List your company

Dialogue State Tracking on WOZ 2.0 (test)

91.37Joint Goal AccuracyAG-DST
60.929268.832176.73584.6379Jun 12, 2016May 5, 2017Mar 29, 2018Feb 19, 2019Jan 13, 2020Dec 5, 2020Oct 29, 2021

Evaluation results

MethodLinks
AG-DST2021.1091.37—
Seq2Seq-DU2021.1091.2—
SUMBTHistory representation=full history2021.0891—
PegasusHistory representation=full history2021.0891—
T5 (span)History representation=full history2021.0891—
T5 (span+ARLM)History representation=full history2021.0891—
Full BERT-based modelNumber of parameters=110M, Storage Size=440MB, Inference Time on CPU=1.465s, Inference Time on GPU=0.113s2019.1090.597.6
Distilled BERT-based modelNumber of parameters=14M, Storage Size=55MB, Inference Time on CPU=0.205s, Inference Time on GPU=0.024s2019.1090.497.7
T5 (ARLM)History representation=full history2021.0889.5—
StateNet_PSI2018.1088.9—
StateNet_PSI2019.1088.9—
StateNet2019.1088.9—
StateNet2021.1088.9—
StateNet-PSIHistory representation=full history2021.0888.9—
G-SATembeddings=Pre-trained GloVe + character n-gram2019.1088.796.9
COMER2021.1088.6—
GCE2018.1288.5197.38
GCE2019.1088.597.4
GCE2019.1088.597.4
GCEHistory representation=full history2021.0888.5—
StateNet_PS2018.1088.2—
GLAD2018.0588.197.1
GLAD2018.1088.1—
GLAD2019.1088.197.1
GLADNumber of parameters=17M2019.1088.197.1
GLAD2021.1088.1—
GLADHistory representation=full history2021.0888.1—
StateNet2018.1087.8—
BERT-DST PS2019.1087.7—
BERT-DST2021.1087.7—
CNN2019.1086.995.4
Belief Tracking: CNN2018.1085.5—
Belief Tracking2021.1085.5—
Belief Tracking: Bi-LSTM2018.1085.1—
NBT-DNNArchitecture=Neural Belief Tracker, Word Vectors=Paragram-SL9992016.0684.491.2
Neural Belief Tracker (NBT)architecture=DNN2018.0584.491.2
Neural Belief Tracker: NBT-DNN2018.1084.4—
Neural Belief TrackerArchitecture=DNN2018.1284.491.2
NBT - DNN2019.1084.491.2
Neural Belief Tracker - DNN2019.1084.491.2
Neural Belief Tracker: NBT-DNNHistory representation=full history2021.0884.4—
NBT-CNNArchitecture=Neural Belief Tracker, Word Vectors=Paragram-SL9992016.0684.291.6
Neural Belief Tracker (NBT)architecture=CNN2018.0584.291.6
Neural Belief Tracker: NBT-CNN2018.1084.2—
Neural Belief TrackerArchitecture=CNN2018.1284.291.6
NBT - CNN2019.1084.291.6
Neural Belief Tracker - CNN2019.1084.291.6
Neural Belief Tracker2021.1084.2—
Neural Belief Tracker: NBT-CNNHistory representation=full history2021.0884.2—
Delexicalisation-Based Model + Semantic Dictionary2016.0683.787.6
Delex. Model + Semantic Dictionarysemantic_dictionary=true2018.0583.787.6
DB Model + Semantic Dictionary2018.1083.7—
Delex. ModelSemantic Dictionary=true2018.1283.787.6
AUXGANWord Vectors=GLOVE-CC, DST Model=Neural Belief Tracker (NBT), Specialization Type=Post-Specialized2018.0983.6—
POST-DFFNWord Vectors=GLOVE-CC, DST Model=Neural Belief Tracker (NBT), Specialization Type=Post-Specialized2018.0982.9—
ATTRACT-REPELWord Vectors=GLOVE-CC, DST Model=Neural Belief Tracker (NBT), Specialization Type=Specialized2018.0981.7—
DistributionalWord Vectors=GLOVE-CC, DST Model=Neural Belief Tracker (NBT)2018.0979.7—
Delexicalisation-Based Model2016.0670.887.1
Delexicalisation-Based Model2018.0570.887.1
Delexicalisation-Based (DB) Model2018.1070.8—
Delex. Model2018.1270.887.1
Delexicalisation Model2019.1070.887.1
AUXGANVector space=FASTTEXT, Specialization protocol=Zero-shot2018.0965.1—
No pre-trainHistory representation=full history2021.0864.5—
DistributionalVector space=FASTTEXT, Specialization protocol=None2018.0962.1—