NLP1 - Intro

주재걸 교수님

NLP processing(이걸로 수업할 예정)

• 학회 : ACL, EMNLP, NAACL

• Low level parsing

  Tokenization, stemming

• Word and phrase level

  NER(Named entity recognition), POS(part-of-speech) tagging, noun-phrase chunking, dependency parsing, coreference resolution

• Sentence level

  Sentiment analysis, machine translation

• Multi-sentence and paragraph level

  Entailment prediction, question answering, dialog systems, summarization

Text mining

• 학회 : KDD, The WebConf(formerly,WWW), WSDM, CIKM, ICWSM

• 텍스트 데이터로부터 information, insight 추출할때

• Document clustering

• computational social science에도 관련됨

Information retrieval

• 학회 : SIGIR, WSDM, CIKM, RecSys

• 정보 검색 분야

Trends of NLP

• Word2Vec : word embedding
• RNN-family models : LSTMs, GRUs

• Attentions & transformers

• self-supervised training setting (no additional labels)

Bag-of-Words

1) Constructing the vocab containing unique words

2) Encoding unique words to one-hot vectors

ex)

NaiveBayes Classifier

특정 문서 d 가 속할 확률이 가장 높은 C

= c가 고정이 되었을 때 d가 나올 확률

각 단어들의 확률을 구할 때는 MLE로 부터 도출이 됨