ClaimLinker

ClaimLinker is a Web service and API that links arbitrary text to fact-checked claims, offering a novel kind of semantic annotation of unstructured content. The system is based on a scalable, fully unsupervised and modular approach that does not require training or tuning and which can serve high quality results at real time.

More information is available at the following paper:

Maliaroudakis E., Boland K., Dietze S., Todorov K., Tzitzikas Y. and Fafalios P., 
"ClaimLinker: Linking Text to a Knowledge Graph of Fact-checked Claims", 
In Companion Proceedings of the Web Conference 2021.

A sub-module system has been implemented, divided respectively by the following modules:

Claimlinker_commons (main library for claim linking)

ClaimLinker
Core NLP API
Association Type
Similarity Measures
Test class

Claimlinker_web (web services)

Web Servlet that returns annotations in JSON
Web Servlet for Bookmarklet that allows a user to select a piece of text in a web page and check if there are fact-checked claims linked to the selected text.
JSP offering a form where the user can give sometext and check if there are fact-checked claims linked to that text.

ElasticSearch_Tools (indexing of claims and search service provision)

Initializer for an ElasticSearch server
Wrapper API for an ElasticSearch server
OpenCSV wrapper class

Getting Started

Installation

We need to install the FEL library to the ClaimLinker_commons pom.

cd ClaimLinker_commons
mvn install:install-file -Dfile=./lib/FEL-0.1.0-fat.jar -DgroupId=com.yahoo.semsearch -DartifactId=FEL -Dversion=0.1.0 -Dpackaging=jar -DgeneratePom=true

Then compile the whole project from the project's root directory:

cd ..
mvn compile package #to compile and package into jar the claimlinker

Get the claim data from a csv, the hash file for FEL library the stopwords file and the punctuations file: i.e.
- data/
  - claim_extraction_18_10_2019_annotated.csv (a dataset of fact-checked claims)
  - english-20200420.hash (used by FEL)
  - stopwords.txt
  - puncs.txt

ElasticSearch setup

You can check here how you can set up a single-node elasticsearch using docker.

Usage

Initializing ElasticSearch:

java ElasticInitializer -f "data.csv" -h "elasticsearch_host"

Running the ClaimlinkerTest class:

java -cp .:ClaimLinker_commons/target/ClaimLinker_commons-1.0-jar-with-dependencies.jar:ClaimLinker_web/target/ClaimLinker_web-1.0.jar:ElasticSearch_Tools/target/ElasticSearch_Tools-1.0.jar: csd.claimlinker.ClaimLinkerTest

Using it as a library:

ClaimLinker CLInstance = new ClaimLinker(elastic_search_threashold, similarityMeasures, stopwords_file, punctuations_file english_hash_FEL, ElasticSearch_host);

CLInstance.claimLink(text, num_of_returned_claims, associationtype, cleanPrevAnnotations)

i.e. (csd.claimlinker.ClaimLinkerTest)

public static void main(String[] args) throws Exception{
        demo_pipeline("Of course, we are 5 percent of the world's population;\n");
}

static Set<CLAnnotation> demo_pipeline(String text) throws IOException, ClassNotFoundException {
    AnalyzerDispatcher.SimilarityMeasure[] similarityMeasures = new AnalyzerDispatcher.SimilarityMeasure[]{
		AnalyzerDispatcher.SimilarityMeasure.jcrd_comm_words,           //Common (jaccard) words
		AnalyzerDispatcher.SimilarityMeasure.jcrd_comm_lemm_words,      //Common (jaccard) lemmatized words
		AnalyzerDispatcher.SimilarityMeasure.jcrd_comm_ne,              //Common (jaccard) named entities
		AnalyzerDispatcher.SimilarityMeasure.jcrd_comm_dissambig_ents,  //Common (jaccard) disambiguated entities BFY
		AnalyzerDispatcher.SimilarityMeasure.jcrd_comm_pos_words,       //Common (jaccard) words of specific POS
		AnalyzerDispatcher.SimilarityMeasure.jcrd_comm_ngram,           //Common (jaccard) ngrams
		AnalyzerDispatcher.SimilarityMeasure.jcrd_comm_nchargram,       //Common (jaccard) nchargrams
		AnalyzerDispatcher.SimilarityMeasure.vec_cosine_sim             //Cosine similarity
	};
	ClaimLinker CLInstance = new ClaimLinker(20, similarityMeasures, "data/stopwords.txt", "data/puncs.txt", "data/english-20200420.hash", "192.168.2.112");
	System.out.println("Demo pipeline started!");
	Set<CLAnnotation> results = CLInstance.claimLink(text, 5, Association_type.all, true);
	return results;
}

Using it as a Web Service returning results in JSON:

Example request for text "You know, interest on debt will soon exceed security spending.":

http://<claimlinker-url>/claimlinker?app=service&text=You%20know,%20interest%20on%20debt%20will%20soon%20exceed%20security%20spending.

