-
Notifications
You must be signed in to change notification settings - Fork 79
/
Makefile
215 lines (193 loc) · 9.53 KB
/
Makefile
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
LANGCODE=it
LANGUAGE=italian
WORK_DIR=./workspace-$(LANGCODE)
PAGES_DIR=$(WORK_DIR)/pages
SOCCER_DIR=$(WORK_DIR)/soccer
SENTENCES_DIR=$(WORK_DIR)/sentences
POS_TAGGED_DIR=$(WORK_DIR)/tagged
SUPERVISED_DIR=$(WORK_DIR)/supervised
UNSUPERVISED_DIR=$(WORK_DIR)/unsupervised
SAMPLE_DIR=$(WORK_DIR)/samples
TREETAGGER_HOME=../treetagger
TREETAGGER=$(TREETAGGER_HOME)/cmd/tree-tagger-$(LANGUAGE)
SOCCER_IDS=extraction/resources/soccer_ids
SAMPLE_LU_NUM=10
SENTENCES_MIN_WORDS=5
SENTENCES_MAX_WORDS=25
SPLITTER_MIN_CHARS=25
DUMP=../$(LANGCODE)wiki-latest-pages-articles.xml.bz2
CL_MAIN_PACKAGE=org.fbk.cit.hlt.dirha
CL_JAVA_OPTS=-Dlog-config=supervised/classifier/log-config.properties -Xmx2G -Dfile.encoding=UTF-8 -cp supervised/classifier/target/fatJar.jar
CL_TRAINING_SET=supervised/resources/training.sample
CL_GAZETTEER=supervised/resources/it/dbpedia-gaz.tsv
CL_SENTENCES_FILE=$(SUPERVISED_DIR)/all_sentences
CL_OUTPUT=$(CL_SENTENCES_FILE).classified
CL_ANNOTATED_GOLD=resources/gold-standard.final
CL_SVM_TRAIN_ARGS=-b 1 -t 0 -m 6000 -s 0
LINK_MODE=twm # twm or nex
MIN_LINK_CONFIDENCE=0.01
CF_RESULTS=resources/crowdflower-results.sample
SCORING_TYPE=weighted-mean
FE_SCORE=both
SCORING_CORE_WEIGHT=2
default:
@echo "Nothing to do. Please launch me with a target!"
extract-pages:
mkdir -p $(WORK_DIR)
bzcat $(DUMP) | python lib/WikiExtractor.py -o $(PAGES_DIR)
find $(PAGES_DIR)/*/* -type f -exec cat '{}' \; > $(WORK_DIR)/all-articles.txt
extract-soccer:
# must run extract-pages before
mkdir -p $(SOCCER_DIR)
python extraction/get_soccer_ids.py $(SOCCER_IDS) --lang $(LANGCODE)
python extraction/process_articles.py $(SOCCER_IDS) \
$(WORK_DIR)/all-articles.txt $(SOCCER_DIR) \
$(WORK_DIR)/wiki-id-to-title-mapping.json
find $(SOCCER_DIR)/* -type f -exec cat '{}' \; > $(WORK_DIR)/all-soccer.txt
extract-verbs:
# must run extract-pages before
cat $(WORK_DIR)/all-articles.txt | $(TREETAGGER) | grep VER | grep -v '<unknown>' | \
cut -f 1,3 | tr '[:upper:]' '[:lower:]' > $(WORK_DIR)/raw-verbs.txt
sort -u $(WORK_DIR)/raw-verbs.txt -o $(WORK_DIR)/token2lemma.txt
cut -f 1 $(WORK_DIR)/token2lemma.txt > $(WORK_DIR)/tokens.txt
cut -f 2 $(WORK_DIR)/token2lemma.txt | sort -u -o $(WORK_DIR)/lemmas.txt
python extraction/bag_of_words.py $(WORK_DIR)/all-articles.txt \
$(WORK_DIR)/vocabulary.txt
pos-tag:
mkdir -p $(POS_TAGGED_DIR)
find $(SENTENCES_DIR) -type f | xargs -I{} -P 4 sh -c 'echo {} && \
$(TREETAGGER) {} > $(POS_TAGGED_DIR)/$$(basename {}) 2> /dev/null';
extract-sentences-baseline:
# must run extract-soccer before
mkdir -p $(SENTENCES_DIR)
python extraction/extract_sentences.py $(WORK_DIR)/all-soccer.txt \
resources/tokens.list $(WORK_DIR)/sentence-to-wikiid.json $(SENTENCES_DIR) \
--min-words $(SENTENCES_MIN_WORDS) --max-words $(SENTENCES_MAX_WORDS)
extract-sentences-synctactic:
# must run extract-sentences-baseline, pos-tag and rank-verbs before
mkdir -p $(SENTENCES_DIR)
python seed_selection/get_meaningful_sentences.py $(POS_TAGGED_DIR) \
$(WORK_DIR)/top-50-tokens.txt $(SENTENCES_DIR)
extract-sentences-lexical:
