跟讀練習: AI agents explained: Chatbots vs. workflows vs. agents - 通過影片學習英語口語

選擇練習模式

每種模式都會開啟全屏練習室。

28 句
如果句子過短或過長,請點選 Edit 進行調整。
1
Let's talk about three layers of memory that your AI agent has.
2
Those layers are state, sessions, and long -term memory.
3
So let's use the travel planning agent as an example.
4
First, state.
5
You can think of state as the agent's working memory.
6
It keeps information the agent needs right now, like remaining budget for your trip or current travel plan.
7
It is like a scratch note that you can put important information on it
8
and pass it among different components in your system.
9
And as the plan changes, the state gets updated too
10
and second sessions a session keeps context of one conversation thread
11
it includes everything the conversation transcript like messages two costs two results
12
and state a session service manages saving
13
and loading the information you could use an in -memory backend for example run for quick
14
and easy context or you can add a persistent database so the session can survive server restarts and continue later.
15
This is really important in the long -running process.
16
Third, long -term memory.
17
Long -term memory carries useful information across conversations.
18
In ADK, we have session service to manage all the sessions and memory service to manage the memory across the sessions.
19
To use them in this trip example, you could tell the agent that you prefer quiet neighborhoods.
20
A session service can store all the conversations you ever talk to the agent.
21
A memory service source and retrieves that information
22
and then brings relevant memories into the context so you can plan it next trip.
23
So memory service can help us manage cross -session context.
24
To summarize, each layer has a different job.
25
State is what the agent is working with right now.
26
Sessions preserve the current conversation.
27
A long -term memory carries useful information into the future.
28
And now your AI is context -aware and ready for production.

下載應用

流利口語所需的一切

AI 發音評分逐句打分
音標練習掌握每個音
詞彙學習建立你的詞庫
詞彙遊戲邊玩邊學

同頻道

FOR YOU

推薦影片

本課的詞彙與口語要點

這節口語課以影片“AI agents explained: Chatbots vs. workflows vs. agents”為素材。 影片中反覆出現的詞有:session, memory, agent, context, conversation。 這段影片共有 28 個句子、310 個單詞可供跟讀。 講話部分時長為 1:54。 說話人語速自然,每分鐘約 163 個詞,接近日常對話。 88% 的單詞屬於英語最常用的 3,000 詞,詞彙容易理解。

影片中的重點詞彙

影片中 10 個值得學習的單詞,附發音和釋義:

單詞發音釋義
layer 名詞/ˈleɪ̯ɚ/層 /層
server 名詞/ˈsɝvɚ/伺服器 /伺服器, 伺服器
thread 名詞/ˈθɹɛd/線 /線
component 名詞/kəmˈpoʊ.nənt/元件, 部件
database 名詞/ˈdeɪtəˌbeɪs/資料庫 /資料庫
scratch 動詞/skɹæt͡ʃ/抓, 劃
persistent 形容詞/pəˈsɪstənt/持續的 /持續的, 固執的 /固執的
transcript 名詞/ˈtɹæn.skɹɪpt/副本, 抄本
retrieve 動詞/ɹɪˈtɹiːv/重拾
summarize 動詞/ˈsʌməˌɹaɪz/摘要, 總結 /總結

需要注意的發音

中文母語者容易讀錯的音:

  • /v/ — 不要讀成 /w/,上齒輕觸下唇: server /ˈsɝvɚ/, preserve /pɹəˈzɝv/, retrieve /ɹɪˈtɹiːv/
  • 詞尾子音 — 要讀清楚,後面不要加母音: thread /ˈθɹɛd/, component /kəmˈpoʊ.nənt/, database /ˈdeɪtəˌbeɪs/, preserve /pɹəˈzɝv/, scratch /skɹæt͡ʃ/
  • /æ/ — 嘴張大,不要讀成 /e/: scratch /skɹæt͡ʃ/, transcript /ˈtɹæn.skɹɪpt/

如何用這段影片練習

  1. 先完整聽一遍影片,不要開口,記下不認識的單詞。
  2. 先用 0.75 倍速逐句跟讀,熟練之後再回到正常速度。
  3. 錄下自己的聲音並與原聲對比,特別注意 layer, server, thread 這類單詞。

什麼是跟讀法?

跟讀法 (Shadowing) 是一種有科學依據的語言學習技巧,最初開發用於專業口譯員的培訓,並由多語言者Alexander Arguelles博士普及。這個方法簡單而強大:您在聽英語母語原聲的同時立即大聲重複——就像是一個延遲1-2秒緊跟說話者的影子。與被動聽力或語法練習不同,跟讀法強迫您的大腦和口腔肌肉同時處理並模模擬實的講話模式。研究表明它能顯著提高發音準確性,語調,節奏,連讀,聽力理解和口語流利度——使其成為雅思口語備考和真實英語交流最有效的方法之一。

影子跟讀法: 閱讀完整分步指南 →