- Run below commands in their respective directories.
- npm install
- npm start
- Press Ctrl+C to exit the localhost app
- source bin/activate
- python app.py
- Press Ctrl+C to exit the localhost app
- deactivate (to deactivate the virtualenv)
Description: Based on sarcastic (The Onion) and non-sarcastic headlines (HuffPost), the system would help in sarcastic headline awareness and detection. We are diving into sarcastic commentary to see if we can predict sarcasm in a headline.
A business needs to drive product marketing, sales, and operations according to the user feedback, product market fit, and public sentiment. The user feedback cycle, driven by dopamine generative experiences, is very clear and there product managers hired with the sole responsibility to make sure the user's are 'hooked' to the product. Companies like Mixpanel have built industry leading software to understand the meaning behind each user activity to understand UX friction to help their customers get closer to product market fit. Social media analysis is currently avaliable via companies like Sprinklr, but they rely heavily on existing marketing, ads, and social sentiment campaign data to tell their clients whether certain sentiment behind a certain ad was postivie or negative. In a nutshell, they can only analyze social media sentiment based on past sentiment and tell their customers if they are doing well or worse.
But there is a missing link. Sarcasm Detection for business use case is that link. We intend to provide a way for businesses to use our tool to help them understand if the media sentiment they see from other tools is actually the sentiment the users/media is expressing or if its sarcasm. We intend to use a Kaggle dataset of news headlines (https://www.kaggle.com/rmisra/news-headlines-dataset-for-sarcasm-detection) and also try to use small sized posts from media platforms and apply similar logic to understand posts behind a 'trending' topic on a social media platform.
Our main goal is to get a working prototype with the Kaggle dataset. The Kaggle dataset's point, according to the dataset description, was to circumvent the noise in Twitter data. We will move to social media data once we can have satisfying outcome from Kaggle. The search can be based on hashtags or profile name. Sarcastic comments related to the input will get displayed along with key KPI’s on the dashboard.
Overview: React <-> tableau-react <-> Flask Python Server <-> TabPy <-> Python ML
- tableau-react: will load Tableau report in React
- TabPy: will update Tableau reports with update from Python
- Datasets: Kaggle dataset (json), Social media APIs, Cornell sarcasm dataset (For training the model)
- ML frameworks: SKlearn, Tensor flow, Pandas, NLTK/Stanford NER
- Visualization: Tableau
- Web application: ReactJS
- Cloud/Infra Platform: AWS EC2, docker, MongoDB
- To be determined
- Product manager: A product manager's responsibilities include making judgements based on market sentiment.
- Sales team member: A sales team member can find correlation between sarcastic sentiment with product performance in a market.
- Marketing team person: A marketing team member can use sarcasm detection to decide market campaign to counter negative sarcasm.
- A product manager can generate a dashboard of user sarcasm sentiment around the company's product in a region to show other stakeholders and take necessary actions to address the sentiment.
- A sales team member can use SarcasmDetection to know which regions under/over performed and why.
- A marketing team member can use SarcasmDetection to adjust ads/outreach campaigns for a product in a region.