Diversify Your Datasets: Analyzing Generalization via Controlled Variance in Adversarial Datasets
| Autor: | Vered Shwartz, Ohad Rozen, Ido Dagan, Roee Aharoni |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2019 |
| Předmět: |
FOS: Computer and information sciences
Training set Computer Science - Computation and Language Generalization Computer science Process (engineering) business.industry Inference 02 engineering and technology Variance (accounting) Machine learning computer.software_genre 03 medical and health sciences Adversarial system 0302 clinical medicine Phenomenon 030221 ophthalmology & optometry 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Artificial intelligence Focus (optics) business computer Computation and Language (cs.CL) |
| Zdroj: | CoNLL |
| Popis: | Phenomenon-specific "adversarial" datasets have been recently designed to perform targeted stress-tests for particular inference types. Recent work (Liu et al., 2019a) proposed that such datasets can be utilized for training NLI and other types of models, often allowing to learn the phenomenon in focus and improve on the challenge dataset, indicating a "blind spot" in the original training data. Yet, although a model can improve in such a training process, it might still be vulnerable to other challenge datasets targeting the same phenomenon but drawn from a different distribution, such as having a different syntactic complexity level. In this work, we extend this method to drive conclusions about a model's ability to learn and generalize a target phenomenon rather than to "learn" a dataset, by controlling additional aspects in the adversarial datasets. We demonstrate our approach on two inference phenomena - dative alternation and numerical reasoning, elaborating, and in some cases contradicting, the results of Liu et al.. Our methodology enables building better challenge datasets for creating more robust models, and may yield better model understanding and subsequent overarching improvements. CoNLL 2019 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|