Dad Jokes - Daily Funny Jokes

Funny jokes, powered by AI. This app is one of its own kind in the market. We leverage AI technology to bring you the best experience.

Highlights:
- Train the AI by reacting with the like and dislike button.
- All jokes are manually screened to ensure they are appropriate and funny.
- Simplistic UI design with relaxing background music.

How it works?
1. We processed more than 15,000 joke entries, by breaking them into word tokens, building token vector representations, and then classifying them into roughly 40 categories using state-of-art Support Vector Machine (SVM) algorithm [1].
2. Out of these entries, we randomly selected 400 of them under uniform distribution (about 10 entries for each category) as the “seed” data list.
3. From this filtered list, we carefully handpicked 100 entries as the initial data.
4. Then we baked these 100 predefined entries into the app for everyone to start with.
5. You react to jokes (by tapping either "like" or "dislike" button)
6. We then perform on-device machine learning to compute a heuristic vector. This is how the algorithm learns about your preference. We use Reinforcement Learning [2] to keep tuning your heuristic vector as you use the app.
7. After training with these 100 “seed” entries, we will start serving fully customized jokes that the algorithm thinks you will most likely enjoy. This pattern matching process involves weighted voting of multiple AI algorithms such as Naive Bayes [3] and K-Nearest Neighbors (KNN) [4].
8. Note that our server is completely stateless - it never stores your preferences. It simply receives your locally computed heuristic vector mentioned above, responds with the best matching joke, and then discards it.

References
[1] Support Vector Machine: https://people.eecs.berkeley.edu/~adj/publications/paper-files/SVM-PKDD08.pdf
[2] Reinforcement Learning:http://rail.eecs.berkeley.edu/deeprlcourse-fa19/static/slides/lec-4.pdf
[3] Naive Bayes: http://artificial-intelligence-class.org/assets/slides/07-Naive-Bayes-and-Perceptron.pdf
[4] K-Nearest Neighbor: https://simons.berkeley.edu/sites/default/files/docs/11886/nn-simons-part1.pdf