Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Improving Compositional Generalization with Latent Structure and Data Augmentation

About

Generic unstructured neural networks have been shown to struggle on out-of-distribution compositional generalization. Compositional data augmentation via example recombination has transferred some prior knowledge about compositionality to such black-box neural models for several semantic parsing tasks, but this often required task-specific engineering or provided limited gains. We present a more powerful data recombination method using a model called Compositional Structure Learner (CSL). CSL is a generative model with a quasi-synchronous context-free grammar backbone, which we induce from the training data. We sample recombined examples from CSL and add them to the fine-tuning data of a pre-trained sequence-to-sequence model (T5). This procedure effectively transfers most of CSL's compositional bias to T5 for diagnostic tasks, and results in a model even stronger than a T5-CSL ensemble on two real world compositional generalization tasks. This results in new state-of-the-art performance for these challenging semantic parsing tasks requiring generalization to both natural language variation and novel compositions of elements.

Linlu Qiu, Peter Shaw, Panupong Pasupat, Pawe{\l} Krzysztof Nowak, Tal Linzen, Fei Sha, Kristina Toutanova• 2021

Related benchmarks

TaskDatasetResultRank
Semantic ParsingGeoQuery (i.i.d.)
Exact Match Accuracy93.3
32
Semantic ParsingGeoQuery compositional
Accuracy89.3
29
Semantic ParsingCOGS (generalization)
Accuracy (Generalization)99
25
Semantic ParsingSMCalFlow-CS (16-C)
Accuracy61.4
20
Instruction FollowingSCAN jump
Accuracy99.7
18
Semantic ParsingGeoQuery (Len.)
Exact Match Accuracy67.8
17
Semantic ParsingCOGS
Accuracy99.5
9
Semantic ParsingSMCalFlow-CS (32-C)
Accuracy70.4
8
Semantic ParsingSMCalFlow CS (8-C)--
8
Semantic ParsingSMCalFlow-CS (0-C)--
8
Showing 10 of 14 rows

Other info

Follow for update