diff --git a/gf.cabal b/gf.cabal
index 7fdbf392f..ee21264fc 100644
--- a/gf.cabal
+++ b/gf.cabal
@@ -86,7 +86,8 @@ Library
                  -- For compatability with ghc < 8
                  -- We need transformers-compat >= 0.6.3, but that is only in newer snapshots where it is redundant.
                  transformers-compat,
-                 ghc-prim
+                 ghc-prim,
+                 text
   hs-source-dirs: src/runtime/haskell
 
   other-modules:
@@ -536,6 +537,7 @@ test-suite lpgf
     process,
     random,
     terminfo,
+    text,
     time,
     transformers-compat,
     unix,
diff --git a/src/compiler/GF/Compile/GrammarToLPGF.hs b/src/compiler/GF/Compile/GrammarToLPGF.hs
index 9ae1545dd..ae03f0e2c 100644
--- a/src/compiler/GF/Compile/GrammarToLPGF.hs
+++ b/src/compiler/GF/Compile/GrammarToLPGF.hs
@@ -18,12 +18,13 @@ import Data.Either (lefts, rights)
 import Data.List (elemIndex, find, groupBy, sortBy)
 import qualified Data.Map as Map
 import Data.Maybe (catMaybes)
+import qualified Data.Text as T
 import Text.Printf (printf)
 
 mkCanon2lpgf :: Options -> SourceGrammar -> ModuleName -> IOE LPGF
 mkCanon2lpgf opts gr am = do
-  (an,abs) <- mkAbstr ab
-  cncs     <- mapM mkConcr cncs
+  (an,abs) <- mkAbstract ab
+  cncs     <- mapM mkConcrete cncs
   let lpgf = LPGF {
     L.absname = an,
     L.abstract = abs,
@@ -36,11 +37,11 @@ mkCanon2lpgf opts gr am = do
   where
     canon@(C.Grammar ab cncs) = grammar2canonical opts am gr
 
-    mkAbstr :: C.Abstract -> IOE (CId, L.Abstr)
-    mkAbstr (C.Abstract modId flags cats funs) = return (mdi2i modId, L.Abstr {})
+    mkAbstract :: C.Abstract -> IOE (CId, L.Abstract)
+    mkAbstract (C.Abstract modId flags cats funs) = return (mdi2i modId, L.Abstract {})
 
-    mkConcr :: C.Concrete -> IOE (CId, L.Concr)
-    mkConcr (C.Concrete modId absModId flags params lincats lindefs) = do
+    mkConcrete :: C.Concrete -> IOE (CId, L.Concrete)
+    mkConcrete (C.Concrete modId absModId flags params lincats lindefs) = do
       let
         paramMap = mkParamMap params
         paramTuples = mkParamTuples params
@@ -61,9 +62,9 @@ mkCanon2lpgf opts gr am = do
                 return $ L.LFConcat v1' v2'
 
               C.LiteralValue ll -> case ll of
-                C.FloatConstant f -> return $ L.LFToken (show f)
-                C.IntConstant i -> return $ L.LFToken (show i)
-                C.StrConstant s -> return $ L.LFToken s
+                C.FloatConstant f -> return $ L.LFToken $ T.pack $ show f
+                C.IntConstant i -> return $ L.LFToken $ T.pack $ show i
+                C.StrConstant s -> return $ L.LFToken $ T.pack s
 
               C.ErrorValue err -> return $ L.LFError err
 
@@ -132,7 +133,7 @@ mkCanon2lpgf opts gr am = do
               C.PreValue pts df -> do
                 pts' <- forM pts $ \(pfxs, lv) -> do
                   lv' <- val2lin lv
-                  return (pfxs, lv')
+                  return (map T.pack pfxs, lv')
                 df' <- val2lin df
                 return $ L.LFPre pts' df'
 
@@ -168,7 +169,7 @@ mkCanon2lpgf opts gr am = do
 
       unless (null $ lefts es) (error $ unlines (lefts es))
 
