Submission ruBert-large finetune

Aug. 18, 2021, 11:55 a.m.

Team: SberDevices

Model url: https://huggingface.co/sberbank-ai/ruBert-large


Total score: 0.62

Dataset Score Metric
LiDiRus 0.235 Matthew`s Corr
RCB 0.356 / 0.5 F1/Acc
PARus 0.656 Accuracy
MuSeRC 0.778 / 0.436 F1a/Em
TERRa 0.704 Accuracy
RUSSE 0.707 Accuracy
RWSD 0.669 Accuracy
DaNetQA 0.773 Accuracy
RuCoS 0.81 / 0.805 F1/EM
Model description:

Russian Bert-large model. 430ml parameters, trained on 30 GB Russian texts.


Parameter description:

-

Diagnostic (Matthew`s Correlation): 0.235

Category Score
LOGIC 0.13965594076204224
KNOWLEDGE 0.25420452490107115
PREDICATE-ARGUMENT STRUCTURE 0.2110557307641125
LEXICAL SEMANTICS 0.23178921890059834
Lexical Semantics - Lexical Entailment 0.1605357781060343
Lexical Semantics - Morphological Negation 0.060993754559283325
Lexical Semantics - Factivity 0.21320071635561044
Lexical Semantics - Symmetry/Collectivity 0.0
Lexical Semantics - Redundancy 0.45652173913043476
Lexical Semantics - Named Entities 0.1767766952966369
Lexical Semantics - Quantifiers 0.1913754029921869
Predicate-Argument Structure Core Args 0.33895960971961925
Predicate-Argument Structure Prepositional Phrases 0.44921394657019476
Predicate-Argument Structure Ellipsis/Implicits 0.14818474149146235
Predicate-Argument Structure Anaphora/Coreference 0.07001400420140048
Predicate-Argument Structure Active/Passive 0.3059179896521843
Predicate-Argument Structure Nominalization 0.0
Predicate-Argument Structure Genitives/Partitives 0.25
Predicate-Argument Structure Datives 0.37796447300922725
Predicate-Argument Structure Relative Clauses 0.13912166872805048
Predicate-Argument Structure Coordination Scopes 0.20894746626614943
Predicate-Argument Structure Intersectivity 0.03545133231866123
Predicate-Argument Structure Restrictivity -0.1455213750217998
Logic Negation 0.08219949365267865
Logic Double Negation 0.0
Logic Interval/Numbers 0.2876779808912305
Logic Conjuction 0.29370130064810657
Logic Disjunction 0.2780367699071248
Logic Conditionals 0.14285714285714285
Logic Universal 0.5324675324675324
Logic Existential 0.10482848367219183
Logic Temporal -0.12434118282549846
Logic Upward Monotone 0.31353483628976775
Logic Downward Monotone -0.09513029883089882
Logic Non-Monotonic 0.0
Knowledge Common Sense 0.3022898014232672
Knowledge World Knowledge 0.17129218721431935

Performance:

Dataset Speed RAM
LiDiRus - -
RCB - -
PARus - -
MuSeRC - -
TERRa - -
RUSSE - -
RWSD - -
DaNetQA - -
RuCoS - -