High-Level Project Summary
First things first, the fact that NASA is moving its data to the cloud and making it available for NASA Space Apps Challenge participants was a motivator for us to explore it in a creative approach. As a solution to the challenge we developed a mobile app (flutter, node) that allows users to input short text phrases, the phrases will then be matched with the imagery described in the dataset, then displayed the result to the user request. We used NLP method called semantic research to match the user input and the data. The app is a way to make NASA's imagery accessible to all humans in a pleasing artistic way.
Link to Final Project
Link to Project "Demo"
Detailed Project Description
The procedure started with a brainstorming phase to confirm all our application features and create a prototype for it in order to start implementation phase.
The app contains simply : welcome page, sign in/up page, home screen where the user inputs the text search to be directed to a results page, for each displayed picture the user can like, save and share it to other social media.
The mobile app allows users to input short text phrases, the phrases will then be matched with the imagery described in the dataset, and in the end, we will display the result to the user request. We will use an NLP method called semantic research to match the user input and the data. Semantic search is a method of data searching where a search query seeks to not only identify keywords but also to establish the intent and context of the words being used.
Cleaning and Tokenizing : Data pre-processing is one of the most important steps in text analytics. The purpose of it is to remove any unwanted words or characters that are written for humans to link important words and make sentences for them to read but won’t contribute to topic modeling in any way (for, the, how...etc ).
Building Word Dictionary : In this step, we will build the vocabulary of the corpus in which all the unique words are given IDs and where their frequency counts are also stored.
Feature Extraction (Bag of Words) : A bag of words model (BoW ) is a way of extracting features from the text for use in modeling machine learning algorithms. It is a representation of text that describes the occurrence of words within a document. It involves two things:
- A collection of well-known words
- A metric of the presence of known words
The doc2bow method of dictionary iterates through all the words in the text, if the word already exists in the corpus, it increments the frequency count, otherwise it inserts the word into the corpus and sets its frequency count to 1.
Build Tf-Idf Model :
Term frequency-Inverse Document Frequency (Tf-IDF) is a popular NLP model that helps you determine the most important words in each document in the corpus. Once the Tf-IDF is built, pass it to the LSI model and specify the number of features to build.
Build LSI Model :
Latent Semantic Indexing is method works by identifying the hidden contextual relationships between words. It helps with:
- Finding relationships Between Words (Semantic)
- Information Retrieval (Indexing)
semantic search :
When we enter a search query, the model will return relevant images using similarity score "score". The query and the image are more similar the greater the similarity score.
further :
https://www.canva.com/design/DAFN11_qiWk/17oDdD9WM05frM1Iu5ZMsQ/view?utm_content=DAFN11_qiWk&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton
Space Agency Data
The provided NASA API Portal was our data source from which we explored NASA's imagery dataset and used it for our AI model then deployed on our application.
Hackathon Journey
To begin with, the idea of participating in NASA Space Apps Challenge was in itself a great step we will be taking as Dev_Art team to dive into a new exceptional experience. The team worked in a harmonic way especially while benefiting from the variety of its skillset ( AI/ML : NLP, Mobile front/back-end development, UX/UI Design, Video Editing ). It is true that we hoped to experience it on-site in Algiers hub yet, despite going online, we kept the motivation to the last second doing our best to provide the best outcome for such a great hackathon. It was a journey of learning, discovery and fulfilling the passion and love of science.
References
Images were taken from Unsplach plugin within Figma.
Videos were taken from pixabay.com & mixkit.com
Video :
https://drive.google.com/file/d/1bknyswSooQ7LzN3eHGowkreOXbW1I73f/view?usp=sharing
Presentation :
https://www.canva.com/design/DAFN4axTyUI/psoKL_NlZ2Mef3JAjfCtsw/view?utm_content=DAFN4axTyUI&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton
Prototype :
https://www.figma.com/proto/0WyvNZ30sPp8tTKM1geC5h/NASA-Space-App-Challenge?page-id=0%3A1&node-id=90%3A428&starting-point-node-id=33%3A617&scaling=scale-down&show-proto-sidebar=1
Tags
#art #ml #ai #nlp #galaxy #nebula #star #earth #planet