Nowadays all platforms with high inter-user interaction as social media cause the increase of information available and provided by any user in the web. The increase of the information content does not correspond to a full truthfulness of the entire information, thus affecting also reputation of all users who contribute with their own posts. However, considering several truth theories, the concept of truthfulness assumes different meanings depending on the philosophical theory considered. This thesis work introduces in literature the opportunity of extracting for a specific topic tweets sets dealing respectively with different truth theories. Analysing in detail all content generated from Twitter platform, each theory definition is reinterpreted and modeled in algorithms based at first on semantic similarity computed between short texts extracted from overall content. Further metrics as user reputation computed for each truth theory considered can help in tweets extraction or results validation. The work also shows future suggestions about adapting algorithms in a general way and enriching literature analysing also other truth theories. Additionally, algorithms could be adapted depending on further features that could be introduced in Twitter.
Oggigiorno le piattaforme con alta interazione inter-utente, come i social media, contribuiscono ad espandere la mole di informazioni accessibili e fornite da chiunque sul Web. Ad un aumento del contenuto informativo non sempre corrisponde una piena veridicità dell'informazione circolante, influenzando inoltre la reputazione degli utenti che contribuiscono con i propri post. Tuttavia, il concetto di veridicità, quindi di verità, assume sfumature diverse a seconda della teoria filosofica considerata. Questa tesi si focalizza sulla possibilità di estrarre insiemi di testi riguardanti uno specifico argomento, ciascuno rispondente a un diverso concetto di verità considerata durante l'estrazione. Analizzando nello specifico contenuti generati in Twitter, le definizioni di ciascuna teoria vengono reinterpretate e modellate tramite algoritmi basati sulla similarità semantica calcolata fra brevi testi. Ulteriori metriche come la reputazione di un utente in base alla teoria considerata possono aiutare nella fase di estrazione, o validare i risultati ottenuti. Il lavoro presenta spunti futuri su un utilizzo ad ampio spettro e ad arricchimenti dell'opera con altre teorie della verità, oltre a possibili adattamenti degli algoritmi derivanti dall'introduzione di funzioni innovative in Twitter.
A methodology for extracting relevant facts from social networks
PIACENTINI, MARIANNA
2014/2015
Abstract
Nowadays all platforms with high inter-user interaction as social media cause the increase of information available and provided by any user in the web. The increase of the information content does not correspond to a full truthfulness of the entire information, thus affecting also reputation of all users who contribute with their own posts. However, considering several truth theories, the concept of truthfulness assumes different meanings depending on the philosophical theory considered. This thesis work introduces in literature the opportunity of extracting for a specific topic tweets sets dealing respectively with different truth theories. Analysing in detail all content generated from Twitter platform, each theory definition is reinterpreted and modeled in algorithms based at first on semantic similarity computed between short texts extracted from overall content. Further metrics as user reputation computed for each truth theory considered can help in tweets extraction or results validation. The work also shows future suggestions about adapting algorithms in a general way and enriching literature analysing also other truth theories. Additionally, algorithms could be adapted depending on further features that could be introduced in Twitter.File | Dimensione | Formato | |
---|---|---|---|
2016_04_Piacentini.pdf
solo utenti autorizzati dal 05/04/2017
Descrizione: Testo della tesi
Dimensione
1.39 MB
Formato
Adobe PDF
|
1.39 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/121048