Fake news detection using semantic and linguistic features

The advent and proliferation of Internet and social networking-websites give rises to a lot of challenging problems for information management, filtering and authenticity of information. In the online world, a growing concern of the users is to handle fake news.  Fake NEWS generally consis

Project Title

Fake news detection using semantic and linguistic features

Project Area of Specialization

Artificial Intelligence

Project Summary

The advent and proliferation of Internet and social networking-websites give rises to a lot of challenging problems for information management, filtering and authenticity of information. In the online world, a growing concern of the users is to handle fake news.  Fake NEWS generally consists of deliberated hoaxes or misleading information related to some entity of the world. The intent is to damage the reputation or to gain material benefits.  Fake NEWS detection is vital problem for online journalism. Online Fake NEWS has several dimensions like: source, target-entity, content, social context, visual contents, in-links, out-links, model of spread of news, social and behavioral analysis of the source and target entities to name a few. This Final Year project proposed a Natural Language based model for Fake-News detection based on content analysis and linguistic feature analysis. The approach belongs to knowledge based fake news detection systems. This Fyp project is limited to content based dimension.

Project Objectives

News such as fake or legitimate news is main source of information for people to help gain insight into the areas they are going to percept. Now a day’s information is readily available but with so much information living online news issues arrive misinformation, something known as confirmation bias which is cognitive bias where we seek out information to prove what we already thought to be true. Over the recent years, the altitude of online social media has dramatically facilitated the people in such way to communicate with each other. Users of online social media share information which gets connected with other people and stay informed about trending events. However, recently, information appearing on social media is wide and rich in content, in some cases, it intends to mislead.This Final Year project proposed a Natural Language based model for Fake-News detection based on content analysis and linguistic feature analysis. The approach belongs to knowledge based fake news detection systems. 

Project Implementation Method

The main contributions are:

(i) linguistic feature extraction from the news

(ii) analysis of syntactic, semantic and entailment analysis

(iii) use of SVM (Support Vector machine) based classifier

(iv) Deep learning models i.e BERT.

There were 7 approaches which were used as projects implementation:

1. Vectorization of lexical chains (semantic feature) using TF-IDF based score for classification

2. Centroid based approach for labelling

3. qlexical word weighting technique based on term frequency and cumulative frequency using TF-IDF

4. Summarization using lexical chains

5. Use of TF-IDF using Tri-gram with char and word vectorizer

6. So using summary from lexical chains now we need to convert these semantic text into words to map in vector space dimensions so we applied word embeddings to those semantic text with BiLstm.We got some good results.

So with all these there were semantic relationship between chains (documents) but there was not context based semantic relationship where we define a same word can appear in two different sentence with different meaning how can we identify that helped us in making a difference between two sentence for using embeddings in multi-dimensional spaces.Below is the solution for it

7.BERT (Bidirectional Encoder Representations from Transformers ) : It solves the problem of context based semantic relationship for eg: (i) I went to the river bank (ii) I need to go to bank to make a deposit .Here we can see "bank" is both sentences have different context so Bert has solved this problem for us.We are getting some good results from this as this uses mutli-embeddings i.e word, sentence and position.

Benefits of the Project

The main aim of the Fyp is to reveal the benefits of artificial intelligence tools used in the detection of fake news and their success levels in different applications.Social networks have become an important way for people to communicate with each other and share their ideas. The most important feature of social networks is the rapid information sharing. In this context, the accuracy of the news or information published is very important. The spread of fake news in social networks has recently become one of the biggest problems. Fake news affects people’s daily life and social order and may cause some negativity.  As a result of the study, it was concluded that the success levels of artificial intelligence tools are over 90%. We all know how fake news can affect the industries and people all over the world, so this project a fake news detection model can help people or industries to prevent from negativity.

Technical Details of Final Deliverable

This Fyp is a research based project although after some experiments we found some good results so this project can be integrated into any website or web application or even mobile application with relevant criteria for developnment process. As it will be integrated with specific domain application. Here are some techincal details :

1. There will be pre-trained model with million of words so an applicaion which constantly recieves news like raw data will be detected as fake or real

2. There must be application or website where news must be  extracted so that it can be labelled as fake or real although this project uses supervised learning approach. Also these extracted labelled results can be shown on gui of mobile or websites.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

IT

Other Industries

Education

Core Technology

Artificial Intelligence(AI)

Other Technologies

Big Data

Sustainable Development Goals

Industry, Innovation and Infrastructure

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com