From ca22f6fe3b2ee25b57f6730222636f289e6d89e2 Mon Sep 17 00:00:00 2001
From: Alice BRENON <alice.brenon@ens-lyon.fr>
Date: Thu, 19 Aug 2021 10:38:04 +0200
Subject: [PATCH] Draft a very simple WebAnno reader ignoring the fields
described in header and taking a lot from TSV and CoNLL
---
lib/Text/TEIWA.hs | 4 +++
lib/Text/TEIWA/Source.hs | 14 ++++++++-
lib/Text/TEIWA/Source/WebAnno.hs | 51 ++++++++++++++++++++++++++++++++
teiwa.cabal | 1 +
4 files changed, 69 insertions(+), 1 deletion(-)
create mode 100644 lib/Text/TEIWA/Source/WebAnno.hs
diff --git a/lib/Text/TEIWA.hs b/lib/Text/TEIWA.hs
index 3024941..d4ad823 100644
--- a/lib/Text/TEIWA.hs
+++ b/lib/Text/TEIWA.hs
@@ -8,6 +8,7 @@ module Text.TEIWA (
, annotateWith
, fromCSV
, fromCoNLLX
+ , fromWebAnno
) where
import Control.Monad.Except (MonadError(..))
@@ -34,3 +35,6 @@ fromCSV = annotate . Source csv
fromCoNLLX :: (MonadError Error m, MonadIO m) => Origin -> Filter m
fromCoNLLX = annotate . Source coNLLX
+
+fromWebAnno :: (MonadError Error m, MonadIO m) => Origin -> Filter m
+fromWebAnno = annotate . Source webAnno
diff --git a/lib/Text/TEIWA/Source.hs b/lib/Text/TEIWA/Source.hs
index 0858fcb..d6b9f66 100644
--- a/lib/Text/TEIWA/Source.hs
+++ b/lib/Text/TEIWA/Source.hs
@@ -10,8 +10,10 @@ module Text.TEIWA.Source (
, runTEIWAParser
, parse
, tsv
+ , webAnno
) where
+import Control.Applicative ((<|>))
import Control.Monad.Except (MonadError(..))
import Control.Monad.Reader (MonadReader(..), ReaderT(..))
import Control.Monad.IO.Class (MonadIO(..))
@@ -28,6 +30,7 @@ import Text.TEIWA.Source.Common (
)
import qualified Text.TEIWA.Source.CoNLLX as CoNLLX (header, sentences)
import qualified Text.TEIWA.Source.SSV as SSV (body, header)
+import qualified Text.TEIWA.Source.WebAnno as WebAnno (header, sentences)
type Format = TEIWAParser Annotation
@@ -69,7 +72,16 @@ csv :: Format
csv = ssv ','
tsv :: Format
-tsv = ssv '\t'
+tsv = webAnno <|> ssv '\t'
+
+webAnno :: Format
+webAnno = do
+ context <- annotationContext =<< WebAnno.header
+ SentenceLevel <$> (
+ WebAnno.sentences >>= mapM (
+ fmap SentenceAnnotation . mapM (annotateToken context)
+ )
+ )
data Origin = File FilePath | Text Text
diff --git a/lib/Text/TEIWA/Source/WebAnno.hs b/lib/Text/TEIWA/Source/WebAnno.hs
new file mode 100644
index 0000000..a20a2a9
--- /dev/null
+++ b/lib/Text/TEIWA/Source/WebAnno.hs
@@ -0,0 +1,51 @@
+{-# LANGUAGE FlexibleContexts #-}
+{-# LANGUAGE OverloadedStrings #-}
+module Text.TEIWA.Source.WebAnno (
+ header
+ , sentences
+ ) where
+
+import Control.Applicative (many)
+import Data.Text.Lazy as Text (pack)
+import Text.Parsec (
+ ParsecT, Stream, char, digit, many1, noneOf, sepBy1, skipMany1, string
+ )
+import Text.TEIWA.Source.Common (
+ Field, Header, Row, TEIWAParser, eol, recordLine
+ )
+
+header :: TEIWAParser Header
+header = do
+ string "#FORMAT=WebAnno TSV " *> version
+ many comment *> pure ()
+ pure [
+ "ID"
+ , "SPAN"
+ , "FORM"
+ , "LABEL"
+ ]
+ where
+ version = skipMany1 digit `sepBy1` char '.' *> eol
+ comment = char '#' *> many (noneOf "\r\n") <* eol
+
+field :: Stream s m Char => ParsecT s u m Field
+field = build <$> many1 (noneOf "\t\n\r")
+ where
+ build "_" = Nothing
+ build s = Just . Text.pack $ unescape s
+ unescape "" = ""
+ unescape ('\\':c:s) = c:unescape s
+ unescape (c:s) = c:unescape s
+
+row :: Stream s m Char => ParsecT s u m Row
+row = recordLine (many1 (field <* char '\t') <* eol)
+
+type Sentence = [Row]
+
+sentence :: Stream s m Char => ParsecT s u m Sentence
+sentence = many comment *> many1 row
+ where
+ comment = char '#' *> many (noneOf "\r\n") <* eol
+
+sentences :: Stream s m Char => ParsecT s u m [Sentence]
+sentences = many (many1 eol *> sentence)
diff --git a/teiwa.cabal b/teiwa.cabal
index 9760cff..bfd3856 100644
--- a/teiwa.cabal
+++ b/teiwa.cabal
@@ -31,6 +31,7 @@ library
, Text.TEIWA.Source.Common
, Text.TEIWA.Source.CoNLLX
, Text.TEIWA.Source.SSV
+ , Text.TEIWA.Source.WebAnno
build-depends: base >=4.12 && <4.15
, bytestring
, mtl
--
GitLab