kudkudak/common-sense-prediction

Issues

• Li filters out of vocab triplets from wiki, I don't

  #70 opened 5 years ago by kudkudak
  2

• Explore regularizers for glove

  #72 opened 5 years ago by kudkudak
  1

• Improvements to train scripts

  #73 opened 5 years ago by kudkudak
  0

• Fixes to experiments

  #74 opened 5 years ago by kudkudak
  0

• Compile Appendix with novel and trivial triplets from 4.1 section.

  #76 opened 5 years ago by kudkudak
  0

• Rethink reporting F1 on imbalance of negative class bucket

  #79 opened 5 years ago by kudkudak
  0

• vocab should be property of dataset, not embedding.

  #33 opened 5 years ago by kudkudak
  2

• Check relation of conceptnet5 to version used in paper

  #48 opened 5 years ago by kudkudak
  0

• Grid to pick hyperparameter for each model

  #65 opened 5 years ago by kudkudak
  2

• Unit test human evaluation

  #68 opened 5 years ago by kudkudak
  3

• Re-evaluate one failed model on wiki buckets

  #78 opened 7 years ago by kudkudak
  0

• Rescore on whole wiki baselines vs Li et al.

  #71 opened 7 years ago by kudkudak
  0

• Think about balance of classes

  #75 opened 7 years ago by kudkudak
  1

• Prepare data for Dima

  #77 opened 7 years ago by kudkudak
  0

• Interhuman agreement of scores

  #69 opened 7 years ago by kudkudak
  1

• Score 900 triplets forming base of wiki evaluation

  #67 opened 7 years ago by kudkudak
  0

• Bucket wiki into 3 buckets - evaluation scripts

  #64 opened 7 years ago by kudkudak
  0

• Fast computing distance script

  #61 opened 7 years ago by kudkudak
  1

• Dev re-make

  #63 opened 7 years ago by kudkudak
  0

• Double check negative sampling

  #66 opened 7 years ago by kudkudak
  0

• Human novelty evaluation

  #62 opened 7 years ago by kudkudak
  0

• Write sketch of text

  #59 opened 7 years ago by kudkudak
  0

• merge LiACLDatasetFromFile and LiACLSplitDataset

  #34 opened 7 years ago by kudkudak
  0

• Improve human evaluation script

  #44 opened 7 years ago by kudkudak
  1

• Code script to add distances, run for wiki and conceptnet, analyze

  #46 opened 7 years ago by kudkudak
  8

• Label negative samples from test

  #58 opened 7 years ago by kudkudak
  0

• Explore wiki distances

  #60 opened 7 years ago by kudkudak
  0

• Examine regularization for glove embeddings when used together with OMCS

  #51 opened 7 years ago by kudkudak
  1

• Negative samples outside of batch

  #52 opened 7 years ago by kudkudak
  4

• Manually label the 5k conceptnet

  #56 opened 7 years ago by kudkudak
  3

• Human examination of split quality

  #57 opened 7 years ago by kudkudak
  1

• Check if distances on conceptnet extrinsic test are useful

  #49 opened 7 years ago by kudkudak
  0

• Negative samples using ArgSim - human based examination

  #50 opened 7 years ago by kudkudak
  0

• Run more models on new negative sample split

  #53 opened 7 years ago by kudkudak
  0

• Small bug in argsim_threshold fnc: avg should count non-zeros

  #40 opened 7 years ago by kudkudak
  1

• Investigate whether we can use shitty triplets from CN as negative examples

  #47 opened 7 years ago by kudkudak
  1

• Create a golden set of "trivial/novel" pairs

  #45 opened 7 years ago by kudkudak
  4

• MaxSim with Fuel

  #36 opened 7 years ago by kudkudak
  0

• Human based negative samples - decide on protocol

  #43 opened 7 years ago by kudkudak
  0

• Perform human evaluation on wiki

  #42 opened 7 years ago by kudkudak
  2

• MaxSim - negative sampling using MaxSim

  #41 opened 7 years ago by kudkudak
  0

• Groundword: do hp optimization for current DNN_CE and Factorized

  #30 opened 7 years ago by kudkudak
  1

• Try argsim and prototypical negative samples - effect on extrinsic and test gen

  #31 opened 7 years ago by kudkudak
  1

• Try to merge ArgSim and MaxSim

  #38 opened 7 years ago by kudkudak
  0

• Closest-neighbour analysis

  #39 opened 7 years ago by kudkudak
  0

• Add conceptnet extrinsic eval

  #37 opened 7 years ago by kudkudak
  0

• Try to figure out why their model works well on extrinsic eval

  #32 opened 7 years ago by kudkudak
  1

• Evaluate on wiki

  #35 opened 7 years ago by kudkudak
  0

• vegab improve meta.json

  #26 opened 7 years ago by mnoukhov
  0

• Add good configs for Factorized to config file

  #27 opened 7 years ago by kudkudak
  0