How Social Mining Works?｜社交挖礦影響要素

Ｗe appreciate all the support from the Lit users, and we understand a guide to a better-social-mining result is desired by all the users.

The ever-improving algorithm is more than a set of mathematical formulas, so it will remain undisclosed.

That being said, we are introducing to you the concept behind the Lit social mining system to help you enjoy social mining more.

Basically, all stories are scored and converted to the number of MITH tokens rewarded.

The exact number of MITH tokens rewarded from social mining is determined by various factors,

Quality Score

Quality Score is related to the quality of contents and other factors.

Accumulated Engagement Score

Accumulated Engagement Score is based on the interaction between a creator and the viewers.

Account Rating

The Account Rating of a user is largely determined by the user’s activity on the Lit platform. It affects the number of tokens rewarded. Also, users rated low will have a lower Accumulated Engagement Score.

Quality Score

What gives your content higher quality score?

１．Unique photos:

Image recognition technology is used for Lit Social Mining and the model reads photos. The more special and unique your photos are, the more diversity they create for Lit, then the higher quality score they have.

２．Dynamic engagement:

Engagement metrics are calculated based on the number of views, likes and messages for each story. When the engagement reaches a specific level, the Quality Score of the story is increased.

３．Invested time:

When viewers spend more time on a specific content creator’s story than other stories, it shows that the viewers are more delighted to read that story, and its Quality Score increases.

What lowers the quality score of your content?

１．Graphic, violent, sexual, inappropriate content; content that is not created by the creator (stolen photos); duplicated content created by the creator or other creators:

To encourage the creation of high-quality content, all the above will not be categorized as “valuable creation” and given a low Quality Score. Account will be blocked for posting nudity.

２．Photo quality:

Blurry photos have a lower quality score.

Accumulated Engagement Score

How to increase accumulated engagement score?

１．Invested time:

When viewers spend more of their time on a specific creator, they are considered as “loyal fans” of this creator. Engagement from loyal fans are more valuable than engagement from regular viewers.

２．Genuine likes:

When a viewer gives fewer likes to all creators, it means the viewer is genuine and does not like everything. The engagement therefore is more valuable than the engagement of other viewers.

What lowers the accumulated engagement score?

１．When a viewer follows a creator but does not watch his/her content for a recorded time:

Engagement from “disloyal fans” are less valuable than engagement from regular viewers.

２．When a viewer watches or likes too fast:

Lit encourages the creation of valuable contents and encourages viewers to watch the contents and give genuine likes. As such, when a viewer watches or likes too fast, the engagement has lower value.

３．Suspicious reciprocal engagement:

It has come to the attention of the Lit team that some users like each other’s content reciprocally. When it is detected by the system, the value is lowered.

４．Multiple accounts:

The Lit team has found that some users create multiple accounts, upload duplicate contents, and sham engagement between accounts. When it is detected by the system, the value is lowered.

Account Rating

What improves your rating?

１．Growth in the number of fans

Fans are defined as users who view each new content of a specific creator. The growth in the number of fans means the content of the creator is recognized, and the rating of the creator will be improved.

What lowers your rating?

１．Spam, repeated and similar messages to different users:

This is considered as inappropriate advertisement and harassment, and the rating of the account falls.

２．Send friend requests to inactive users:

Lit, as a social platform, expects real and authentic interaction among users. Sending follow requests to inactive users (new users excluded) is considered as abnormal usage, and the rating of the account drops.

３．Spend too little time on each story and like stories too frequently:

The Lit team deploys various standards for skimming through stories and liking each story too often. When the frequency reaches a level, it is labeled as zombie likes and abuse of the platform. The account is blocked automatically.

４．Story reported by other users and the red flag is confirmed:

The rating of the creator falls when a red flag is confirmed.

