Content Curation

The curation classification and processes

Overview

Content curation is a process where relevant content gets acquired, selected, and distributed for users to deep read.

We understand that users’ attention is valuable, not just to ourselves, but to the entire network to perpetuate knowledge, culture, and foster meaningful connections. And the value can only be maximized if our time is well spent on high-quality content.

Hence, we want to leverage crowds’ and experts’ wisdom to curate content for all readers, use this process to surface relevant content, and recognize its contribution of it to the generation and distribution of Time Points.

There are 3 types of content curation classifications on t2:

• Algorithmic curation Algorithmic curation is the type of curation that's performed automatically with algorithm.

• Permissionless curation Permissionless curation is the type of curation in which territory members perform with pre-set rules.

• Open curation Open curation is the type of curation where every user can perform after reading a piece of content, without the need of pre-qualification.

In addition, both permissionless curation and open curation are done purely with human effort involved, hence are also classified as human curation.

A preview of the architecture of the curation design can be seen below. L1, L2, L3 stand for Layer 1, Layer 2 and Layer 3 respectively. A higher level indicates a more granular classification.

Algorithmic Curation

Algorithmic Curation is the curation which is achieved through recommendation and ranking algorithm of T2.

We leverage user's Social Graph and consuming preferences to curate the content on users' feeds.

The goal of this curation is to balance the relevance and discovery of a user's feed, where users can easily find the content that suits his/ her preference most, as well as explore the content that he/ she would be interested in.

The result of the algorithmic curation are displayed on the home page of a user, separated by 3 feeds:

• Following: Posts from followed territories, writers, users

• Discovery: Recommended world-wide posts based on our algorithm

• *Perspective: Customizable feel with parameters set by the user (Future Roadmap)

Each feed consists of the same types of content but in different scopes and orders. The basic element of a feed is a post, with total 4 types of posts:

• Writers’ content publications

• Territory’s content listing updates

• Users’ highlights/comments/reactions/sharing to a content

• Users’ open curation of a content

Rewards Mechanism

For each minute of deep reading on the content curated by algorithmic curation, there will be 1 Time Point generated to award the reader, $\alpha_1,\alpha_2,\alpha_3$ Time Points generated for the related writers, curators, and propagators.

Depending on the type of posts, the variables are set differently. The bottom line is to reward the reader, writer, and last-touch inspirators (curators or propagators).

Permissionless Curation

Permissionless curation is the type of curation where curation is achieved through hard work of t2 community members, with a pre-set of rules.

Depending on the participants and the rules, we further can classify them as collective curation and whitelisted curation. Collective curations involve multiple territory content curators while whitelisted curations only involve one whitelisted curators.

Content that passed the permissionless curation are shown on territory page public facing all readers.

Collective Curation (through Voting and Ranking)

In the collective curation, writers can submit content proposals for territories or sub-territories, which are then voted on by curators. The top-voted content is then listed on the territory, and if found to be incorrect or unsafe, can be delisted through a counter-proposal and get removed.

• Voting and Ranking Process

  • A writer can submit a proposal at time $t$

  • From time $t$to time $t$ + 72 hours, the content would be listed if all of the following conditions are met:

    • There are no less than $x\%$ of total votes casted, we tentatively set $x$ to be 20.

    • Among all the votes casted, 66% or more votes are in favor of the content.

If any of the conditions are not met at the time $t$ + 72 hours, the content got rejected.

• Proposal Frequency, Voting Power and Voting Frequency

Users’ proposal power, voting power and voting frequency are determined by their level, which is dictated by amount of time points they own.

• Rewards Mechanism

Curators are rewarded based on their effective curation, taking into account both frequency of curation and successful rate of curation. Recall that for each minute of deep reading, there will be 0.2 time points generated for territory treasury, from which $t$ time points ($t \leq 0.2$) are used to reward curators. For a curator $j$, he/ she can assign certain voting power on proposal $i$ , namely, $w_i$. And the outcome of the proposal, $D_i$ is 1 if the proposal is successful and 0 other wise. For this curator, his/ her voting effectiveness $e_j=\sum_{i\in I}D_iw_i$, and the time points reward this curator gets will be $t_j=t *\frac{e_j}{\sum_{j \in J}e_j}$

Whitelisted Curation

In the whitelisted curation, a whitelisted curator can directly list a content he/ she deem as high-quality content.

• Rewards mechanism

A whitelisted curator contributes by directly listing high-quality content which users can deep-read, which is similar to the other curation processes. Therefore, a whitelisted curator can be rewarded with all the TP that were intended to be used for curator reward ( $tp_{curation =}tp_{deep\_read}*0.2^-$), multiplied with a multiplier $m$, where $m$ is tentatively set to be 0.7.

There are two types of content curations within a territory: curation through voting and ranking, and permission-less curation by whitelisted curators.

Open Curation

Open curation is the curation in which everyone can improvise curating their own collections. After users read a content, they can choose to curate the content, adding the content to one of their collections on their profile page.

Rewards Mechanism

For open curation, for each minute of deep reading, we generated 1 time point for reader, $\alpha_1,\alpha_2$ time points for writers and the curators respectively. The variables are subject to our design but are tentatively set to be 0.2 and 0.1.