SCKR-Course

Peking University Semantic Computation and Knowledge Retrieval Course Project

WordSimilarity

Calculate the word similarity using different methods

• data

  MTURK-771.csv: Data set with the groud truth.

  text8: The corpus from Wikipedia for training the Word2vec.(Please add glove.6B.300d.txt in dataset youself.)

• result

  corpus: The result and model of word2vec.

  web_search: The results of jaccard, overalap, pmi and dice.

  wordnet: The results of path, wup, lch, res, lin and jcn.

• word_similarity.py: Codes.

• report.pdf: The report of Course Project.

SentimentClassification

Tweet Sentiment Classification (SemEval2017 Task 4 Subtask A)

• data

  glove.6B.300d：Pre-trained word vectors (dimension = 300). (Please add glove.6B.300d.txt in dataset youself.)

  twitter-2016train-A/twitter-2016dev-A/twitter-2016test-A: Tweets, divided into training set, valid set and test set.

• code: Codes.

• report.pdf: The report of Course Project.

DBQA

Document-based Question Answering task (DBQA)

• data

  hanlp-wiki-vec-zh.txt: Pre-trained word vectors (dimension = 300). (Please add hanlp-wiki-vec-zh.txt in dataset youself.)

  stop_words.txt: Chinese stop words.

  nlpcc-iccpol-2016.dbqa.training-data/nlpcc-iccpol-2016.dbqa.testing-data/test.txt: NLPCC2017DBQA data, divided into training set, valid set and test set.

• code

  DBQA_CNN&Attention1: CNN with Attention Model 1.

  DBQA_CNN&Attention2: CNN with Attention Model 2.

• report.pdf: The report of Course Project.