Example of results in JSON: 

{"_results":[{
   "text":"You know, interest on debt will soon exceed security spending.",
   "sentencePosition":0,
   "association_type":"same_as",
   "linkedClaims":[
       { "claimReview_claimReviewed":"'Within a few years, we will be spending more on interest payments than on national security.'",
	 "_score":39.261948,
	 "extra_title":"Mitch Daniels says interest on debt will soon exceed security spending",
	 "rating_alternateName":"false",
	 "creativeWork_author_name":"Mitch Daniels",
	 "claimReview_url":"http://www.politifact.com/truth-o-meter/statements/2011/feb/17/mitch-daniels/mitch-daniels-says-interest-debt-will-soon-exceed-/",
	 "claim_uri":"http://data.gesis.org/claimskg/creative_work/16076cfe-34c2-542b-9ffa-41a4c8677ced"},
       { "claimReview_claimReviewed":"'In just 17 years, spending for Social Security, federal health care and interest on the debt will exceed ALL tax revenue!'",
         "_score":28.52819,
	 "extra_title":"Brat says entitlement and debt payments will consume all taxes in 2032",
	 "rating_alternateName":"mostly true",
	 "creativeWork_author_name":"Dave Brat",
	 "claimReview_url":"http://www.politifact.com/virginia/statements/2015/jun/16/dave-brat/brat-says-entitlement-and-debt-payments-will-consu/",
	 "claim_uri":"http://data.gesis.org/claimskg/creative_work/928f16c7-1ae8-53bb-a6d2-8aa6a59dbd45"},
       { "claimReview_claimReviewed":"'By 2022, just the interest payment on our debt will be greater than the defense of our country.'",
         "_score":28.45221,
         "extra_title":"Will interest on the debt exceed defense spending by 2022?",
         "rating_alternateName":"mostly true",
         "creativeWork_author_name":"Joe Manchin",
         "claimReview_url":"http://www.politifact.com/truth-o-meter/statements/2018/may/10/joe-manchin/will-interest-debt-exceed-defense-spending-2022/",
         "claim_uri":"http://data.gesis.org/claimskg/creative_work/3579711f-d846-57ba-998a-6c0983c2c2cb"},
       { "claimReview_claimReviewed":"'The debt will soon eclipse our entire economy.'",
         "_score":18.63715,
	 "extra_title":"Paul Ryan, in State of the Union response, says U.S. debt will soon eclipse GDP",
	 "rating_alternateName":"true",
	 "creativeWork_author_name":"Paul Ryan",
	 "claimReview_url":"http://www.politifact.com/truth-o-meter/statements/2011/jan/26/paul-ryan/paul-ryan-state-union-response-says-us-debt-will-s/",
	 "claim_uri":"http://data.gesis.org/claimskg/creative_work/7951dee8-b793-512c-bd13-074f648a052a"},
       { "claimReview_claimReviewed":"'Our spending has caught up with us, and our debt soon will eclipse the entire size of our national economy.'",
         "_score":17.90907,
	 "extra_title":"House Speaker John Boehner has the right count on the magnitude of the federal debt",
	 "rating_alternateName":"true",
	 "creativeWork_author_name":"John Boehner",
	 "claimReview_url":"http://www.politifact.com/ohio/statements/2011/jan/10/john-boehner/house-speaker-john-boehner-has-right-count-magnitu/",
	 "claim_uri":"http://data.gesis.org/claimskg/creative_work/4c906dfc-c6af-5d38-aec4-0808edb2dcb9"}]}],"timeElapsed":1117}

malvag / claimlinker Goto Github PK

claimlinker's Introduction

ClaimLinker

Claimlinker_commons (main library for claim linking)

Claimlinker_web (web services)

ElasticSearch_Tools (indexing of claims and search service provision)

Getting Started

Installation

ElasticSearch setup

Usage

claimlinker's People

Contributors

Stargazers

Watchers

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent