The project aims to design and develop a TSN (Tailored Social Network) platform that offers advanced functionalities beyond traditional platforms like Facebook. The platform must be based on a graph structure, inspired by the FOAF (Friend of a Friend) Ontology, enabling users to connect, interact, and share content in a more personalized and intuitive manner.
We recently enabled GA (General Availability) of our project, making it accessible online, by everyone. You can see it here : https://transmit-project.vercel.app/about
Run pnpm install to get started and install all dependencies.
If you don't have pnpm installed, simply run npm install -g pnpm.
Run pnpm dev for a dev server ; then, navigate to http://localhost:4200/. The application will automatically reload if you change any of the source files. You don't need to run anything else in order to get started ; our project completely relies on on-device-computing making it backendless, and the database is hosted online.
Posts are ranked based on User behaviour, User friendship, User tags and date. For more information : download the algorithm_explanation.pdf.
Yes, networking is cool, but a network is not here just to exist. It is here for you to share content with your relations. A social network without the ability to share content is useless, thus Transmit allows you to share your opinion, eventually joining an image to it.
What is interesting with humans is that everybody has different opinions. And from these differences arises interesting conversations. However conversations need an appropriate support to exist, which we believed to be a comment section under each main post. Comments are also updated in realtime.
As writing a "i love this content" comment for each post you like is not the best user experience, we chose a seamless approach by implement a small "Like" button to show your love for the content you appreciate.
The force graph is the component displaying the kinship between you and your followers. Each people is represented by a node; the lighter the node is, the further this person is from you.
The graph displayed is powered by the d3 library ; we chose to use the force graph because it's perfect to represent a network with nodes and directed links. The force-graph component is solely used for rendering and is not doing any logic ; it's all within the foaf-service.
Algorithm
The algorithm is an simple implementation of Breadth First Search (BFS).
We start by adding the user connected (UC) to the queue and then iterating through it. For each user in the queue, we fetch their friends and add them inside. To avoid infinite loops, we keep track of fetched users and do not add them to the queue if we have already fetched their friends.
Data Structure
To store the relationship between people, we chose an adjacency list due to its numerous advantages.
- Perfectly suited to represent directional & unweighted graph
- Enables fast lookup to retrieve friends of someone
- Efficient storage as only IDs are stored
When someone is fetched, their name is added to the adjacency list as a key, and all their friends are associated as the value.
Depth Control
If we complete the BFS entirely, it will result in fetching all the network in local, which is very time-consuming and not private at all.
Therefore, to keep track of the depth of each user, we employ the following algorithm: when we initiate the first iteration of the loop with the user connected, we add it to the adjacency list with an attribute called "depth" initialized at 0. Consequently, all individuals fetched have a depth equal to their friend's depth, plus 1, stored in the adjacency list.
With this system, we can specify a maximum depth as an argument and exclude friends of a user that exceed the maximum depth, resulting in a stack dwindling to nothing.
The default depth for network visualization is 2 and the maximum is 5 limiting network visibility and keeping the graph view decluttered. However, friends and post recommendation features can search beyond this limit.
Complexity
Thanks to our depth control feature, a user is virtually placed in a sub-network of depth 5, making it finite. Therefore, the complexity of our system depends on the number of users within that sub-network and the number of relationships they have.
The time complexity of our BFS algorithm is O(U + L) with N the number of users and R the number of links.
For the adjacency list, we store all the friends as key with a list of their friends as value. The space complexity is O(U+F) with U the number of fetched user and F the number of friends
The side peek view is a panel that displays the last user we clicked on. A request is made to get user data based on its id and then displayed in a friendly way. However, the number of followers displayed is not accurate. It is relative to the current depth of the graph. For instance, if the graph is set with a depth of 1, we only have our friends, so our number of followers is zero. But the bigger the depth becomes, the more accurate the number of followers is.
The user recommendation is done using 2 methods : recommendation based on tags and based on friends of friends. For more information : download the algorithm_explanation.pdf.
We provide a basic interface to modify user information. The forms are automatically updated when the user click outsides, meaning the data is always up to date.
Relations are an essential concept in Transmit. This is why having an easy access to manage them is required. Friend breakups may appear in many occasions, so removing a friendship link should be done quickly, as well as removing followers that are not wanted anymore.
Ideas an opinions come and go. You might say something one day and regret saying it a few days later. Same goes for posts published online, sometimes you do not want some content to be visible anymore. Thus, a content management panel was required to review ones own posts and delete them if the wil is present.