Fake News Detection using Machine Learning

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 4 Issue 2, February 2020 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD30014 | Volume – 4 | Issue – 2 | January-February 2020 Page 512
Fake News Detection using Machine Learning
Prasanth. K1, Praveen. N1, Vijay. S1, Auxilia Osvin Nancy. V2
1B.Tech Student, IT, SAEC, Chennai, Tamil Nadu, India
2Assistant Professor, IT, SAEC, Chennai, Tamil Nadu, India
ABSTRACT
Recently, fake news has been incurring many problems to our society. As a
result, many researchers have been working on identifying fakenews.Mostof
the fake news detection systems utilize the linguistic feature of the news.
However, they have difficulty in sensing highly ambiguous fake news which
can be detected only after identifying meaning and latest related information.
In this paper, to resolve this problem, we shall present a new Korean fake
news detection system using fact DB which is built and updated by human’s
direct judgement after collecting obvious facts. Our system receives a
proposition, and search the semantically relatedarticlesfromFactDBinorder
to verify whether the given proposition is true or not by comparing the
proposition with the related articles in fact DB. To achieve this, we utilize a
deep learning model, Bidirectional Multi-Perspective Matching for Natural
Language Sentence(BiMPM), which hasdemonstrateda goodperformancefor
the sentence matching task. However, BiMPM has some limitations inthatthe
longer the length of the input sentence is, the lower its performance is, and it
has difficulty in making an accurate judgement when an unlearned word or
relation between words appear. In order to overcome the limitations,weshall
propose a new matching technique which exploits article abstraction as well
as entity matching set in addition to BiMPM. In our experiment, weshall show
that our system improves the whole performance for fake news detection.
KEYWORDS: Fake news detection, Sentence matching, Natural Language
Processing, Deep learning, BiLSTM model, Machine Learning
How to cite this paper: Prasanth. K |
Praveen. N | Vijay. S | Auxilia Osvin Nancy.
V "Fake News Detection using Machine
Learning" Published
in International
Journal of Trend in
Scientific Research
and Development
(ijtsrd), ISSN: 2456-
6470, Volume-4 |
Issue-2, February
2020, pp.512-514, URL:
www.ijtsrd.com/papers/ijtsrd30014.pdf
Copyright © 2019 by author(s) and
International Journal ofTrendinScientific
Research and Development Journal. This
is an Open Access article distributed
under the terms of
the Creative
CommonsAttribution
License (CC BY 4.0)
(http://creativecommons.org/licenses/by
/4.0)
INRODUCTION
Fake news has been incurring many problemstooursociety.
Fake news is the ones that the writer intends to mislead in
order to achieve his/her interestspoliticallyoreconomically
on purpose [1]. With the generation of a huge volume of
internet news and social media. It becomes much more
difficult to identify fake news personally. Recently, many
researchers have worked on fake news detection system
which automaticallydeterminesifanyopinionclaimedinthe
article contains fake content [2]. In a large context,theforms
of their research are carried out with the method that
connects the linguistic patternofnewstodeception,andthat
verifies deception by utilizing external knowledge [3]. The
first approach can quickly verify fake news at a low cost.
However, in order to detect clever fake news, it is necessary
to grasp the semantic content of the article rather than
partial patterns and verify it through external facts updated
by human. Therefore, we search the in put proposition and
related articles from the Fact DB, and develop the fake news
detection system to verify if the found articles and
proposition are semantically related.
Related Works:
Recently, as the deep learning in the NLP field has been
developed, various types of the sentence matching
techniques have been introduced. We introduce the related
research of the sentence matching techniques as we divide
the works into the unsupervised learning, and supervised
learning based works.
A. Unsupervised Learning:
One of the most important elements in the sentence
matching is the way of expressing a word into a data
structure. The existing method of expressing words is one-
hot encoding vector. However, this method requires lots of
dimension to express a single word, and cannot express the
relation between words. Overcoming these shortcomings,
the word-to-vector (word2vec) [5] method was proposed
which maps significant information into the vector of fixed
dimensions. The word2vec is enabled to learn the weight to
increase the probability that the nearby words will appear
for the main word, and uses the corresponding weight as a
vector. As an extended research of word2vec, sentence-to-
vector (sent2vec).
B. Supervised Learning:
Recently, the research of the machine comprehension is
developed with attention mechanism and BiLSTM. LSTM
resolved the vanishinggradientproblemofRecursiveNeural
Network (RNN) by adding the layer that forgets the past
information, and remembers the current one to the cell.
Since LSTM handles sequential inputs, it is often used for
encoding and decoding of sentences. However, as the length
of LSTM becomes longer, the model loses the information,
and it shows a tendency to remember the latter information.
Therefore, scholars worked on improving the performance
of the existing LSTM through attention mechanism which
reminds important information selectively. They also
showed using BiLSTM together can improve performance.
The Bi-Directional Attention Flow(BiDAF)[7]minimizesthe
IJTSRD30014

