What’s the Next Word?

class: center, middle, inverse, title-slide

.title[
# What’s the Next Word?
]
.author[
### Abril Susana
]
.date[
### March 7, 2025
]

---

# Welcome to "What's the Next Word?"

- **Goal:** Predict the next word based on user input  
- **Collaboration:** Data Science Capstone Project with SwiftKey  
- **Technologies Used:**  
  - **shiny**: To create the interactive app  
  - **tm**, **text2vec**, **dplyr**: For text processing and data manipulation  
- **Data Source:** English text from HC Corpora (blogs, Twitter, and news)

**How It Works:**  
The app provides predictions by analyzing the user input and suggesting the top three most likely next words.

<img src="https://raw.githubusercontent.com/abrilsm114/data-science-capstone/main/app.gif" width="45%" style="display: block; margin: auto;" />
---

# The Prediction Algorithm

- **Data Preparation:**  
  - Cleaned 1 million lines of text (removing noise, punctuation, and stopwords).
  - Tokenized data to generate **n-grams** (up to 6-grams) and computed frequency distributions.
- **Algorithm Used:**  
  - The app uses a **back-off model** to predict the next word based on the user’s input.  
  - **Maximum Likelihood Estimation** (MLE) is used for calculating n-gram probabilities.  
  - The top 3 predictions are presented based on these probabilities, improving accuracy.

---

# The Shiny App Interface & How It Works

- **Functionality:**  
  - **Real-Time Predictions:** As you type, the app predicts the next word.
  - **Interactive Feature:** Click any of the predicted words to auto-complete your sentence.
  - **Instant Feedback:** Predictions are updated instantly based on your input.

---

# Try it Out & Learn More

- **Live Demo:** Check out the app on [shinyapps.io](https://asusana.shinyapps.io/predict-next-word/)

- **How to Use the App:**
  1. Enter text in the input box.
  2. See the top 3 predicted words appear below.
  3. Click a word to add it to your input.