From bb98f03df59dd8942e23bbb43a4cdcb5128eb53a Mon Sep 17 00:00:00 2001
From: Joerg Tiedemann <jorg.tiedemann@helsinki.fi>
Date: Wed, 22 Jan 2020 13:33:28 +0200
Subject: [PATCH] backtranslate bugfix

---
 TODO.md                | 14 ++++++++++++++
 backtranslate/Makefile | 19 ++++++++++++++++---
 2 files changed, 30 insertions(+), 3 deletions(-)

diff --git a/TODO.md b/TODO.md
index 4f18d8ab..814aec25 100644
--- a/TODO.md
+++ b/TODO.md
@@ -1,8 +1,22 @@
 
 # Things to do
 
+
+## Backtranslation
+
+* status: basically working, need better integration?!
 * add backtranslations to training data
 * can use monolingual data from tokenized wikipedia dumps: https://sites.google.com/site/rmyeid/projects/polyglot
 * https://dumps.wikimedia.org/backup-index.html
 * better in JSON: https://dumps.wikimedia.org/other/cirrussearch/current/
 
+## Fine-tuning and domain adaptation
+
+* status: basically working
+* do we want to publishfine-tuned data or rather the fina-tuning procedures? (using a docker container?)
+
+
+## Show-case some selected language pairs
+
+* collaboration with wikimedia
+* focus languages: Tagalog (tl, tgl), Central Bikol (bcl), Malayalam (ml, mal), Bengali (bn, ben), and Mongolian (mn, mon)
diff --git a/backtranslate/Makefile b/backtranslate/Makefile
index 6653551f..29d9a84e 100644
--- a/backtranslate/Makefile
+++ b/backtranslate/Makefile
@@ -28,6 +28,12 @@ MODELHOME = ../models/${LANGPAIR}
 MODELZIP  = ${lastword ${sort ${wildcard ${MODELHOME}/*-20*.zip}}}
 MODELNAME = ${patsubst %.zip,%,${notdir ${MODELZIP}}}
 
+ifeq (${MODELNAME},)
+  MODELHOME = ../work-spm/models/${LANGPAIR}
+  MODELZIP  = ${lastword ${sort ${wildcard ${MODELHOME}/*-20*.zip}}}
+  MODELNAME = ${patsubst %.zip,%,${notdir ${MODELZIP}}}
+endif
+
 
 LOAD_MODULES = 	module use -a /projappl/nlpl/software/modules/etc/ && \
 		module load nlpl-udpipe nlpl-opus &&
@@ -46,7 +52,6 @@ WIKI_TRG = ${LANGPAIR}/${WIKISOURCE}.${PART}_${MODELNAME}.${LANGPAIR}.${TRG}.gz
 ## don't delete translated text if the process crashes
 .PRECIOUS: ${WIKI_TRG}
 
-
 ## find wiki downloads
 WIKI_JSON = $(shell grep -o '${LANGID}${WIKISOURCE}-[0-9]*-cirrussearch-content.json.gz' index.html | head -1)
 
@@ -68,12 +73,14 @@ all: index.html
 	${MAKE} ${WIKI_SRC} ${WIKI_TRG}
 
 
-WIKISOURCES = wiki wikibooks wikinews wikiquote wikisource
+WIKISOURCES = wiki wikibooks wikinews wikiquote wikisource wiktionary
 
 all-wikis:
 	for w in ${WIKISOURCES}; do \
 	  ${MAKE} WIKISOURCE=$$w prepare-data; \
-	  ${MAKE} WIKISOURCE=$$w HPC_CORES=1 WALLTIME=72 translate.submit; \
+	  if  [ `find ${WIKI_DIR} -name '$$w.${LANGID}.${PART}.gz' | wc -l` -gt 0 ]; then \
+	    ${MAKE} WIKISOURCE=$$w HPC_CORES=1 WALLTIME=72 translate.submit; \
+	  fi \
 	done
 
 
@@ -83,6 +90,12 @@ all-wikilangs: index.html
 	done
 
 
+# Tagalog (tl, tgl), Central Bikol (bcl), Malayalam (ml, mal), Bengali (bn, ben), and Mongolian (mn, mon)
+focus-wikis:
+	for l in tl bcl ml bn mn; do \
+	  ${MAKE} SRC=$$l TRG=en all-wikis; \
+	done
+
 extract-text: ${WIKI_TXT}
 prepare-model: ${LANGPAIR}/decoder.yml
 prepare-data: ${WIKI_PRE}