added a script for RTE with multiple parsers

Author: kmineshima

ja/rte_ja_mp.sh

@@ -0,0 +1,209 @@
+#!/bin/bash
+#
+#  Copyright 2016 Pascual Martinez-Gomez
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+# Script to Recognize Textual Entailment of problems in Japanese, using
+# multiple CCG parsers (Jigg and depccg at the moment).
+# This script receives a file with several sentences (one per line), where all
+# sentences are premises except the last one, which is a conclusion. It returns
+# 'yes' (the premises entail the conclusion), 'no' (there is a contradiction) or
+# 'unknown' (none of the former).
+# You can use it as:
+# 
+# ./rte_ja_mp.sh <sentences.txt> <semantic_templates.yaml>
+# 
+# E.g.
+# ./rte_ja_mp.sh ja/sample_ja.txt ja/semantic_templates_ja.yaml
+
+USAGE="Usage: ./rte_ja_mp.sh <sentences.txt> <semantic_templates.yaml>"
+
+# Create a file named "parser_location_ja.txt" at the "ja" directory and
+# write a list of CCG parsers installed, as in:
+# $ cat ja/parser_location_ja.txt
+# jigg:/path/to/ccg2lambda/ja/jigg-v-0.4
+# depccg:/path/to/depccg/build
+
+# Check that the number of arguments is correct.
+if [ "$#" -ne 2 ]; then
+  echo "Error: Number of arguments invalid".
+  echo $USAGE
+  exit 1
+fi
+
+# This variable contains the filename where the category templates are.
+category_templates=$2
+if [ ! -f $category_templates ]; then
+  echo "Error: File with semantic templates does not exist."
+  echo $USAGE
+  exit 1
+fi
+
+# This variable contains the name of the dataset (fracas or jsem).
+sentences_fname=$1
+sentences_basename=${sentences_fname##*/}
+if [ ! -f $sentences_fname ]; then
+  echo "Error: File with plain sentences does not exist."
+  echo $USAGE
+  exit 1
+fi
+
+# These variables contain the names of the directories where intermediate
+# results will be written.
+plain_dir="ja_plain" # tokenized sentences.
+parsed_dir="ja_parsed" # parsed sentences into XML or other formats.
+results_dir="ja_results" # HTML semantic outputs, proving results, etc.
+mkdir -p $plain_dir $parsed_dir $results_dir
+
+# Copy the input text to plain_dir
+cp $sentences_fname ${plain_dir}/${sentences_basename}
+
+function timeout() { perl -e 'alarm shift; exec @ARGV' "$@"; }
+
+# Set parser locations
+if [ ! -f "ja/parser_location_ja.txt" ]; then
+  echo "Error: File ja/parser_location_ja.txt does not exist."
+  exit 1
+fi
+for parser in `cat ja/parser_location_ja.txt`; do
+  parser_name=`echo $parser | awk -F':' '{print $1}'`
+  parser_dir=`echo $parser | awk -F':' '{print $2}'`
+  if [ "${parser_name}" == "jigg" ]; then
+    jigg_dir=${parser_dir}
+    if [ ! -d "${jigg_dir}/jar" ]; then
+      echo "Parser directory does not exist. Exit."
+      exit 1
+    fi
+    if [ ! -e "${parser_dir}"/jar/ccg-models-*.jar ]; then
+      echo "Japanese CCG models not found. Refer to Jigg instructions to download them."
+      exit 1
+    fi
+  fi
+  if [ "${parser_name}" == "depccg" ]; then
+    depccg_dir=${parser_dir}
+    if [ ! -d "${depccg_dir}" ] || [ ! -e "${depccg_dir}"/src/run.py ]; then
+      echo "depccg parser directory incorrect. Exit."
+      exit 1
+    fi
+  fi
+done
+
+# Set a variable with the command to invoke Jigg
+parser_cmd="java -Xmx4g -cp \"${jigg_dir}/jar/*\" jigg.pipeline.Pipeline \
+  -annotators ssplit,kuromoji,ccg \
+  -ccg.kBest 5 -file"
+
+tagging_cmd="java -Xmx4g -cp \"${jigg_dir}/jar/*\" jigg.pipeline.Pipeline \
+  -annotators ssplit,kuromoji -file"
+
+function parse_jigg() {
+  # Parse using jigg.
+  base_fname=$1
+  eval $parser_cmd ${plain_dir}/$base_fname \
+    > ${parsed_dir}/${base_fname}.log.std \
+    2> ${parsed_dir}/${base_fname}.log.err
+  mv ${plain_dir}/${base_fname}.xml ${parsed_dir}/${base_fname}.jigg.jigg.xml
+}
+
+function parse_depccg() {
+  # Parse using depccg.
+  base_fname=$1
+  eval $tagging_cmd ${plain_dir}/$base_fname \
+    > ${parsed_dir}/${base_fname}.log.std \
+    2> ${parsed_dir}/${base_fname}.log.err
+  mv ${plain_dir}/${base_fname}.xml ${parsed_dir}/${base_fname}.tagged.xml
+  env PYTHONPATH=$depccg_dir/src:$PYTHONPATH \
+    python ja/rte.py \
+    ${depccg_dir}/../models/ja_headfinal \
+    ${parsed_dir}/${base_fname}.tagged.xml \
+    > ${parsed_dir}/${base_fname}.depccg.jigg.xml
+}
+
+function semantic_parsing() {
+  parser=$1
+  sentences_basename=$2
+  python scripts/semparse.py \
+    $parsed_dir/${sentences_basename}.${parser}.jigg.xml \
+    $category_templates \
+    $parsed_dir/${sentences_basename}.${parser}.sem.xml \
+    --arbi-types \
+    2> $parsed_dir/${sentences_basename}.${parser}.sem.err
+}
+
+function proving() {
+  parser=$1
+  sentences_basename=$2
+  start_time=`python -c 'import time; print(time.time())'`
+    timeout 100 python scripts/prove.py \
+      ${parsed_dir}/${sentences_basename}.${parser}.sem.xml \
+      --graph_out ${results_dir}/${sentences_basename}.${parser}.html \
+      > ${results_dir}/${sentences_basename}.${parser}.answer \
+      2> ${results_dir}/${sentences_basename}.${parser}.err
+  rte_answer=`cat ${results_dir}/${sentences_basename}.${parser}.answer`
+  echo "judging entailment for ${parsed_dir}/${sentences_basename}.${parser}.sem.xml $rte_answer"
+  proof_end_time=`python -c 'import time; print(time.time())'`
+  proving_time=`echo "${proof_end_time} - ${start_time}" | bc -l | \
+       awk '{printf("%.2f\n",$1)}'`
+  echo $proving_time > ${results_dir}/${sentences_basename}.time
+}
+
+function select_answer() {
+  parser=$1
+  fname=${results_dir}/${sentences_basename}.${parser}.answer
+  if [ ! -e $fname ]; then
+    echo "" > $fname
+  fi
+  fname_answer=`cat ${fname}`
+  if [ "current_answer" = "no" ] && [ "$fname_answer" = "yes" ]; then
+    current_answer="unknown"
+  elif [ "current_answer" = "yes" ] && [ "$fname_answer" = "no" ]; then
+    current_answer="unknown"
+  elif [ "$fname_answer" = "yes" ]; then
+    current_answer="yes"
+    prediction_fname=`echo ${fname##*/} | sed 's/.answer//g'`
+  elif [ "$fname_answer" = "no" ]; then
+    current_answer="no"
+    prediction_fname=`echo ${fname##*/} | sed 's/.answer//g'`
+  else
+    :
+  fi
+  if [ ! -z "${prediction_fname}" ]; then
+    cp ${parsed_dir}/${prediction_fname}.jigg.xml ${parsed_dir}/${sentences_basename}.xml
+    cp ${parsed_dir}/${prediction_fname}.sem.xml ${parsed_dir}/${sentences_basename}.sem.xml
+    cp ${results_dir}/${prediction_fname}.answer ${results_dir}/${sentences_basename}.answer
+    cp ${results_dir}/${prediction_fname}.html ${results_dir}/${sentences_basename}.html
+  fi
+}
+
+# Set the current answer
+current_answer="unknown"
+prediction_fname="${sentences_basename}.jigg"
+
+# CCG parsing, semantic parsing and theorem proving
+for parser in `cat ja/parser_location_ja.txt`; do
+  parser_name=`echo $parser | awk -F':' '{print $1}'`
+  parser_dir=`echo $parser | awk -F':' '{print $2}'`
+  if [ ! -e ${parsed_dir}/${sentences_basename}.${parser_name}.jigg.xml ]; then
+    echo "${parser_name} parsing ${plain_dir}/${sentences_basename}"
+    parse_$parser_name $sentences_basename
+  fi
+  if [ ! -e ${parsed_dir}/${sentences_basename}.${parser_name}.sem.xml ]; then  
+    echo "semantic parsing $parsed_dir/${sentences_basename}.${parser_name}.sem.xml"
+    semantic_parsing $parser_name $sentences_basename 
+  fi
+  if [ ! -e ${results_dir}/${sentences_basename}.${parser_name}.answer ]; then  
+    proving $parser_name $sentences_basename 
+    select_answer ${parser_name}
+  fi
+done