BBCS 2023 P3 Advanced Track Submission

Introduction

In the travel industry, often, Artificial Intelligence Tools recommending destinations for travel seem to either consider past travel activity from users (which is counterproductive since users may prefer going to many new places or going to similar places, unpredictable by the AI), or are mostly merely showing descriptions of any recommended destinations, not being very personalized for the user. This is not ideal, since

Idea

Create an Artificial Intelligence tool that references a database of destinations (only places in Singapore in this project) and their features, and tries to find matches with user preferences (different from past travel activity, considering that users could want to visit new places; in this case, user input). The AI would first generate 3 top suggestions for possible destinations, then the AI tool would generate sentences to explain its choices to the users and give details about the destinations. This would both preserve the ability of being able to enter open-ended prompts, being more user-friendly, while still allowing users to find destinations with tools that do not consider the user's past history or assume that the user likes what the general public likes.

Details on AI Tool

The AI tool is a Naive Bayes Classifier, taking in user input as a query, lemmatizing the words and removing stop words as preprocessing, to decide good destinations to recommend to the user. The second part will be a series of transformers to generate a sentence to justify its recommendation to the user, with each run of it taking in a destinations as an output from the first part, as well as the words from user input after preprocessing in the first part.

An important note is that the descriptions will be generated based on what the users put in as well, not merely the destiantion, and it is planned to be more based on input rather than destinations recommended, to increase personalization and make it (hopefully) better than merely showing a fixed description each destination.

Usage of the tool

At the homepage (picture below), you will type in a search query into the search bar, and it will redirect you to the search page.

The following is an image of the search page:

Running the WebApp

There is one missing file in this repository, which is a Word2Vec model released by google, to be named GoogleNews-vectors-negative300.bin in the root folder, and it can be downloaded at this link: [https://drive.google.com/file/d/0B7XkCwpI5KDYNlNUTTlSS21pQmM/edit?resourcekey=0-wjGZdNAUop6WykTtMip30g]. (It was not included in this repository since it was a very large file)

To run this WebApp, download this repository and run app.py. Note that it will print out some loading messages.

Ending words

We hope that this project is helpful for the travel industry, and will turn out useful for users.