iofu728 / Task

🤧Some tasks for competition by gunjianpan

🤧Some tasks for competition by gunjianpan

❱ ❱❱ ❱❱❱ ❱❱❱❱ ❱❱❱❱ ❱❱❱❱❱ ❱❱❱❱❱❱ ❱❱❱❱❱❱❱

Navigation

Name	Classify	Type	Method	Modify
NLP senior Task3: Chinese traditional Sequence Annotation	Sequence annotation	Classify	BiLSTM-CRF, Bert-CRF	190626
DataMining Project	Semantic Representation	Classify	LSTM-ATT	190620
NLP cbb Task2: Medicine Corpus processing with Corpus	Sequence annotation	Classify	BiLSTM-CRF, CRF	190601
NLP senior Task2: SemEval2013-Task13-WordSenseInduction	WordSenseInduction	Cluster, LM	BiLM + Clustering/WordNet	190528
Semantic Course Task3: NLPCC 2019 Task2	Semantic Parsing	Generation	Seq2Seq	190520
Deecamp 2019 Exam A	Basic Knowledge	Test	-	190427
NLP cbb Task1: Medicine Corpus processing	Linguistics	Generate	Jieba + Dict + Rules	190418
Semantic Task2: SemEval2017-Task4-SentimentAnalysis	Sentiment	Classify	TextCNN & Bert	190414
NLP senior Task1: SemEval2018-Task7-RelationClassification	Semantic Relation	Multi Classify	TextCNN & LR & LinearSV	190411
Concrete	Feature	Classify	lightGBM	190328
Semantic Course Task1	Word Similarity	Regression	word2vec & Bert & WordNet	190324
Interview	Feature	Classify	lightGBM	190318
Elo	Feature	Classify	lightGBM	190304
Titanic	Feature	Classify	lightGBM	181220

NLP senior Task3: Chinese traditional Sequence Annotation

• Task: Chinese traditional Sequence Annotation, CWS & NER
• Code: iofu728/Chinese_T-Sequence-annotation
• Final paper: iofu728/Chinese_T-Sequence-annotation
• Result: CWS: 97.78%/Dev F1; NER: 96.92%/Dev F1

DataMining Project

• Task Detecting Incongruity Between News Headline and Body Text via a Deep Hierarchical Encoder
• Final paper dataMining

NLP cbb Task2: Medicine Corpus processing with Corpus

• Task: Chinese word segmentation, part-of-speech tagging, named-entity recognition
• Code: pku-nlp-forfun/CWS_POS_NER
• Final paper: pku-nlp-forfun/CWS_POS_NER
• Result: CWS: 90.55/Dev F1, 91.16%/Test F1; POS: 96.06/Dev F1, 95.86%/Test F1(Top 10)

NLP senior Task2: SemEval2013-Task13-WordSenseInduction

• Task: SemEval2013-Task13-WordSenseInduction

  • Word Sense Induction (WSI) seeks to identify the different senses (or uses) of a target word in a given text in an automatic and fully-unsupervised manner. It is a key-enabling technology that aims to overcome the limitations associated with traditional knowledge-based & supervised Word Sense Disambiguation (WSD) methods, such as:
  • their limited adaptation to new languages and domains
  • the fixed granularity of senses
  • their inability to detect new senses (uses) not present in a given dictionary

• Code: pku-nlp-forfun/SemEval2013-WordSenseInduction

• Final paper: Multi-fusion on SemEval-2013 Task13: Word Sense Induction

• Result: 11.06%/Fuzzy NMI, 57.72%/Fuzzy B-Cubed, 25.27%/Average

Semantic Course Task3: NLPCC 2019 Task2 Open Domain Semantic Parsing

• Task: Senantic Parsing

  • In this task, a Multi-perspective Semantic ParSing (or MSParS) dataset will be released, which can be used to evaluate the performance of a semantic parser from different aspects. This dataset includes more than 80,000 human-generated questions, where each question is annotated with entities, the question type and the corresponding logical form. We split MSParS into a train set, a development set and a test set

• Code: semantic/task3

• Final paper: semantic/task3

• • Result: BLEU 0.538

DeeCamp 2019 Exam A

• Problem: DeeCamp 2019 Exam A

NLP cbb Task1: Medicine Corpus processing

• Task: Chinese word segmentation, part-of-speech tagging, named-entity recognition in Rule
• Code: nlpcbb/task1
• Final paper: nlpcbb/task1

Semantic Course Task2: SemEval2017-Task4-SentimentAnalysis

• Task: SemEval2017-Task4-SentimentAnalysis

  • Subtask A. (rerun): Message Polarity Classification: Given a message, classify whether the message is of positive, negative, or neutral sentiment.

• Code: iofu728/SemEval2017-Task4-SentimentAnalysis

• Final paper: Bert & TextCNN on SemEval-2017 Task4 subTask A: Sentiment Analysis in twitter

• Result: 69.39%/Recall, 69.91%/Macro-F1, 70.03%/accuracy-subTaskA(No.1)

NLP senior Task1: SemEval2018-Task7-RelationClassification

• Task: SemEval-2018 task 7 Semantic Relation Extraction and Classification in Scientific Papers

  • The NLP tasks that underlie intelligent processing of scientific documents are those of information extraction: identifying concepts and recognizing the semantic relation that holds between them. The current task adresses semantic relation extraction and classification into 6 categories, all of them specific to scientific literature.

• Record: pku-nlp-forfun/SemEval-2018-RelationClassification

• Code: pku-nlp-forfun/SemEval-2018-RelationClassification

• Final paper: pku-nlp-forfun/SemEval2018-Task7-Paper

• Result: Macro_f1: 67.74%/subTask1.1(No.8), 77.35%/subTask1.2(No.7)

Concrete

• Task: Concrete Warning
• code: concrete

Semantic Course Task1

• Task: Word similarity

  • Two way to calculate word similarity. one for dictionary, one for corpus.
  • Data: Mturk-771
  • http://www2.mta.ac.il/~gideon/mturk771.html

• Final result: semanticTask1

• code: semantic/Task1

Interview

• Task:

  • Description: This is a very simple binary classification task, you can design your model in any way you want.
  • Evaluation: AUC, area under the ROC curve
  • Data: The data file are provided in your answer submission folder in SharePoint (the link shown above).
  • train.csv: used to train model.
  • test.csv: used to metric your model's performance.
  • Feedback / Submission:
  • Your AUC score at the test dataSet.
  • A brief description of your model and feature engineering.
  • Your code.

• Final result: Code record

• Code: interview

• Result: macro_f1: 84.707%/Train, 82.429%/Test

Elo

• Task: Elo Merchant Category Recommendation
• code: elo

Titanic

• Task: Titanic: Machine Learning from Disaster
• code: titanic