[Paper Review] Machine Translation, Sentiment Analysis, Text Similarity, Topic Modelling, and Tweets: Understanding Social Media Usage Among Police and Gendarmerie Organizations

Emre Cihan Ateş, Erkan Bostancı+1 more|arXiv (Cornell University)|Jan 29, 2021

Sentiment Analysis and Opinion Mining38 references4 citations

TL;DR

This study analyzes social media engagement of police and gendarmerie organizations across Turkey, Italy, France, and Spain using Twitter data, applying machine translation, sentiment analysis, text similarity, and topic modeling. It finds that Turkey's Jandarma has the highest influence among the organizations studied, highlighting disparities in public engagement effectiveness across nations.

ABSTRACT

It is well known that social media has revolutionized communication. Nowadays, citizens, companies, and public institutions actively use social media in order to express themselves better to the population they address. This active use is also carried out by the gendarmerie and police organizations to communicate with the public with the purpose of improving social relations. However, it has been seen that the posts by the gendarmerie and police organizations did not attract much attention from their target audience from time to time, and it has been discovered that there was not enough research in the literature on this issue. In this study, it was aimed to investigate the use of social media by the gendarmerie and police organizations operating in Turkey (Jandarma - Polis), Italy (Carabinieri - Polizia), France (Gendarmerie - Police) and Spain (Guardia Civil - Policía), and the extent to which they can be effective on the followers, by comparatively examining their activity on twitter. According to the obtained results, it was found that Jandarma (Turkey) has the highest power of influence in the twitter sample, and the findings were comparatively presented in the study.

Motivation & Objective

• To investigate how police and gendarmerie organizations in Turkey, Italy, France, and Spain use social media to engage the public.
• To identify factors affecting the influence and reach of their social media content.
• To compare the effectiveness of communication strategies across national law enforcement agencies using Twitter.
• To assess the role of NLP techniques—such as sentiment analysis and topic modeling—in understanding public engagement patterns.

Proposed method

• Collected and analyzed Twitter posts from police and gendarmerie organizations in Turkey, Italy, France, and Spain.
• Applied machine translation to standardize multilingual content for cross-national comparison.
• Used sentiment analysis to evaluate the emotional tone of official posts.
• Employed text similarity techniques to detect content repetition or templating across posts.
• Conducted topic modeling to identify recurring themes and communication priorities.
• Measured influence through engagement metrics such as likes, retweets, and replies, comparing performance across countries.

Experimental results

Research questions

• RQ1 How do police and gendarmerie organizations in Turkey, Italy, France, and Spain differ in their social media communication strategies on Twitter?
• RQ2 What is the level of public engagement (likes, retweets, replies) for official posts from these organizations?
• RQ3 Which organization demonstrates the highest influence on social media, and what factors contribute to this?
• RQ4 How do sentiment and topic distribution in official posts correlate with audience engagement?
• RQ5 To what extent do multilingual posts and content similarity affect the reach and effectiveness of law enforcement communications?

Key findings

• Turkey's Jandarma demonstrated the highest level of influence on Twitter compared to other police and gendarmerie organizations studied.
• The study found significant variation in engagement levels across countries, with Jandarma posts receiving more interaction than those from counterparts in Italy, France, and Spain.
• Sentiment analysis revealed that positive sentiment was more common in high-engagement posts, suggesting emotional tone affects public response.
• Topic modeling identified recurring themes such as public safety, community service, and operational updates, with Jandarma showing greater thematic diversity.
• Text similarity analysis indicated that some organizations reused content templates, but this did not consistently correlate with higher engagement.
• Machine translation enabled valid cross-national comparisons despite multilingual content, confirming its utility in international social media analysis.