mcnguyen / nlpproject Goto Github PK

The problem is to extract information about events mentioned in sentences. Input includes articles describing events and a schema for that domain such as bombing domain. Output is the information about the events in a tabular form of triples where each tuple denotes the relation between two actors and a second table for each actor and all its mentions. Event extraction is a key prerequisite for generating structured, machine-readable representations of natural language. Such representations can aid various tasks like question answering, machine translation, novelty detection.

1. Jars

• json-simple-1.1.1
• stanford-corenlp-3.5.1-models
• stanford-corenlp-3.5.1

2. relgramtuples-app
3. textual-entailment running as service

• go to textual-entailment directory and run following from terminal
  • sbt
  • service/reStart

• Run EventExtraction.java

1. Copy required schema lines in manual_schema_original.txt
2. Steps to run relgramtuples-app on news articles and store output in document_relgram_original.txt

• save news article in news_articles.txt
• run ParseDocumentForRelgramsApp.java, it will output news_relgram_input.txt
• copy news_relgram_input.txt in relgramtuples-app directory
• go to relgramtuples-app directory from terminal and run below command
  • java -cp lib/relgramtuples-1.0.0-SNAPSHOT-jar-with-dependencies.jar edu.washington.cs.knowitall.relgrams.apps.RelgramTuplesApp --wnHome resources/wordnet3.0 --wnTypesFile resources/wordnet-selected-classes.txt --ne7ModelFile resources/english.muc.7class.nodistsim.crf.ser.gz --ne3ModelFile resources/english.all.3class.nodistsim.crf.ser.gz -mpp resources/engmalt.linear-1.7.mco news_relgram_input.txt document_relgram_original.txt
• copy generated document_relgram_original.txt in workspace

3. Run following parsers to parse above files

• ParseDocument.java
• ParseSchema.java

4. Run TextEntail.java to see desired output

• results.txt file
• Schema_tuple - confidence - Document_tuple