASKELL__ >= 810
+main :: IO ()
+main = do
+    putStrLn "Temporarily, doctests are ignored for GHC >= 8.10 due to bugs:\n"
+    putStrLn "  https://github.com/sol/doctest/issues/301"
+#elif __GLASGOW_HASKELL__ >= 808 && defined(mingw32_HOST_OS)
+main :: IO ()
+main = do
+    putStr "Temporarily, doctests are ignored for GHC >= 8.8 on Windows "
+    putStrLn "due to bugs:\n"
+    putStrLn "  https://github.com/sol/doctest/issues/300"
+#else
 {-# OPTIONS_GHC -F -pgmF doctest-discover -optF test/doctest.json #-}
+#endif