# must run extract-short-sentences and rank-verbs before
mkdir -p $(SENTENCES_DIR)
pyhon seed_selection/generate_lexicalizazion_patterns.py \
$(WORK_DIR)/top-50-tokens.txt SoccerPlayer SoccerClub SoccerLeague \
SoccerTournament SoccerLeagueSeason $(WORK_DIR)/lexicalizations
sort -ou $(WORK_DIR)/lexicalizations
find $(SHORT_DIR) -type f -exec sh -c 'echo $$1 && grep -irhwf \
$(WORK_DIR)/lexicalizations $$1 | ifne "cat > $(SENTENCES_DIR)/$$(basename $$1)"' \
'extract-sentences-lexical' '{}' \;
extract-sentences-splitter:
# must run extract-soccer before
mkdir -p $(SENTENCES_DIR)
python seed_selection/split_sentences.py resources/italian-splitter.pickle \
$(SOCCER_DIR) $(SENTENCES_DIR) --min-length $(SPLITTER_MIN_CHARS)
rank-verbs:
# must run extract-verbs before
cut -f 2 $(WORK_DIR)/token2lemma.txt | python verb_ranking/make_lemma_freq.py - \
$(WORK_DIR)/lemma-freq.json
cut -d '"' -f 2 workspace-it/lemma-freq.json | grep -vP \
'avere|essere|fare|potere|volere|dovere' | sed -n '2,52'p | sort \
> $(WORK_DIR)/top-50-lemmas.txt
grep -wf $(WORK_DIR)/top-50-lemmas.txt $(WORK_DIR)/token2lemma.txt | sort -u | \
cut -f 1 > $(WORK_DIR)/top-50-tokens.txt
grep -wf $(WORK_DIR)/top-50-lemmas.txt $(WORK_DIR)/token2lemma.txt \
> $(WORK_DIR)/top-50-token2lemma.txt
python verb_ranking/tf_idfize.py $(SOCCER_DIR) $(WORK_DIR)/top-50-tokens.txt \
--variance-out $(WORK_DIR)/verbs-variance.json \
--stdevs-out $(WORK_DIR)/verbs-stdev.json \
--threshold-rank-out $(WORK_DIR)/verbs-rank-threshold.json \
--tfidf-rank-out $(WORK_DIR)/verbs-tfidf-rank.json
python verb_ranking/compute_stdev_by_lemma.py $(WORK_DIR)/top-50-token2lemma.txt \
$(WORK_DIR)/verbs-stdev.json $(WORK_DIR)/lemma-stdev.json
supervised-build-classifier:
cd supervised/classifier && mvn compile assembly:assembly
supervised-learn-roles:
java $(CL_JAVA_OPTS) $(CL_MAIN_PACKAGE).RoleTrainingSetToLibSvm \
-t $(CL_TRAINING_SET) -g $(CL_GAZETTEER)
svm-train $(CL_SVM_TRAIN_ARGS) $(CL_TRAINING_SET).iob2.svm \
$(CL_TRAINING_SET).iob2.model
supervised-learn-frames:
java $(CL_JAVA_OPTS) $(CL_MAIN_PACKAGE).FrameTrainingSetToLibSvm \
-t $(CL_TRAINING_SET) -g $(CL_GAZETTEER)
svm-train $(CL_SVM_TRAIN_ARGS) $(CL_TRAINING_SET).frame.svm \
$(CL_TRAINING_SET).frame.model
supervised-run-interactive:
java $(CL_JAVA_OPTS) -Dtreetagger.home=$(TREETAGGER_HOME) \
$(CL_MAIN_PACKAGE).Annotator -g $(CL_GAZETTEER) -m $(CL_TRAINING_SET) \
-l $(LANGCODE) -i -n
supervised-run-batch:
mkdir -p $(SUPERVISED_DIR)
for s in $$(find $(SENTENCES_DIR) -type f); do echo $$(basename $$s) >> $(CL_SENTENCES_FILE); \
cat $$s >> $(CL_SENTENCES_FILE); echo >> $(CL_SENTENCES_FILE); done
python date_normalizer/rpc.py 2>/dev/null &
java $(CL_JAVA_OPTS) -Dtreetagger.home=$(TREETAGGER_HOME) \
$(CL_MAIN_PACKAGE).Annotator -g $(CL_GAZETTEER) -m $(CL_TRAINING_SET) \
-l $(LANGCODE) -r $(CL_SENTENCES_FILE).eval -a $(CL_SENTENCES_FILE) \
-o $(CL_OUTPUT) -n
supervised-evaluate:
python date_normalizer/rpc.py 2>/dev/null &
java $(CL_JAVA_OPTS) -Dtreetagger.home=$(TREETAGGER_HOME) \
$(CL_MAIN_PACKAGE).Annotator -g $(CL_GAZETTEER) -m $(CL_TRAINING_SET) \
-l $(LANGCODE) -r $(CL_ANNOTATED_GOLD).report \
-a $(CL_ANNOTATED_GOLD).sentences -e $(CL_ANNOTATED_GOLD) \