-      return (mdi2i modId, L.Concr {
+      return (mdi2i modId, L.Concrete {
         L.lins = lins
       })
 
diff --git a/src/runtime/haskell/LPGF.hs b/src/runtime/haskell/LPGF.hs
index 66b3adee2..85be9d481 100644
--- a/src/runtime/haskell/LPGF.hs
+++ b/src/runtime/haskell/LPGF.hs
@@ -1,6 +1,8 @@
--- | Linearisation-only PGF format
--- Closely follows description in Section 2 of Angelov, Bringert, Ranta (2009)
--- "PGF: A Portable Run-Time Format for Type-Theoretical Grammars"
+{-# LANGUAGE OverloadedStrings #-}
+
+-- | Linearisation-only grammar format.
+-- Closely follows description in Section 2 of Angelov, Bringert, Ranta (2009):
+-- "PGF: A Portable Run-Time Format for Type-Theoretical Grammars".
 module LPGF where
 
 import PGF (Language)
@@ -9,9 +11,9 @@ import PGF.Expr (Expr)
 import PGF.Tree (Tree (..), expr2tree, prTree)
 
 import Data.Binary (Binary, get, put, encodeFile, decodeFile)
-import Data.Char (toUpper)
-import Data.List (isPrefixOf)
 import qualified Data.Map as Map
+import Data.Text (Text)
+import qualified Data.Text as T
 import Text.Printf (printf)
 
 import Prelude hiding ((!!))
@@ -20,18 +22,18 @@ import qualified Prelude
 -- | Linearisation-only PGF
 data LPGF = LPGF {
   absname   :: CId,
-  abstract  :: Abstr,
-  concretes :: Map.Map CId Concr
+  abstract  :: Abstract,
+  concretes :: Map.Map CId Concrete
 } deriving (Show)
 
--- | Abstract syntax
-data Abstr = Abstr {
+-- | Abstract syntax (currently empty)
+data Abstract = Abstract {
 } deriving (Show)
 
 -- | Concrete syntax
-data Concr = Concr {
-  -- lincats :: Map.Map CId LinType, -- ^ assigning a linearization type to each category
-  lins    :: Map.Map CId LinFun  -- ^ assigning a linearization function to each function
+data Concrete = Concrete {
+  -- lincats :: Map.Map CId LinType, -- ^ a linearization type for each category
+  lins    :: Map.Map CId LinFun  -- ^ a linearization function for each function
 } deriving (Show)
 
 -- | Abstract function type
@@ -49,15 +51,15 @@ data Concr = Concr {
 data LinFun =
   -- Additions
     LFError String -- ^ a runtime error, should probably not be supported at all
-  | LFBind
-  | LFSpace
-  | LFCapit
-  | LFAllCapit
-  | LFPre [([String], LinFun)] LinFun
+  | LFBind -- ^ join adjacent tokens
+  | LFSpace -- ^ space between adjacent tokens
+  | LFCapit -- ^ capitalise next character
+  | LFAllCapit -- ^ capitalise next word
+  | LFPre [([Text], LinFun)] LinFun
 
   -- From original definition in paper
   | LFEmpty
-  | LFToken String
+  | LFToken Text
   | LFConcat LinFun LinFun
   | LFInt Int
   | LFTuple [LinFun]
@@ -80,15 +82,15 @@ instance Binary LPGF where
       concretes = concs
     }
 
-instance Binary Abstr where
+instance Binary Abstract where
   put abs = return ()
-  get = return $ Abstr {}
+  get = return $ Abstract {}
 
-instance Binary Concr where
+instance Binary Concrete where
   put concr = put (lins concr)
   get = do
     ls <- get
-    return $ Concr {
+    return $ Concrete {
       lins = ls
     }
 
@@ -105,16 +107,24 @@ encodeFile = Data.Binary.encodeFile
 readLPGF :: FilePath -> IO LPGF
 readLPGF = Data.Binary.decodeFile
 
--- | Main linearize function
+-- | Main linearize function, to 'String'
 linearize :: LPGF -> Language -> Expr -> String
-linearize lpgf lang =
+linearize lpgf lang expr = T.unpack $ linearizeText lpgf lang expr
+
+-- | Main linearize function, to 'Data.Text.Text'
+linearizeText :: LPGF -> Language -> Expr -> Text
+linearizeText lpgf lang =
   case Map.lookup lang (concretes lpgf) of
-    Just concr -> linearizeConcr concr
+    Just concr -> linearizeConcreteText concr
     Nothing -> error $ printf "Unknown language: %s" (showCId lang)
 
