# What do we know about non-pharmaceutical interventions?

# Team-1
## Team Members
P Advaith Alenkrith (170010024)\
Vaibhav Kumar Rai (1913107)\
V prakhyath sree harsgha (170010023)\
Mood Suman Chauhan (1913103)\
Gorthi Jaswanth (170010025)\
Maganuru jayasurya (170010026)

# Introduction
- Covid-19 has been one of the worst nightmare for the world till now, deciding how to take control over is very difficult since there is no vaccine or any specific medecine available.
- So first thing which comes in mind is what are the precautions we must take? This will come under NPIs.
- Nonpharmaceutical Interventions (NPIs) are actions, apart from getting vaccinated and taking medicine, that people and communities can take to help slow the spread of illnesses like pandemic influenza (flu).
- Given all the data available on kaggle about different research articles about epidemics and pandemics.
- We will find relevant article for NPIs.
- Our Approach is to do preprocessing and find all articles containing some keywords related to NPIs. Then we will do sentiment analysis of that.
- We also plotted graphs for available data about dignosed and onset patients. Compared situation in China and Italy with time.

## How we did preprocessing?

First we create a dataframe using the file metadata.csv with following columns : 'cord_uid','sha','source_x','title','doi' etc. Then we will import all the json files into single variable named filenames and then we use lists to seperate the content and title of articles form other necessary columns for preprocessing.

- Stemming:
- While doing the search in json files we use the stemming funtion.Stemming is reducing the words that arised from the same root word into the corresponding root word. At first we are searching the files based on word search algorithm.The use of stemming is explained below with an example.Without stemming when we search for word "gathering' The search will show only the files that contains the words "gathering".But when we use stemming now the word 'gathering' is truncated to 'gather'and we use both the words 'gathering' and 'gather' for search. So by this we are searching the files more accurately as 'gethering' and 'gather' are very close words as one word is created by adding a suffix to another word(root word).

- Lemmatization
Lemmatization unlike streaming reduces the infected words properly ensuring that the root word belongs to the language. In lemmatization the root is called lemma. We used the python package nltk (Natural Language Toolkit) and used the function WordNetLemmatizer() to improve our search results. Should go through whole Wordnet Corpus so, it takes more time than stemming. But this is useful for accurate predictions

- Stop Words:
- Stopwords are those words that do not provide any useful information to decide in which category a text should be classified. This may be either because they don't have any meaning or because they are too frequent in the classification context.Stop words generally have high frequency and have no meaning.

## Sentiment Analysis
After doing text cleaning and preprosessing we found all relevant articles and then we did sentiment analysis of those articles,Sentiment analysis is the interpretation and classification of emotions (positive, negative and neutral) within text data using text analysis techniques. what is the result of applying NPIs? Is it positive, negative or neutral?
We did Fine-grained Sentiment Analysis, for that we used nltk library for tokenization and finding out whether the words used in abstract of research paper is positive or negative. We have implemented tfidf vectorization and have done k-mean clustring where k is taken as 5. Then those cluster values are saved in frame_array and after that it is given to the function analyzesentiment to predict the type of sentiment and then store all those sentiment in a csv file. Now these documents can be used to review how NPIs were sucessful and what should we do to enforce it.

# Graphical Analysis
We also did plotting of different pie, bar graph to analyze the situation in China and Italy.Based on the data available from WHO and Worldometers. We analyzed how many patient are dignosed daily and how many onset patients are there. We also compared situation in china and italy based on the number of dignosed patients. It shows how condition in italy become worse with time.

### References
1)https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge/tasks?taskId=587
2)https://towardsdatascience.com/sentiment-analysis-with-text-mining-13dd2b33de27