社交挖礦影響要素

一直以來我們非常感謝所有使用者對 Lit 的支持，也知道關於 Lit 社交挖礦數量的計算方式一直是使用者很希望了解的一塊。

由於演算法不斷在調整，同時有一些並不是純粹的「數字公式」，因此沒辦法公布最精確的算法給大家。不過這邊還是整理了最新的 Lit 社交挖礦評估元素，希望能夠幫助使用者們更順利地進行社交挖礦。

基本上每一則使用者發布的動態，都會透過公式算出基本分數，然後根據此分數轉換成對應的秘銀數。

然而，實際挖掘到的秘銀幣數量還是會受到諸多變因影響，包含三大項：

貼文品質分數

與用戶的貼文的品質、及其相關因素有關。

互動累積分數

互動累積分數來自使用者獲得來自觀看者的互動。

用戶評級

用戶評級分數則與使用者本人的Lit使用方式有關，用戶評級分數會影響到使用者貼文時的實際挖礦數量，此外，用戶評級低的用戶給予的互動累積分數也會較低。

貼文品質分數

會提升貼文品質分數的要素

１．照片獨特性：

Lit的社交挖礦有使用到圖像辨識技術，其模型會辨別動態中的物件，物件的獨特性越高、增加了 Lit 內容多元性的貼文，貼文品質分數越高。

２．動態的互動數：

每則動態的觀看數、按讚數、訊息數都會被記為互動數，互動次數達到一定程度會提升貼文品質分數。

３．觀看者花特別多時間在作者的特定貼文：

觀看者觀看作者某貼文的時間，高過於其平均觀看其他貼文的時間，代表觀看者更樂於觀看此貼文，此則貼文品質分數提升。

會降低貼文品質分數的要素

１．血腥暴力/色情/會使觀看者不適的內容、上傳非自身拍攝之內容（盜圖）、與其他使用者或自己過去所上傳內容完全重複：

為鼓勵優質的內容創作，以上內容會被歸類為不符合「有價值的創作」，貼文品質分數低落。其中，若有露點情形，會直接封鎖帳號。

２．照片品質：

若照片模糊看不清物件，則貼文品質分數較低。

互動累積分數

會提升互動累積分數的要素

１．觀看者花特別多時間在作者的帳號：

觀看者在使用 Lit 的時間中，花費在某作者的比例越高，代表其為此作者的「忠實粉絲」；來自忠實粉絲的觀看的互動累積分數高於來自一般觀看者的互動累積分數。

２．觀看者的真心按讚

若觀看者使用期間給予所有作者的按讚數總和較少，代表此為真心按讚，而非看到什麼都按讚，則互動累積分數高於來自一般按讚者的互動累積分數。

會降低互動累積分數的要素

１．觀看者追蹤了作者卻長期未觀看其貼文：

來自「非忠實粉絲」的互動累積分數低於來自一般使用者的互動累積分數。

２．觀看者瀏覽過快、按讚速度過快：

Lit 鼓勵有價值內容的創作，也因此鼓勵使用者認真觀看內容並發自內心按讚。因此，若瀏覽速度/按讚速度過快，會降低此觀看者給予的互動累積分數。

３．關係過於緊密的群體互相刷互動：

Lit 團隊有注意到許多使用者之間會互相無差別按讚，若被系統辨識到，互動累積分數會較低。

４．不鼓勵一人擁有多重帳號互贊互粉：

Lit 團隊有注意到使用者一人創立多個帳號，並上傳相同內容，帳號之間互相刷互動，若被系統辨識到，互動累積分數降低。

用戶評級

會提升用戶評級的要素

１．粉絲人數成長：

「粉絲」為每次作者上傳動態就會來觀看的用戶，使用者的粉絲人數成長代表著使用者的內容被肯定，因此能夠提升使用者的用戶評級。

會降低用戶評級的要素

１．大量傳送重複且類似的訊息給予不同用戶。

我們視其為不當廣告宣傳、騷擾行為，會降低用戶評級。

２．大量傳送交友邀請給予沒有活動的用戶：

Lit 為社交平台，我們希望使用者的互動都是真實有意義的，若是大量傳送交友邀請給予沒有活動的用戶（新創建帳號的用戶不算），行為視同非正常使用，同樣會降低用戶評級。

３．觀看每則動態時間過短，並頻繁按讚每一篇動態：

關於快速觀看以及頻繁按讚每一篇動態，Lit團隊有不同等級的標準，若頻率高到一定程度，會被判別為使用機器人刷讚，為不當使用行為，會自動封鎖該帳戶。

然而，即使頻率沒有高到被判斷為機器人，只要頻率高到一定程度，仍會判定該用戶「手動刷讚」，而Lit官方做為鼓勵「創作有價」的平台，我們希望讓高品質的內容被看見、獲得獎勵，也因此無差別性地手動刷讚並非官方鼓勵之行為，會降低用戶評級。

４．被其他用戶檢舉動態，並經查證屬實：

被檢舉屬實會降低用戶評級。