stanford-NLP-CLASS1课堂笔记

CLASS1 — NLP INTRODUCTION SUMMARY

Applications of NLP

1. Information Extraction 信息抽取
2. Information Extraction & Sentiment Analysis 信息抽取与情感分析
3. Machine Translation

---

Three kind of language technology

A. mostly solved

1. spam detection
2. part-of-speech(POS) tagging词性标签
3. named entity recognition(NER)

B. making good progress

1. sentiment analysis
2. coreference resolution
3. word sense disambiguous词义消歧(WSD)

C. still really hard

1. question answering(QA)
2. paraphrase反义句
3. summarization
4. dialog

---

What makes NLP hard?

Ambiguity — crash blossoms

---

Why else is NL understanding difficult?

1. non-standard English
2. segmentation issue
3. idioms
4. neologisms新词
5. world knowledge
6. tricky entity names

---

What tools do we need?

1. knowledge about language
2. knowledge about the world
3. a way to combine knowledge sources

---

How we generally do this?

• probabilities models built from language data
• rough text features can often do half the job.