This repository contains a simple but efficient implementation of an argument-based recommender system, which makes use of NLP techniques and a taxonomy and lexicon of connectors to extract argument graphs from the proposals and citizen debates available in the Decide Madrid e-participation platform.
This work (v1.0) was presented as a paper at Joint Workshop of the 3rd Edition of Knowledge-aware and Conversational Recommender Systems (KaRS) and the 5th Edition of Recommendation in Complex Environments (ComplexRec) co-located with 15th ACM Conference on Recommender Systems (RecSys 2021). Virtual Event, Amsterdam, The Netherlands, September 25, 2021. Paper and presentation slides can be found here.
This work (v2.0) was presented as a paper at 23st Annual International Conference on Digital Government Research. Virtual Event, Seoul National University, South-Korea, June 15, 2022. Paper and presentation slides can be found here.
Part of this work (v2.6.4) has been submitted as a paper to Government Information Quarterly. A draft of the paper can be found here.
This system is composed of 2 main modules, which are:
- Argument Miner: automatic argument extractor based on NLP techniques and a lexicon of connectors.
- Argument-based Recommender System: It makes recommendations of proposals and arguments based on topics and aspects of interest to the user.
This project uses the two-level taxonomy of argument relations and the set of linguistic connectors (in English and Spanish) published in the argrecsys/connectors repository.
We propose an heuristic method aimed to automatically identify and extract arguments from textual content, which is evaluated on citizen proposals and comments from the Decide Madrid e-participatory platform. The method follows a simple but effective algorithm to address the three basic tasks of argument mining, namely argument detection, argument constituent identification, and argument relation recognition using the two-level taxonomy.
We present some statistics from a preliminary offline test on the automatic identification and extraction of arguments from the citizen proposals available in the Decide Madrid database:
- From the full list of 318 connectors in Spanish, 11,645 arguments are identified and extracted (9,676 from simple sentences and 1,969 from compound sentences).
- Arguments were automatically extracted as follows: 2025 in the proposals and 9620 in the proposal comments.
- Of the 11,645 arguments extracted (some proposals had more than one argument), 5,136 (44.1 %) were identified with connectors from the CONTRAST category, 4,669 (40.1 %) from the CONSEQUENCE category, 1,316 (11.3 %) from the CAUSE category, 505 (4.3 %) from the ELABORATION category, and 19 (0.2 %) from the CLARIFICATION category.
- The descriptive analysis of the result of the automatic annotation can be seen in the following report.
- Network analysis to find the argumentative threads within the proposals can be seen in the following report.
All the results generated by the Argument Miner and Recommender System can be consulted in the following folder.
Example in JSON format of an argument extracted from a citizen proposal about public transportation by the Argument Miner.
{
"5717-0-1-1": {
"proposalID": 5717,
"majorClaim": {
"entities": "[]",
"text": "Allowing pets on public transport",
"nouns": "[pets, transport]"
},
"sentence": "We are almost forced to use public transport in the city but pets are not allowed in EMT",
"claim": {
"entities": "[]",
"text": "We are almost forced to use public transport in the city",
"nouns": "[use, transport, city]"
},
"linker": {
"value": "but", "intent": "attack",
"category": "CONTRAST", "subCategory": "OPPOSITION"
},
"premise": {
"entities": "[EMT]",
"text": "pets are not allowed in EMT",
"nouns": "[pets]"
},
"mainVerb": "forced",
"pattern": {
"value": "[S]-[conj_LNK]-[S]-[PUNCT]",
"depth": 1
},
"syntacticTree": "(sentences
(S (sn (PRP We)) (group.verb (VBP are) ... ))
(conj but)
(S (sn (NNS pets)) (group.verb (VBP are) (RB not) ... ))
(PUNCT .))"
}
}Example in XML format of recommendations of citizen proposals and arguments about public transportation generated by the Argument-based Recommender System
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<recommendations>
<proposals quantity="5">
<proposal id="20307" topics="buses" categories="mobility" date="2017-12-10" districts="Tetuán">
Urban buses connecting San Chinarro and Las Tablas with Cuatro Caminos</proposal>
<proposal id="1432" topics="environment" categories="mobility" date="2015-09-18" districts="city">
Public transportation in Madrid Río</proposal>
<proposal id="5717" topics="pets" categories="mobility" date="2015-11-18" districts="city">
Allowing pets on public transport</proposal>
<proposal id="4671" topics="public transport" categories="mobility" date="2015-11-05" districts="city">
Public transport price</proposal>
<proposal id="2769" topics="transport pass" categories="mobility" date="2015-10-07" districts="city">
The Transport Pass should expire in one month</proposal>
</proposals>
<topics quantity="1">
<topic value="transport" aspects="subway,use,price,transports" quantity="4">
<aspect value="subway" quantity="2">
<argument id="20307-1">
<claim>The PAU of Norte Sanchinarro Las Tablas are poorly served by public transport</claim>
<connector category="cause" subcategory="reason" intent="support">due to</connector>
<premise>the ineffectiveness of light subway</premise>
</argument>
<argument id="1432-1">
<claim>The Madrid Rio park was created promising that public transport would reach there</claim>
<connector category="contrast" subcategory="opposition" intent="attack">but</connector>
<premise>it is false, the Legazpi subway is far away and buses are non-existent</premise>
</argument>
</aspect>
<aspect value="use" quantity="1">
<argument id="5717-1">
<claim>The use of public transport in the city is almost forced</claim>
<connector category="contrast" subcategory="opposition" intent="attack">but</connector>
<premise>in EMT pets are not allowed</premise>
</argument>
</aspect>
<aspect value="price" quantity="1">
<argument id="4671-1">
<claim>Lower the price of transportation</claim>
<connector category="cause" subcategory="reason" intent="support">because</connector>
<premise>it is very expensive</premise>
</argument>
</aspect>
<aspect value="transport" quantity="1">
<argument id="2769-1">
<claim>The Madrid Transport Pass expires in 30 days</claim>
<connector category="contrast" subcategory="opposition" intent="attack">but</connector>
<premise>not all months have 30 days, there are several months that have 31 days</premise>
</argument>
</aspect>
</topic>
</topics>
</recommendations>The other results (in file form) of the argument extractor and recommender system can be viewed here.
The implemented solutions depend on or make use of the following libraries and .jar files:
- JDK 16
- Stanford CoreNLP 4.3.1
- MySQL Connector 8.0.22
- MongoDB Java Driver 3.12.10
- Snake YAML 1.9
- JSON Java 20210307
Please read the contributing and code of conduct documentation.
Created on Apr 10, 2021
Created by:
This project is licensed under the terms of the Apache License 2.0.
This work was supported by the Spanish Ministry of Science and Innovation (PID2019-108965GB-I00).
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent