اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدRefining Language Models with Compositional Explanations
77
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Pre-trained language models have been successful on text classification tasks, but are prone to learning spurious correlations from biased datasets, and are thus vulnerable when making inferences in a new domain. Prior works reveal such spurious patterns via post-hoc explanation algorithms which compute the importance of input features. Further, the model is regularized to align the importance scores with human knowledge, so that the unintended model behaviors are eliminated. However, such a regularization technique lacks flexibility and coverage, since only importance scores towards a pre-defined list of features are adjusted, while more complex human knowledge such as feature interaction and pattern generalization can hardly be incorporated. In this work, we propose to refine a learned language model for a target domain by collecting human-provided compositional explanations regarding observed biases. By parsing these explanations into executable logic rules, the human-specified refinement advice from a small set of explanations can be generalized to more training examples. We additionally introduce a regularization term allowing adjustments for both importance and interaction of features to better rectify model behavior. We demonstrate the effectiveness of the proposed approach on two text classification tasks by showing improved performance in target domain as well as improved model fairness after refinement.
قيم البحث
اقرأ أيضاً
Advances in machine reading comprehension (MRC) rely heavily on the collection of large scale human-annotated examples in the form of (question, paragraph, answer) triples. In contrast, humans are typically able to generalize with only a few examples
, relying on deeper underlying world knowledge, linguistic sophistication, and/or simply superior deductive powers. In this paper, we focus on teaching machines reading comprehension, using a small number of semi-structured explanations that explicitly inform machines why answer spans are correct. We extract structured variables and rules from explanations and compose neural module teachers that annotate instances for training downstream MRC models. We use learnable neural modules and soft logic to handle linguistic variation and overcome sparse coverage; the modules are jointly optimized with the MRC model to improve final performance. On the SQuAD dataset, our proposed method achieves 70.14% F1 score with supervision from 26 explanations, comparable to plain supervised learning using 1,100 labeled instances, yielding a 12x speed up.
Targeted syntactic evaluation of subject-verb number agreement in English (TSE) evaluates language models syntactic knowledge using hand-crafted minimal pairs of sentences that differ only in the main verbs conjugation. The method evaluates whether l
anguage models rate each grammatical sentence as more likely than its ungrammatical counterpart. We identify two distinct goals for TSE. First, evaluating the systematicity of a language models syntactic knowledge: given a sentence, can it conjugate arbitrary verbs correctly? Second, evaluating a models likely behavior: given a sentence, does the model concentrate its probability mass on correctly conjugated verbs, even if only on a subset of the possible verbs? We argue that current implementations of TSE do not directly capture either of these goals, and propose new metrics to capture each goal separately. Under our metrics, we find that TSE overestimates systematicity of language models, but that models score up to 40% better on verbs that they predict are likely in context.
This thesis is about the problem of compositionality in distributional semantics. Distributional semantics presupposes that the meanings of words are a function of their occurrences in textual contexts. It models words as distributions over these con
texts and represents them as vectors in high dimensional spaces. The problem of compositionality for such models concerns itself with how to produce representations for larger units of text by composing the representations of smaller units of text. This thesis focuses on a particular approach to this compositionality problem, namely using the categorical framework developed by Coecke, Sadrzadeh, and Clark, which combines syntactic analysis formalisms with distributional semantic representations of meaning to produce syntactically motivated composition operations. This thesis shows how this approach can be theoretically extended and practically implemented to produce concrete compositional distributional models of natural language semantics. It furthermore demonstrates that such models can perform on par with, or better than, other competing approaches in the field of natural language processing. There are three principal contributions to computational linguistics in this thesis. The first is to extend the DisCoCat framework on the syntactic front and semantic front, incorporating a number of syntactic analysis formalisms and providing learning procedures allowing for the generation of concrete compositional distributional models. The second contribution is to evaluate the models developed from the procedures presented here, showing that they outperform other compositional distributional models present in the literature. The third contribution is to show how using category theory to solve linguistic problems forms a sound basis for research, illustrated by examples of work on this topic, that also suggest directions for future research.
Generating context-aware language that embodies diverse emotions is an important step towards building empathetic NLP systems. In this paper, we propose a formulation of modulated layer normalization -- a technique inspired by computer vision -- that
allows us to use large-scale language models for emotional response generation. In automatic and human evaluation on the MojiTalk dataset, our proposed modulated layer normalization method outperforms prior baseline methods while maintaining diversity, fluency, and coherence. Our method also obtains competitive performance even when using only 10% of the available training data.
To obtain high-quality sentence embeddings from pretrained language models (PLMs), they must either be augmented with additional pretraining objectives or finetuned on a large set of labeled text pairs. While the latter approach typically outperforms
the former, it requires great human effort to generate suitable datasets of sufficient size. In this paper, we show how large PLMs can be leveraged to obtain high-quality embeddings without requiring any labeled data, finetuning or modifications to the pretraining objective: We utilize the generative abilities of PLMs to generate entire datasets of labeled text pairs from scratch, which can then be used for regular finetuning of much smaller models. Our fully unsupervised approach outperforms strong baselines on several English semantic textual similarity datasets.
الأسئلة المقترحة
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
