在過去，毒品濫用一直是一個嚴重的議題。吸毒以及成癮的比例不斷地在上升，根據研究報告指出，年輕人會去接觸毒品的動機主要是因為好奇與同儕的影響。透過分析這些人想 要接觸毒 品的意圖有助於防止毒品的濫用。我們認為 從使用者的輸入的問句中 可以 檢測出潛在的風險並獲得警示，並提供給專業的諮商師進行評估。此外，強化年輕族群對於毒品的認識也是另一個有效阻止毒品氾濫的方法。知識庫問答是近年來流行的一種方法，可以從知識庫中提取準確而簡潔的回答。 在本文中，我們提出了一個基於Bert 的 行為認知的反毒聊天機器人， 透過分層分類器架構來分析使用者的輸入，從而對可能吸毒和對毒品感到好奇的人們進行風險評估並交由專家進行後續追蹤。此外， 利用 新聞以及反毒網站爬蟲 作為 語料 能夠根據問題檢索以及知識庫回答有關毒 品的問題以加深人們對毒品的認識。對現有的反毒題庫的實驗結果表明，引入知識庫問答具有較強的效能。;In the past, drug abuse has always been a serious issue. The proportion of drug abuse and addiction keep rising. According to research, the motivation of the youth to access drugs is due to curiosity and peer influence. Analyzing the intention of these people to access drugs can help to prevent drug abuse. We believe that potential risks can be detected and warnings can be obtained from the user′s input questions, and provided to professional consultants for evaluation. In addition, deepening the young people’s awareness of drugs is another effective way to stop the spread of drugs. Knowledge base question answering is a popular method, which can reply accurate and concise answers from the knowledge base. In this paper, we propose an anti-drug cognitive behavioral chatbot with Bert, which analyzes the user’s input through a hierarchical classifier, so as to conduct a risk evaluation for people who may take drugs and who are curious about drugs and hand them over to experts trace. In addition, using news and anti-drug website crawlers as corpus can answer drug-related questions based on question retrieval and knowledge bases to deepen people′s understanding of drugs. The experimental results of the existing anti-drug question bank show that the introduction of knowledge base question answering has a strong effect.