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
loss of information by applying the attention mechanism at
each time stamp of LSTM. In particular, BiMPM [8] applied
BiLSTM and attention mechanism to sentence matching.
Proposed System:
This section discusses the proposed solution for fake news
detection by combining Fake News with SentimentAnalysis.
Proposed solution is shown in Fig. 1. It consists of various
steps as below:
Step 1 : Merged data set was prepared using from different
data sets namely Politifact, Kaggle and Emergent
datasets.
Step 2 : The different text preprocessing techniques like
bigrams (series of two words taken from a given
text) ,trigrams (continuous series of three words
taken from example text), CountVectorizer (count of
terms in vector/ text , term frequency-inverse
document frequency (tf-idf) vectorizer.
Step 3 : We have used tf-idf vectorizer on twitter dataset
along with cosine similarity to build our vocabulary.
Then Naive Bayes classifier was used to predict the
sentiment of news statement oftestdata set(Merged
data set) as shown in Fig. 2.
Step 4 : We added additional columns: tf-idf scores,
sentiments and Cosine similarity scores in Merged
data set.
Step 5 : Training model was built using Naive Bayes and
Random Forest (train-test ratio: 3:1)
Step 6 : Performance is evaluated and compared using
accuracy.
Proposed solution consists of important steps 2 to 4 as
preprocessing. It uses tf-idf Vectorizer with cosine
similarities method for tokenizing a collection of text
documents along with building a vocabulary of pre-existing
words. Further we encoded the novel documents using that
vocabulary. The encoded vector is returned with length of
the entire vocabulary (bag of words) and an integer count
for the number of times each word appeared had in the
document.
System Evaluation:
We evaluate the performance of proposed system in this
section. Given the relevant article on the input proposition,
the evaluation verifies the ability to determine whether the
semantic content of the input propositioncan befoundinthe
relevant article. We train the BiMPM which isthefoundation
of our system, and the experiments identify how much the
performance improves by adding modules proposed
previously. We first build the data set directly to train the
BiMPM to output true or false when given a short article
consisting of three or four sentencesandpropositions.In the
datasets construction, the following policy is set up to
proceed with the learning.
1. Extract one sentence from a short article, and use it into
an input proposition.
2. Generate the data, which is true through variations of
thesaurus, a change of word orders, and omission of
some contents.
3. Distort some information such as numbers, nouns, and
verbs or omit words to generate false data.
Fig.2. Training accuracy of test set for each epoch.
Proved the sentences used in the new testsetarelonger,and
consist of new words that are not in the previous dataset. In
terms of the using True Positive Rate (TPR) as the y-axisand
False Positive Rate (FPR) as the x-axis.
Conclusion:
In this paper, we have proposed the fake news detection
system using Machine learningwhichisbuiltandupdatedby
human’s direct judgement. Our system receives a

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
proposition as an input to verify, and search the related
articles so that it verifies if the article found by the entered
proposition can be semantically concurred with the
proposition. To achieve this, we utilized model which is a
deep learning model for sentences matching and machine
learning. However, even though has shown good
performance in variousdatasets,ithassomelimitationssuch
that the longer the length of the input sentence is, the lower
its performance is, and it has difficulty inmakinganaccurate
judgement when an unlearned word or relation between
words appear. In order to overcome the limitations,wehave
presented the new matching technique which makes use of
article abstraction as well as entity matching set besides
BiMPM. In our experiment, we have shown that our system
improved the whole performance for fake news detection.
Reference:
[1] H. Allcott and M. Gentzkow, “Social media and fake
news in the 2016 election,” Journal of economic
perspectives, vol. 31, no. 2, pp. 211–36, 2017.
[2] N. J. Conroy, V. L. Rubin, and Y. Chen, “Automatic
deception detection: Methodsforfindingfake news,”in
Proceedings of the 78th ASIS&T Annual Meeting:
Information Science with Impact: Research in and for
the Community, p. 82, American Society for
Information Science, 2015.
[3] V. L. Rubin, Y. Chen, and N. J. Conroy, “Deception
detection for news:threetypesoffakes,”inProceedings
of the 78th ASIS&T Annual Meeting: Information
Science with Impact: Research in and for the
Community, p. 83, American Society for Information
Science, 2015.
[4] S. R. Bowman, G. Angeli, C. Potts, and C. D. Manning, “A
large annotated corpus for learning natural language
inference,” arXiv preprintarXiv: 1508.05326, 2015.
[5] M. Potthast, S. K¨opsel, B. Stein, and M. Hagen, “Click
bait detection,” in Advances in Information Retrieval.
Springer, 2016.
[6] M. Pagliardini, P. Gupta, and M. Jaggi, “Unsupervised
learning of sentence embeddings using compositional
n-gram features,” arXiv preprintarXiv: 1703.02507,
2017.
[7] V. Piek et al., “Newsreader: How semantic web helps
natural language Recognizing click bait as false news,”
in ACM MDD, 2015.
[8] W. W. Yang. ” liar, liar pants on fire: A new benchmark
dataset for fake news detection,” arXiv preprint arXiv:
1705.00648, 2017.
[9] Z. Wang, W. Hamza, and R. Florian, “Bilateral multi-
perspective matching for natural language sentences,”
arXiv preprint arXiv: 1702.03814,2017.
[10] K. Shu, D. Mahudeswaran, and H. Liu. ”Fake News
Tracker: a tool for fake news collection, detection, and
visualization,” Computational and Mathematical
Organization Theory, pp.1-2, 2018.

Fake News Detection using Machine Learning

Recommended

More Related Content

What's hot (20)

Similar to Fake News Detection using Machine Learning (20)

More from ijtsrd (20)

Recently uploaded (20)

Fake News Detection using Machine Learning