| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 1 | SRC_DIR ?= test/resources/DNB |
| 2 | BUILD_DIR = build |
| 3 | TARGET_DIR ?= target |
| 4 | |
| Marc Kupietz | 3d82f56 | 2024-03-16 10:19:03 +0100 | [diff] [blame] | 5 | .PHONY: all clean test krill |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 6 | |
| Marc Kupietz | fc32862 | 2024-03-16 14:31:34 +0100 | [diff] [blame] | 7 | .PRECIOUS: %.zip %.tree_tagger.zip %.ud.zip %.spacy.zip %.i5.xml %.tar |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 8 | |
| 9 | all: $(TARGET_DIR)/dnb.i5.xml |
| 10 | |
| Marc Kupietz | 3d82f56 | 2024-03-16 10:19:03 +0100 | [diff] [blame] | 11 | krill: $(TARGET_DIR)/dnb.krill.tar |
| 12 | |
| Marc Kupietz | fa445da | 2024-03-16 10:46:22 +0100 | [diff] [blame] | 13 | KORAPXML2CONLLU ?= java -jar lib/korapxml2conllu.jar |
| 14 | |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 15 | $(TARGET_DIR)/dnb.i5.xml: $(patsubst $(SRC_DIR)/%.epub,$(TARGET_DIR)/%.i5.xml,$(wildcard $(SRC_DIR)/*.epub)) |
| 16 | head -n -1 xslt/idsCorpus-template.xml > $@ |
| 17 | cat $^ >> $@ |
| 18 | tail -n 1 xslt/idsCorpus-template.xml >> $@ |
| 19 | |
| Marc Kupietz | dfbc3f4 | 2024-03-16 11:46:41 +0100 | [diff] [blame] | 20 | test: $(TARGET_DIR)/dnb.i5.xml |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 21 | xmllint --noout --valid $< |
| 22 | |
| 23 | $(BUILD_DIR)/%: $(SRC_DIR)/%.epub |
| 24 | mkdir -p $@ |
| 25 | echo "Converting $< to $@" |
| 26 | unzip -q -o $< -d $@ |
| 27 | |
| Marc Kupietz | dfbc3f4 | 2024-03-16 11:46:41 +0100 | [diff] [blame] | 28 | $(TARGET_DIR)/%.i5.xml: $(BUILD_DIR)/% xslt/epub2i5.xsl xslt/idsCorpus-template.xml |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 29 | mkdir -p $(TARGET_DIR) |
| 30 | echo "Converting $< to $@" |
| 31 | java -jar lib/saxon9ee.jar -xsl:xslt/epub2i5.xsl $</*/content.opf > $@ |
| 32 | |
| 33 | %.zip: %.i5.xml |
| 34 | tei2korapxml -l warn -s -tk - < $< > $@ |
| 35 | |
| 36 | %.tree_tagger.zip: %.zip |
| Marc Kupietz | fa445da | 2024-03-16 10:46:22 +0100 | [diff] [blame] | 37 | $(KORAPXML2CONLLU) $< | pv | docker run --rm -i korap/conllu2treetagger -l german | conllu2korapxml > $@ |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 38 | |
| 39 | %.spacy.zip: %.zip |
| Marc Kupietz | fa445da | 2024-03-16 10:46:22 +0100 | [diff] [blame] | 40 | $(KORAPXML2CONLLU) $< | pv | docker run --rm -i korap/conllu2spacy | conllu2korapxml > $@ |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 41 | |
| 42 | %.ud.zip: %.zip |
| Marc Kupietz | fa445da | 2024-03-16 10:46:22 +0100 | [diff] [blame] | 43 | $(KORAPXML2CONLLU) $< | pv | ./scripts/udpipe2 | conllu2korapxml > $@ |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 44 | |
| Marc Kupietz | 22ad4b7 | 2024-03-17 16:24:21 +0100 | [diff] [blame^] | 45 | %.krill.tar: %.zip %.ud.zip %.tree_tagger.zip %.spacy.zip |
| Marc Kupietz | 3d82f56 | 2024-03-16 10:19:03 +0100 | [diff] [blame] | 46 | mkdir -p $(basename $@) |
| Marc Kupietz | fc32862 | 2024-03-16 14:31:34 +0100 | [diff] [blame] | 47 | korapxml2krill archive --quiet -w -z -cfg krill-korap4dnb.cfg --non-word-tokens --meta I5 -i $< -i $(word 2,$^) -i $(word 3,$^) -o $(basename $@) |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 48 | |
| Marc Kupietz | fc32862 | 2024-03-16 14:31:34 +0100 | [diff] [blame] | 49 | %.json: %.krill.tar |
| 50 | rm -rf $@ |
| 51 | mkdir -p $@ |
| 52 | for f in $<; do tar -C $@ -xf $$f; done |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 53 | |
| Marc Kupietz | 22ad4b7 | 2024-03-17 16:24:21 +0100 | [diff] [blame^] | 54 | %.index: %.json |
| 55 | rm -rf $@ |
| 56 | mkdir -p $@ |
| 57 | chmod a+rwX $@ |
| 58 | docker run --rm -v ./$(TARGET_DIR):/data:z korap/kustvakt:latest-full Krill-Indexer.jar -c /kustvakt/kustvakt.conf -i /data/dnb.json -o /data/dnb.index/ |
| 59 | chmod o-w $@ |
| 60 | |
| 61 | %.index.tar.xz: %.index |
| 62 | tar -I 'xz -T0' -C $(dir $<) -cf $@ $(notdir $<) |
| 63 | |
| Marc Kupietz | 1a42266 | 2024-03-16 09:34:10 +0100 | [diff] [blame] | 64 | clean: |
| 65 | rm -rf $(BUILD_DIR) $(TARGET_DIR) |
| 66 | |