-r $(CL_ANNOTATED_GOLD).report -o $(CL_ANNOTATED_GOLD).classified -n
supervised-results-to-assertions:
python supervised/produce_triples.py $(CL_OUTPUT) \
$(WORK_DIR)/wiki-id-to-title-mapping.json $(CL_OUTPUT).triples.nt \
$(CL_OUTPUT).triples.scores.nt --format nt --sentence-score $(SCORING_TYPE) \
--core-weight $(SCORING_CORE_WEIGHT) --fe-score $(FE_SCORE) --format nt
supervised-plot-results:
python supervised/plot.py $(CL_ANNOTATED_GOLD).report.frames.confusion.csv \
precall_bars &
python supervised/plot.py $(CL_ANNOTATED_GOLD).report.frames.confusion.csv \
precall_scatter &
python supervised/plot.py $(CL_ANNOTATED_GOLD).report.frames.confusion.csv confmat &
python supervised/plot.py $(CL_ANNOTATED_GOLD).report.roles.confusion.csv \
precall_bars &
python supervised/plot.py $(CL_ANNOTATED_GOLD).report.roles.confusion.csv \
precall_scatter &
python supervised/plot.py $(CL_ANNOTATED_GOLD).report.roles.confusion.csv confmat &
crowdflower-create-input:
# TODO generate twm ngrams and textpro chunks somehow
mkdir -p $(WORK_DIR)/linked
python lib/entity_linking.py $(LINK_MODE) $(SAMPLE_DIR)/sentences $(WORK_DIR)/linked
python unsupervised/produce_labeled_data.py $(WORK_DIR)/linked \
$(WORK_DIR)/labeled_data.json
# TODO change combine_chunks so that you can specify where to get individual parts
mv $(WORK_DIR)/linked resources/twm-links
python crowdflower/combine_chunks.py resources/ $(WORK_DIR)/chunks.json
python crowdflower/create_crowdflower_input.py resources/labeled_data.sample \
$(WORK_DIR)/chunks.json -o $(WORK_DIR)/crowdflower_input.csv
python crowdflower/generate_crowdflower_interface_template.py \
$(WORK_DIR)/crowdflower_input.csv $(WORK_DIR)/crowdflower_template.html
crowdflower-to-training:
python crowdflower/crowdflower_results_into_training_data.py $(CF_RESULTS) \
$(TAGGED_DIR) $(WORK_DIR)/training-data.tsv -d
unsupervised-run:
# you need to run extract-sentences-* before
mkdir -p $(WORK_DIR)/unsupervised/linked
python lib/entity_linking.py --debug -d -c $(MIN_LINK_CONFIDENCE) $(LINK_MODE) \
$(SENTENCES_DIR) $(WORK_DIR)/unsupervised/linked
python unsupervised/produce_labeled_data.py $(WORK_DIR)/unsupervised/linked \
$(WORK_DIR)/labeled_data.json --score $(SCORING_TYPE) \
--core-weight $(SCORING_CORE_WEIGHT) --score-fes
python unsupervised/labeled_to_assertions.py $(WORK_DIR)/labeled_data.json \
$(WORK_DIR)/wiki-id-to-title-mapping.json $(WORK_DIR)/unsupervised/processed \
$(WORK_DIR)/unsupervised/discarded $(WORK_DIR)/unsupervised/dataset.nt
unsupervised-evaluate:
mkdir -p $(WORK_DIR)/unsupervised/evaluation/{linked,sentences}
i=0; while read f; do \
printf -v j "%04d" $$i; \
echo $$f > $(WORK_DIR)/unsupervised/evaluation/sentences/$$j; \
let i++; \
done < $(GOLD_SENTENCES)
python lib/entity_linking.py -d -c $(MIN_LINK_CONFIDENCE) $(LINK_MODE) \
$(WORK_DIR)/unsupervised/evaluation/sentences \
$(WORK_DIR)/unsupervised/evaluation/linked
python unsupervised/produce_labeled_data.py $(WORK_DIR)/unsupervised/evaluation/linked \
$(WORK_DIR)/unsupervised/evaluation/test_data.json --score $(SCORING_TYPE) \
--core-weight $(SCORING_CORE_WEIGHT) --score-fes
python unsupervised/evaluate.py $(WORK_DIR)/unsupervised/evaluation/test_data.json \
$(GOLD_STANDARD) --partial