唔好侮辱維尼 2019-2-13 16:46:19

:^(

i-vtec找J/屎眼數 2019-2-13 17:06:22

:^(

讓位專員 2019-2-13 17:07:30

:^(

唔好侮辱維尼 2019-2-13 17:09:36

:^(

答沉船 2019-2-13 17:40:15 IT狗思考方法

:^(

:^(

:^(

答沉船 2019-2-13 17:42:10 一係BA/PM

一係 network

一係做看更

心碎 2019-2-13 17:48:24 做看更啦
反正做IT到中年都係做看更

室見立華 2019-2-13 18:05:18

:^(

等投胎窮三代IT狗 2019-2-13 18:05:47 打坐好工

:^(

食屎的過程 2019-2-13 18:11:36 打code 打多幾年都係做看更
不如早啲做看更拎經驗
之後升做高級看更搵多啲

:^(

答沉船 2019-2-13 18:14:04 IFC樓下個D看更要識好多language

:^(

:^(

:^(

i-vtec 2019-2-13 18:42:45 做幾年IT邊儲錢邊買收息股

:^(

頭幾年IT最值錢

:^(

跟著去做看更仲正呀 做果D中學通宵夜更保安
自己一個守更亭又無危險
又唔洗煩公司人事問題又唔洗用腦
一邊睇戲打機上連登, 睇到悶悶地訓下教

:^(

:^(

:^(

真正退休生活

讓位專員 2019-2-13 18:47:57 係咪尖吵咀個間

青花瓷 2019-2-13 19:26:20 召喚強手

假設有10000句phase, eg "coffee workshop", "cat owner meeting" 要搵相似既phase做cluster

Step 1: English normalisation, 將meeting, met 變做meet. ..

Step2: for each unique normalized word, get a dictionary of occurrence. "meet":6％, "coffee" 0.5％, workshop 5%....

‌Step3: build a co-exist-probability matrix for the phases.
"Cat ownership talk" vs "cat owner meeting" = prob of "cat" * prob of "own"
..........

Step4: build a graph, each phase is a node, if the co-exist-probability of two node is smaller than x％， build an edge between the two node.

Step 5: use graph algorithm to find all cluster.


會唔會做多左，定係已經有library做緊呢D

:^(

:^(

:^(

心碎 2019-2-13 19:31:05 可能已經有人做左🤔你入返你d training data 就okay
搵下有冇open source

青花瓷 2019-2-13 19:51:15 我堆text 好多specific noun , 驚用library太loose

:^(

有冇人知build 個matrix 係咪一定要 o(n^2）

逆水行舟 2019-2-13 20:09:56 睇你能力

實驗羊 2019-2-13 20:13:41 想做Quant Dev搵真銀

求教路

青花瓷 2019-2-13 20:24:13 buzz word, 入假quant扮真quant

實驗羊 2019-2-13 20:25:34 你同手黏好似就係Quant Dev / Data Engine