=Paper=
{{Paper
|id=Vol-3395/T5-7
|storemode=property
|title=Classification of COVID-19 Tweets
|pdfUrl=https://ceur-ws.org/Vol-3395/T5-7.pdf
|volume=Vol-3395
|authors=Sumana Madasu
|dblpUrl=https://dblp.org/rec/conf/fire/Madasu22
}}
==Classification of COVID-19 Tweets==
<pdf width="1500px">https://ceur-ws.org/Vol-3395/T5-7.pdf</pdf>
<pre>
Classification of COVID-19 Tweets
Sumana Sree Madasu
Indian Institute of Science Education and Research Tirupati,Andhra Pradesh,India


                                      Abstract
                                      Classification is a vital work to human beings in day today life as it breaks down complex subjects. In
                                      the same way, text classification is very important to understand and realize the subject of the text.

                                      Keywords
                                      data preprocessing, Support Vector Machine, Doc2Vec.




1. Introduction
A classification pipeline is constructed to classify the given set of COVID -19 tweets. Here,
Data set containing COVID-19 tweets are classified into three different classes using a specific
classification pipeline and this classification solves a real world problem. Data is preprocessed,
Vectorized using a Doc2Vec model and then Support Vector Machine classifier is used to train
the model. The model is then saved and can be used to predict new tweets.


2. Data Preprocessing
Given data have different characteristics, numbers, symbols etc. and this makes the text
preprocessing one of the most critical steps in the classification pipeline. Firstly, necessary
libraries are imported and then the data is preprocessed by using a few methods depending
on the classification task. Here, two functions called clean_text and clean_numbers are used,
they return modified text [1](after HTML decoding, lowercase text, replaces symbol by space,
deleting few symbols, stopwords and numbers).


3. Implementing Doc2Vec
To implement Gensim’s Doc2Vec, every document has to be labeled [2]. Here ,this is done using
the TaggedDocument method. Then data is split into train and test sets using the train_test
_split from Scikit-Learn library and stratify parameter is used on the data column ‘label’. For
representing each tweet , a Doc2Vec model is built with each vector as 300 dimension, which
iterates over the training corpus 30 times, minimum word count is set to 2 to discard rarely
occurring words and alpha is set to 0.065. Vector representation is obtained using the above
trained Doc2Vec model for the vocabulary of the data.


Envelope-Open madasu@students.iisertirupati.ac.in (S. S. Madasu)
                                    © 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
 CEUR
 Workshop
 Proceedings
               http://ceur-ws.org
               ISSN 1613-0073
                                    CEUR Workshop Proceedings (CEUR-WS.org)
4. Classification Algorithm
Support Vector Machine :

  Now, the training data vector list is fitted on the Support Vector Machine (SVM)
classifier,which is a traditional machine learning technique [3] and this model is used
to predict the labels on the testing data vector list(i.e., validation data list).


5. Classification Report
Classification metrics like f1 score, precision and accuracy are calculated to evaluate
the model.

  Accuracy - 0.5563

                      precision recall f1-score support
         Antivax           0.53 0.54       0.53     216
         Neutral           0.59 0.62       0.60     327
          ProVax           0.54 0.51       0.52     336
        accuracy                           0.56     879
      macro avg            0.55 0.55       0.55     879
    weighted avg           0.56 0.56       0.56     879


6. Conclusion
Classification performances of the classifiers depends on quality of training text cor-
puses to some extent. Most text classification problems are linearly separable [4] and
SVM helps in finding those linear separators and this says that SVM can perform well
in text classification [5] when over fitting issue is taken care of. For large data samples,
Deep Learning techniques show better performance than Traditional Machine Learning
techniques.



References
[1] A. Kadhim, An evaluation of preprocessing techniques for text classification,
    International Journal of Computer Science and Information Security, 16 (2018)
    22–32.
[2] Q. Le, T. Mikolov, Distributed representations of sentences and documents, in:
    International conference on machine learning, PMLR, 2014, pp. 1188–1196.
[3] E. Ikonomakis, S. Kotsiantis, V. Tampakas, Text classification using machine learning
    techniques, WSEAS transactions on computers 4 (2005) 966–974.
[4] T. Joachims, Text categorization with support vector machines: Learning with
    many relevant features, in: European conference on machine learning, Springer,
    1998, pp. 137–142.
[5] A. Basu, C. Walters, M. Shepherd, Support vector machines for text categoriza-
    tion, in: 36th Annual Hawaii International Conference on System Sciences, 2003.
    Proceedings of the, 2003, pp. 7 pp.–. doi:10.1109/HICSS.2003.1174243 .
  References:

</pre>