--- | Language-specific linearize function
-linearizeConcr :: Concr -> Expr -> String
-linearizeConcr concr expr = lin2string $ lin (expr2tree expr)
+-- | Language-specific linearize function, to 'String'
+linearizeConcrete :: Concrete -> Expr -> String
+linearizeConcrete concr expr = T.unpack $ linearizeConcreteText concr expr
+
+-- | Language-specific linearize function, to 'Data.Text.Text'
+linearizeConcreteText :: Concrete -> Expr -> Text
+linearizeConcreteText concr expr = lin2string $ lin (expr2tree expr)
   where
     lin :: Tree -> LinFun
     lin tree = case tree of
@@ -127,7 +137,7 @@ linearizeConcr concr expr = lin2string $ lin (expr2tree expr)
 -- | Evaluation context is a sequence of terms
 type Context = [LinFun]
 
--- | Operational semantics, Table 2
+-- | Operational semantics
 eval :: Context -> LinFun -> LinFun
 eval cxt t = case t of
   LFError err -> error err
@@ -150,7 +160,7 @@ eval cxt t = case t of
   _ -> t
 
 -- | Turn concrete syntax terms into an actual string
-lin2string :: LinFun -> String
+lin2string :: LinFun -> Text
 lin2string l = case l of
   LFEmpty -> ""
   LFBind -> "" -- when encountered at beginning/end
@@ -160,15 +170,16 @@ lin2string l = case l of
   LFConcat (LFPre pts df) l2 -> lin2string $ LFConcat l1 l2
     where
       l2' = lin2string l2
-      matches = [ l | (pfxs, l) <- pts, any (`isPrefixOf` l2') pfxs ]
+      matches = [ l | (pfxs, l) <- pts, any (`T.isPrefixOf` l2') pfxs ]
       l1 = if null matches then df else head matches
-  LFConcat l1 (LFConcat LFBind l2) -> lin2string l1 ++ lin2string l2
+  LFConcat l1 (LFConcat LFBind l2) -> lin2string l1 `T.append` lin2string l2
   LFConcat l1 (LFConcat LFSpace l2) -> lin2string $ LFConcat l1 l2
-  LFConcat LFCapit l2 -> let l = lin2string l2 in toUpper (head l) : tail l
-  LFConcat LFAllCapit l2 -> let tks = words (lin2string l2) in unwords $ map toUpper (head tks) : tail tks
-  LFConcat l1 l2 -> unwords $ filter (not.null) [lin2string l1, lin2string l2]
-  x -> printf "[%s]" (show x)
+  LFConcat LFCapit l2 -> let l = lin2string l2 in T.toUpper (T.take 1 l) `T.append` T.drop 1 l
+  LFConcat LFAllCapit l2 -> let tks = T.words (lin2string l2) in T.unwords $ T.toUpper (head tks) : tail tks
+  LFConcat l1 l2 -> T.unwords $ filter (not.T.null) [lin2string l1, lin2string l2]
+  x -> T.pack $ printf "[%s]" (show x)
 
+-- | List indexing with more verbose error messages
 (!!) :: (Show a) => [a] -> Int -> a
 (!!) xs i
   | i < 0 = error $ printf "!!: index %d too small for list: %s" i (show xs)
diff --git a/testsuite/lpgf/run.hs b/testsuite/lpgf/run.hs
index b33daa211..8f602ccb6 100644
--- a/testsuite/lpgf/run.hs
+++ b/testsuite/lpgf/run.hs
@@ -53,7 +53,7 @@ doGrammar' gname cncs = do
       Just tree = readExpr ast
       -- Do some linearization
       langs =
-        [ printf "%s: %s" (showLanguage lang) (linearizeConcr concr tree)
+        [ printf "%s: %s" (showLanguage lang) (linearizeConcrete concr tree)
         | (lang,concr) <- Map.toList (concretes lpgf)
         ]
     mapM_ putStrLn langs