Merge branch 'master' into link2concept

annotations/D18-1332.txt

@@ -0,0 +1,7 @@
+# Title: Accelerating Asynchronous Stochastic Gradient Descent for Neural Machine Translation
+# Online location: https://www.aclweb.org/anthology/D18-1332.pdf
+train-parallel
+optim-adam
+norm-gradient
+arch-gru
+task-seq2seq

annotations/D19-1555.txt

@@ -0,0 +1,6 @@
+# Title: Leveraging Structural and Semantic Correspondence for Attribute-Oriented Aspect Sentiment Discovery
+# Online location: https://www.aclweb.org/anthology/D19-1555.pdf
+# CHECK: confidence=0.9, justification=Matched regex attention
+arch-att
+pre-word2vec
+pre-use

annotations/N18-1158.txt

@@ -0,0 +1,10 @@
+# Title: Ranking Sentences for Extractive Summarization with Reinforcement Learning
+# Online location: https://www.aclweb.org/anthology/N18-1158.pdf
+optim-adam
+pool-max
+arch-lstm
+arch-cnn
+nondif-reinforce
+task-extractive
+task-seq2seq
+pre-word2vec

annotations/N18-2097.txt

@@ -0,0 +1,9 @@
+# Title: A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents
+# Online location: https://www.aclweb.org/anthology/N18-2097.pdf
+arch-bilstm
+arch-att
+arch-copy
+arch-coverage
+search-beam
+optim-adagrad
+task-seq2seq

annotations/N19-1015.txt

@@ -0,0 +1,9 @@
+# Title: Topic-Guided Variational Auto-Encoder for Text Generation
+# Online location: https://www.aclweb.org/anthology/N19-1015.pdf
+reg-dropout
+arch-gru
+arch-att
+latent-vae
+latent-topic
+task-lm
+task-seq2seq

annotations/N19-1154.txt

@@ -0,0 +1,8 @@
+# Title: One Size Does Not Fit All: Comparing NMT Representations of Different Granularities
+# Online location: https://www.aclweb.org/anthology/N19-1154.pdf
+optim-sgd
+arch-lstm
+arch-att
+arch-subword
+task-seqlab
+task-seq2seq

annotations/N19-1157.txt

@@ -0,0 +1,3 @@
+# Title: Quantifying the morphosyntactic content of Brown Clusters
+# Online location: https://www.aclweb.org/anthology/N19-1157.pdf
+not-neural

annotations/P19-1511.txt

@@ -0,0 +1,8 @@
+# Title: Sequence-to-Nuggets: Nested Entity Mention Detection via Anchor-Region Networks
+# Online location: https://www.aclweb.org/anthology/P19-1511.pdf
+arch-bilstm
+arch-cnn
+pre-glove
+optim-adadelta
+struct-crf
+task-seqlab

concepts.md

@@ -65,7 +65,8 @@ and no other tags need to be applied.
 * Transfer Learning: [`train-transfer`](http://pfliu.com/pl-nlp2019/bs/train-transfer.html)
 * Active Learning: [`train-active`](http://pfliu.com/pl-nlp2019/bs/train-active.html)
 * Data Augmentation: [`train-augment`](http://pfliu.com/pl-nlp2019/bs/train-augment.html)
-
+* Curriculum Learning: `train-curriculum`
+* Parallel Training: `train-parallel`
 
 ## Sequence Modeling Architectures
 
@@ -106,6 +107,8 @@ and no other tags need to be applied.
 * Memory: [`arch-memo`](http://pfliu.com/pl-nlp2019/bs/arch-memo.html)
 * Copy Mechanism: [`arch-copy`](http://pfliu.com/pl-nlp2019/bs/arch-copy.html)
 * Bilinear, Biaffine Models: [`arch-bilinear`](http://pfliu.com/pl-nlp2019/bs/arch-bilinear.html)
+* Coverage Vectors/Penalties: `arch-coverage`
+* Subword Units: `arch-subword`
 
 ### Standard Composite Architectures
 
@@ -151,6 +154,7 @@ and no other tags need to be applied.
 * Skip-thought: [`pre-skipthought`](http://pfliu.com/pl-nlp2019/bs/pre-skipthought.html) (implies `arch-lstm`, `task-seq2seq`)
 * ELMo: [`pre-elmo`](http://pfliu.com/pl-nlp2019/bs/pre-elmo.html) (implies `arch-bilstm`, `task-lm`)
 * BERT: [`pre-bert`](http://pfliu.com/pl-nlp2019/bs/pre-bert.html) (implies `arch-transformer`, `task-cloze`, `task-textpair`)
+* Universal Sentence Encoder (USE): `pre-use` (implies `arch-transformer`, `task-seq2seq`)
 
 ## Structured Models/Algorithms
 

template.cpt

@@ -28,6 +28,8 @@ train-multiling	train-multitask	cross-lingual|multi-lingual|cross language
 train-transfer	null	transfer learning|domain adaptation
 train-active	null	active learning
 train-augment	null	data augmentation|Data Augmentation
+train-curriculum	null	data curriculum
+train-parallel	null	parallelism
 # ---------- Activation Functions ----------
 activ-tanh	null	Hyperbolic Tangent|hyperbolic tangent
 activ-relu	null	Rectified Linear Units|rectified linear
@@ -56,6 +58,8 @@ arch-gating	null	gating connections|Highway
 arch-memo	null	memory network|external memory
 arch-copy	null	copy mechanism|copying mechanism
 arch-bilinear	null	bilinear|bi-linear|biaffine|bi-affine
+arch-coverage	null	coverage
+arch-subword	null	subword|BPE|sentencepiece
 # ---------- Standard Composite Architectures ----------
 arch-transformer	arch-selfatt	Transformer
 # ---------- Model Combination ----------
@@ -71,9 +75,10 @@ search-gumbel	search-sampling	Gumbel Max|gumbel max
 pre-word2vec	null	word2vec|Word2vec
 pre-glove	null	glove|GloVe
 pre-pv	null	paragraph vector|ParaVector
-pre-skipthought	task-seq2seq	Skip-thought|skip-thought
+pre-skipthought	task-seq2seq	Skip-thought|skip-thought|skipthought
 pre-elmo	task-lm	ELMo
 pre-bert	arch-transformer	BERT
+pre-use	null	Universal Sentence Encoder|universal sentence encoder
 # ---------- Pre-trained Embedding Techniques ----------
 struct-hmm	null	Hidden Markov Models|hidden markov
 struct-crf	null	Conditional Random Fields|conditional random fields|CRF
@@ -92,7 +97,7 @@ adv-examp	null	Adversarial Examples|adversarial examples
 adv-train	adv-examp	Adversarial Training|adversarial trainin
 # ---------- Latent Variable Models ----------
 latent-vae	null	Variational Auto-encoder|variational auto-encoder|latent variable
-latent-topic	null	Topic Model|topic model
+latent-topic	null	topic model
 # ---------- Loss Functions ----------
 loss-cca	null	Canonical Correlation Analysis|canonical correlation analysis
 loss-svd	null	Singular Value Decomposition|SVD|singular value decomposition