1 |
Parameter-Efficient Neural Reranking for Cross-Lingual and Multilingual Retrieval ...
|
|
|
|
BASE
|
|
Show details
|
|
7 |
Data for paper: "Evaluating Resource-Lean Cross-Lingual Embedding Models in Unsupervised Retrieval" ...
|
|
|
|
BASE
|
|
Show details
|
|
8 |
Crossing the Conversational Chasm: A Primer on Natural Language Processing for Multilingual Task-Oriented Dialogue Systems ...
|
|
|
|
BASE
|
|
Show details
|
|
9 |
On Cross-Lingual Retrieval with Multilingual Text Encoders ...
|
|
|
|
BASE
|
|
Show details
|
|
10 |
Evaluating Multilingual Text Encoders for Unsupervised Cross-Lingual Retrieval ...
|
|
|
|
BASE
|
|
Show details
|
|
11 |
RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational Language Models ...
|
|
|
|
BASE
|
|
Show details
|
|
12 |
LexFit: Lexical Fine-Tuning of Pretrained Language Models ...
|
|
|
|
BASE
|
|
Show details
|
|
13 |
Verb Knowledge Injection for Multilingual Event Processing ...
|
|
|
|
BASE
|
|
Show details
|
|
14 |
Is supervised syntactic parsing beneficial for language understanding tasks? An empirical investigation
|
|
|
|
BASE
|
|
Show details
|
|
15 |
Evaluating multilingual text encoders for unsupervised cross-lingual retrieval
|
|
|
|
BASE
|
|
Show details
|
|
16 |
Training and domain adaptation for supervised text segmentation
|
|
|
|
BASE
|
|
Show details
|
|
17 |
AraWEAT: Multidimensional Analysis of Biases in Arabic Word Embeddings ...
|
|
|
|
BASE
|
|
Show details
|
|
18 |
XCOPA: A Multilingual Dataset for Causal Commonsense Reasoning ...
|
|
|
|
BASE
|
|
Show details
|
|
19 |
On the Limitations of Cross-lingual Encoders as Exposed by Reference-Free Machine Translation Evaluation ...
|
|
|
|
BASE
|
|
Show details
|
|
20 |
Orthogonal Language and Task Adapters in Zero-Shot Cross-Lingual Transfer ...
|
|
|
|
Abstract:
Adapter modules, additional trainable parameters that enable efficient fine-tuning of pretrained transformers, have recently been used for language specialization of multilingual transformers, improving downstream zero-shot cross-lingual transfer. In this work, we propose orthogonal language and task adapters (dubbed orthoadapters) for cross-lingual transfer. They are trained to encode language- and task-specific information that is complementary (i.e., orthogonal) to the knowledge already stored in the pretrained transformer's parameters. Our zero-shot cross-lingual transfer experiments, involving three tasks (POS-tagging, NER, NLI) and a set of 10 diverse languages, 1) point to the usefulness of orthoadapters in cross-lingual transfer, especially for the most complex NLI task, but also 2) indicate that the optimal adapter configuration highly depends on the task and the target language. We hope that our work will motivate a wider investigation of usefulness of orthogonality constraints in language- and ...
|
|
Keyword:
Computation and Language cs.CL; FOS Computer and information sciences
|
|
URL: https://dx.doi.org/10.48550/arxiv.2012.06460 https://arxiv.org/abs/2012.06460
|
|
BASE
|
|
Hide details
|
|
|
|