Natural Language Processing" (1) Curriculum - Overview and Hands-on Practice

Subject: Natural Language Processing (NLP)

Through my work in implementing AI chatbots and taking the Natural Language Processing (NLP) course, I realized that since the introduction of the "Attention Is All You Need" paper and the breakthrough models like "Transformer", "BERT", and "GPT-2", there has been a significant increase in both corporate and academic interest in dialog systems such as AI chatbots and customer support services, with NLP technologies at their core.

This shift can be attributed to the advancements in NLP technologies, which have reached a point where natural language services are widely accepted without significant resistance.

In this context, I want to explore what Natural Language Processing (NLP) is, why the interest in it has grown rapidly, and what technologies have made this possible. Additionally, I will present the requirements and methods for developing AI services based on NLP technology.

Approach/Goal:

  • Understanding NLP
  • Developing an AI chatbot using BERT
  • Timeline: From September 2020 onwards.




1. Overview

  • AI vs Natural Language Processing
  • SQuAD vs KorQuAD

2. Mathematical Background

3. Basics of Natural Language Processing

  • NLP Steps
  • Morphological Analysis
  • Word Embedding

4. Language Models (LM)

  • Definition of Models
  • Markov Probability Model
  • RNN (Recurrent Neural Network)
  • Attention Models
  • Self-Attention
  • Multi-head Self-Attention Model
  • Multi-head Self-Attention Encoder Model

5. Key Research Papers

  • Background
  • Feed Forward
  • Softmax
  • Positional Encoding
  • BERT vs OpenAI GPT vs ELMo
    • BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
    • OpenAI GPT-1: Improving Language Understanding by Generative Pre-Training
    • ELMo: Deep Contextualized Word Representations

6. Hands-on Practice (Environment Setup and Review of Trial-and-Error)

  • Korean Morphological Analyzer
  • Preprocessing
  • Evaluation of Sequence Labeling
  • Speech-to-Text (STT)
  • Text-to-Speech (TTS)
  • AI Chatbot

This curriculum provides a comprehensive introduction to NLP, its mathematical foundations, key models, and practical applications in areas like AI chatbots and speech processing.


Reference

- 서 적) 텐서플로 2 와 머신러닝으로 시작하는 자연어 처리

- 외부 강의 ) 자연어 언어모델 ‘BERT (Bidirectional Encoder Representations form       Transformer)’ * T Academy & 솔트룩스 강의

- http://hunkim.github.io/ml/ 모두를 위한 머신러닝/딥러닝 강의 - 홍콩과기대 김성훈-

시즌 NLP - Deep NLP

(TBA)

Comments

Post a Comment

Popular posts from this blog

[MaritimeCyberTrend] Relationship and prospects between U.S. Chinese maritime operations and maritime cybersecurity

Image
U.S. Sanctions on Chinese Ships & Cybersecurity Compliance The U.S. Trade Representative (USTR)’s sanctions on Chinese shipping and shipbuilding are expected to heighten the importance of cybersecurity regulations in vessel operations. In particular, as the U.S. increasingly frames Chinese-built ships and shipping companies as cybersecurity risks, compliance with maritime cybersecurity standards will become a critical issue for global shipping stakeholders. The United States is increasingly likely to classify Chinese-built vessels as national security and cybersecurity threats, using this as a basis for additional regulations and sanctions. In particular, drawing from past sanctions on Huawei and ZTE, the U.S. may argue that ships built in Chinese shipyards and equipped with Chinese IT systems (navigation, communication, and monitoring equipment) pose risks to the digital maritime infrastructure of the U.S. and its allies. As a result, the U.S. Coast Guard (USCG) is expected to s...
Read more

인공지능 서비스 - 챗봇, 사전에 충분한 지식을 전달하고 함께 학습 하기!

Image
우리는 앞선 글을 통해 성공적으로  AI 서비스를  개발하거나 도입하는 것을 목적으로  프로젝트의 공식 착수 이전에  목표로 하는 AI 서비스에 대하여, 소속된 조직 내부는 물론 기업내 협력이 요구되는 조직들  각자가 서로 다른 정의와 기대 효과 그리고 활용 방안을 가지고 있던 상황을 해결할 필요가 있는것을  잘 알고 있습니다. 이를 위해 가장 먼저, 이해 관계자들이 그 추진 방향과 목표를 공감 할 수 있도록 잘  알려진 선행 사례와 수치를 활용하여 필요성 측면에서 사전 논쟁을 유도 하였으며  앞으로의 프로젝트 진행 단계 별 추진 방향에 대한 의사 결정의 매 순간  마다 이해 관계자들에게 정확한  정보를 전달 할 것임을 암시적으로 표현 하기도 하였습니다. 이제 남은 것은 관련 분야에 경험이나 지식이 없는 상황에서 잘못된 의사 결정이나 행동을 미연에 방지하기 위해   "사전에  충분한 지식을 전달하고 함께 학습 " 하는  것이라 생각 합니다. (나던 남이던 알면 얼마나 더 알겠는가.. 인터넷 검색만 해도 다 나오는 한 줌도 되지 않는 지식으로 자존심 세우려 하지 말고, 함께 학습 하며 결과물의 완성도를 높이는 대 힘쓰자~! 그것이 상생이다. 꼰대가 되지 말자~!) 이는 본인의 경험을 바탕으로 작성된 글임을 다시 한번 밝히며, 이러한 사항이 AI 업계에 종사 하시는 분들에게도 도움이 되기를 바라는 마음에  그 사례를 소개 하고자 합니다. (관련 전문 용어나 이미지 등은 하단의 reference 들을 참조 하여 작성 하였습니다.) 챗봇의 작동 원리 그리고 최적 성능 확보를 위한 노력   역은이: In-Sung Lee 모든 챗봇이 인간의 언어를 완벽히 이해하고 자연스럽게 대화하는 것을 목표로 가져야 할까요 ? 아마도 사람처럼 말하고 답변 하지 않는 단순 챗봇이라 하더라도 고객의 문제를 해결해 줄 수 있다면 그 자체로 의미가 있...
Read more

[Curriculum] Sungkyunkwan University - Department of Information Security - Course Sequence by Areas of Interest

Image
The order in which subjects are approached may vary depending on the student's major, interests, and learning goals. However, generally, taking courses in the following sequence allows for efficient learning. (Source: Sungkyunkwan University GSIC ) 1. Basic Courses First, it is essential to understand fundamental theories and basic concepts. The following courses help build foundational knowledge: Introduction to Digital Forensics (FSI5056) : Establishes the basics of digital forensics and teaches various methods for collecting and analyzing digital evidence. Korea University Graduate School of Information Security Introduction to Cryptography (GSIS001) : Covers the fundamental principles and applications of encryption technologies for data protection. Life Coding Database (GSID003) : Introduces key concepts and practical skills for data storage and management. Operating Systems (GSID021) : Helps understand the structure and functions of operating systems, which are the core of com...
Read more