Why Natural Language Processing Is the Jewel in the Crown of AI

Can machines understand text like we humans do? This was the initial fantasy of artificial intelligence. Today, it has become the core area of artificial intelligence—Natural Language Processing (NLP). Natural Language Processing is a science that integrates linguistics, computer science, and artificial intelligence, addressing the problem of “enabling machines to understand natural language”—a privilege that so far belongs exclusively to humans. Therefore, it is hailed as the jewel in the crown of artificial intelligence.

Although the term “Natural Language Processing” is not as familiar as big data or machine learning, we use or benefit from it every day. A typical example is Xiaoice, a chatbot on China’s Weibo that brings conversation into our daily lives. Millions of young Chinese users exchange information through Xiaoice, often chatting with it when they break up, lose a job, or feel depressed. So far, Xiaoice has covered three languages: Chinese, Japanese, and English, accumulating over a hundred million users, averaging 23 rounds of conversation, with an average chat duration of about 25 minutes.

The chatbot Xiaoice is just a glimpse of the applications of Natural Language Processing. Here are some broader examples of the use of Natural Language Processing technologies:

Last autumn, Google Translate launched a newly upgraded AI translation engine. As a result, Google Translate, which was once known for producing awkward but usable translations, has begun to produce fluent and highly accurate translated texts. For those without professional translation training, this text output is almost indistinguishable from human translation. For example, the translation of the current passage is shown in Figure 1.

Figure 1 Google Translate Illustration

In applications such as automatic spam detection, there are only two classifications: spam and non-spam. In other cases, classifiers can have multiple categories, such as organizing news reports by topic or organizing academic papers by field. But what if a blog post discusses both sports and entertainment? How does a classifier choose the correct category among multiple options? That depends on the specific application: it can simply choose the most likely option, but sometimes it makes sense to assign multiple categories to a text.

From the birth of Siri in 2011 to Google Now, Cortana, and Alexa, these voice assistants are essentially question-answering systems. They are all publicly-oriented question-answering systems that help us in our daily lives by setting alarms, making calls, navigating, searching for questions, and occasionally telling jokes, making our lives increasingly convenient.

Figure 2 Apple Siri Illustration

Especially after 2010, the application of deep learning in the field of Natural Language Processing has gradually introduced a series of product features into our lives. Major companies are also actively laying out related industries, investing heavily to recruit talent in this field. In China, three publicly listed companies have emerged in the field of language information processing, in the order of their listing: Hanwang, specializing in pattern recognition, followed by iFlytek, focusing on speech recognition, and then Tuoshu, specializing in information retrieval and text mining.

Figure 3 Job Posting from a Recruitment Website

The Deep Blue Academy, focusing on cutting-edge technology online education, has partnered with Professor Zhou, a leading author from international top conferences, to offer two online live courses: “Introduction to Natural Language Processing and Algorithm Practice” and “Practical NLP Based on Deep Learning”. Outstanding students from these courses can be directly recommended for internships and employment at well-known companies such as Baidu, Sogou, and Toutiao.

Professor Zhou, a master’s supervisor, received his PhD from the Institute of Automation, Chinese Academy of Sciences, primarily engaged in research on Natural Language Processing and Deep Learning. He has published over 20 papers in international journals and at top academic conferences such as ACL and has won the Best Paper Award at international conferences twice. He is currently undertaking more than ten projects, including the National Natural Science Foundation and 973 sub-projects.

1. Top conference authors as speakers, insight into technology trends;

2. Theory combined with practice, foundational courses paired with intensive courses;

3. Online live Q&A during class, group chat Q&A after class;

4. Certificates awarded to outstanding students, recommendations for internships at top companies;

5. Course PPT and source code will be made available to students in advance.

Part I: Basics of NLP and Algorithm Practice (10 hours)

1. Syntactic and Semantic Analysis (2 hours)

1.1 Dependency Parsing

1.2 Semantic Role Labeling

1.3 Relevant Datasets and Tools Introduction

2. Opinion Mining and Sentiment Analysis (2 hours)

2.1 Sentence-Level Sentiment Analysis

2.2 Document-Level Sentiment Analysis

2.3 Cross-Language Sentiment Analysis

2.4 Cross-Domain Sentiment Analysis

2.5 Relevant Datasets and Tools Introduction

3. Information Extraction: Part 1 (2 hours)

3.1 Named Entity Recognition and Extraction

3.2 Entity Disambiguation

3.3 Relevant Datasets and Tools Introduction

4. Information Extraction: Part 2 (2 hours)

4.1 Entity Relation Extraction

4.2 Event Extraction

4.3 Relevant Datasets and Tools Introduction

5. Question-Answering Systems (2 hours)

5.1 Retrieval-Based Question Answering

5.2 Community Question Answering

5.3 Knowledge Base Question Answering

5.4 Relevant Datasets and Tools Introduction

Part II: Practical NLP Based on Deep Learning (24 hours)

6. Lexical Analysis Based on Deep Learning (4 hours)

6.1 Chinese Word Segmentation Based on Deep Learning

6.2 Part-of-Speech Tagging Based on Deep Learning

6.3 Named Entity Recognition Based on Deep Learning

6.4 Code Module Demonstration, Common Tools, and Public Datasets

7. Syntactic and Semantic Analysis Based on Deep Learning (4 hours)

7.1 Graph-Based Dependency Parsing

7.2 Transition-Based Dependency Parsing

7.3 Shallow Semantic Role Labeling

7.4 Code Module Demonstration, Common Tools, and Public Datasets

8. Sentiment Analysis Based on Deep Learning (4 hours)

8.1 Building Sentiment Lexicons Based on Deep Learning

8.2 Sentence-Level Sentiment Analysis Based on Deep Learning

8.3 Document-Level Sentiment Analysis Based on Deep Learning

8.4 Cross-Language Sentiment Analysis Based on Deep Learning

8.5 Code Module Demonstration, Common Tools, and Public Datasets

9. Information Extraction Based on Deep Learning: Part 1 (4 hours)

9.1 Entity Relation Extraction Based on Deep Learning

9.2 Entity Disambiguation Based on Deep Learning

9.3 Demonstration of Representative System Modules, Common Tools, and Public Datasets

10. Information Extraction Based on Deep Learning: Part 2 (4 hours)

10.1 Event Extraction Based on Deep Learning

10.2 Knowledge Base Representation Based on Deep Learning

10.3 Knowledge Base Completion Based on Deep Learning

10.4 Code Module Demonstration, Common Tools, and Public Datasets

11. Question-Answering Systems Based on Deep Learning (4 hours)

11.1 Community Question Answering Based on Deep Learning

11.2 Complex Question Parsing Based on Deep Learning

11.3 Knowledge Base Question Answering Based on Deep Learning

11.4 Code Module Demonstration, Common Tools, and Public Datasets

Register now to receive a 150 RMB coupon. Classes start on January 6, with online live sessions every Saturday and Sunday from 7 PM to 9 PM, with unlimited online replays available for one year.

Please add the staff member “Deep Blue Academy“Teaching Assistant to register.