Install these using pip install
Python3
Flask==1.0.2
Flask-SQLAlchemy==2.4.0
pandas
Flask-Dance
wit
newsapi-python==0.2.6
tweepy
tweet-preprocessor
python -m flask run
Sentiment analysis of user reviews for particular disease and medicine Using Wit.Ai with Covid-19 Chatbot
We are trying to give the user fair information about what others think about that medicine he is using and what other
best-rated medicines suggested for this particular disease and User can also compare two Medicines to know the best one.
We are using the UCI drug review dataset for getting reviews having more than 2 million reviews.
We have implemented a chatbot for all your queries on Covid-19 and it works on both Messenger and Telegram.
Our aim is to give a platform that will be useful for both medical professionals and other Users. On this platform, they can study their situation and medicines, they can easily see the reviews for the work of medicine in that condition, and by our monitoring tool, they can get the latest information about the disease and its medicine. It would be really helpful in these pandemic situations like Covid-19 created. These situations create shortages of medicines. So, through our platform medical professionals and users can easily check alternate best medicine for the same situation and much other information that they need.
⌨️ 1.Typing Search
You can normally search through our search box and Suggestions will be provided beneath it for better help to get your desired name.
🎙️ 2.Voice Search
An interactive Voice search option is also available. Just click on the MIC icon to trigger voice search and an interactive modulated voice will help you throughout your search.
At the bottom of the search box, there is an option of selecting a relevant source ( Ex. UCI Dataset, Twitter, Drug.Com(WIP) and other NewsAPI ) from where data is taken and then broken into tokens and Using WIT.AI inbuilt NLP and Sentiment Analysis the Sentiment is negative, positive or neutral is taken out and displayed in results.
We get the relevant data from the user-desired platform and then tokens are generated from the data then Using NLP, Sentiment Analysis with the help of Wit.Ai we get the sentiment in each data and the model predicts the average sentiment of all persons and displays them in results.
Using Sentiment analysis the results for the medicine and disease are shown as written in the previous section.
Now, in results, we also display other medicines that are rated good for the same disease and the other disease that can be cured by the same medicine. You can also download the result as a pdf softcopy.
We also display the Graph to better understand both Sentiment Analysis results and other best-rated medicines.
You can navigate to this feature from the main page navbar and then you can compare two of your medicines that you are confused in and you will get the results according to other reviews. The result will be in two forms, side by side Graph and Comparison Table for the full detailed comparison of both in tabular form.
From here you can get all the records for your previous search with date and time, name of medicine, and disease. You can also see the result again by clicking the See results Button.
You can see what's going on recently or the history of medicine. For the section, the application uses NewsAPI and tweets to show the sentiment of the user changing with the time for that medicine or disease.
Available with both ⌨️Typing Search & 🎙️Voice Search
A chatbot to solve all your queries regarding the Covid-19 situation and tries to help you in the best possible ways.
The chatbot works on both Messenger and Telegram.
Analyzing 2 million of the dataset in just some seconds. That makes it more User-Friendly and Time-Efficient.
This is the overall architecture of our application. on the left side, the red-colored sections are those with which the user interacts and on it right the brief internal working of our application is shown.
Sentiment Analysis is used on the pre-saved dataset (over 2 million) and the fixed data that is being Scraped from Twitter and Drug.com. Then data is broken into tokens and Using this dataset we train our model over Wit.ai, Wit.ai Speech API is also used to make our application voice interactive.
On running our Model on a pre-stored dataset we get the accuracy of about 85% and when the twitter and drug.com dataset is added accuracy increases to 87.05 %.
This project has a very wide area to which we can explore and we can add many new features in it. For the future, We are thinking of adding a feature to suggest the best Doctors and Best hospitals for the searched medicine. and it can be according to the city of the user and also best in his country By getting user location and Analyzing data from different platforms about the best doctors and hospitals for that disease.
- Frontend & UI :-
HTMLCSSBootstrapJavascript
- Backend :-
Flask
- Database :-
PostgreSQL
- Voice intraction :-
Annayang.js
- Authentication :-
Google Auth0
- Model :-
Wit.ai
- Support for Model :-
NLPUCI
☐ It helps the user to better understand the medicine he is using and provide helpful feedback.
☐ Deep sentiment analysis is performed for over 2 million datasets.
☐ It has the ability to suggest best-rated medicines for that disease.
☐ It also gives the list of other diseases on which the same medicine works.
☐ Both graphical and statistical representations of data for better understanding.
☐ Feature to download your result in Pdf softcopy.
☐ The history section for all your searched results so you can check anytime and revisit.
☐ Compare the Meds section for knowing the best medicine B/W 2 if you are confused.
Brain Tumor Detection using Web App (Flask) that can classify if patient has brain tumor or not based on uploaded MRI image.
The image data that was used for this project is Brain MRI images for Brain tumor detaction.(https://www.kaggle.com/navoneel/brain-mri-images-for-brain-tumor-detection)
Data set Link- https://www.kaggle.com/navoneel/brain-mri-images-for-brain-tumor-detection
- Backend :-
Flask
- Model :-
Keras VGG16
MRI images have been considered for this project since it gives the clear structure of the brain, without any surgery it scans and gives the structure of the brain this helps in further processing in the detection of the tumor. Human prediction in classifying the tumor from the MRI leads to misclassification. This motivates our project to construct the algorithm to predict the tumor. Machine learning plays a key role in predicting tumor. In this proposed paper, we have constructed a framework for detecting the brain tumor and classifying its type. The approach goes under pre-processing to filter and smooth the image. The segmentation is carried out by using morphological operation followed by masking, which increases the accuracy in the classification step. The multiple feature extraction methods are utilized to extract the feature from the masked image, and for classification, the kernel SVM is used.