While recent language models have the ability to take long contexts as input, relatively little is known about how well they use longer context. We analyze the performance of language models on two tasks that require identifying relevant information in their input contexts: multi-document question answering and key-value retrieval. We find that performance can degrade significantly when changing the position of relevant information, indicating that current language models do not robustly make use of information in long input contexts. In particular, we observe that performance is often highest when relevant information occurs at the beginning or end of the input context, and significantly degrades when models must access relevant information in the middle of long contexts, even for explicitly long-context models. Our analysis provides a better understanding of how language models use their input context and provides new evaluation protocols for future long-context language models.

We present Backpacks: a new neural architecture that marries strong modeling performancewith an interface for interpretability and control. Backpacks learn multiple non-contextual sense vectors for each word in a vocabulary, and represent a word in a sequence as a context-dependent, non-negative linear combination ofsense vectors in this sequence. We find that, after training, sense vectors specialize, each encoding a different aspect of a word. We can interpret a sense vector by inspecting its (non-contextual, linear) projection onto the output space, and intervene on these interpretable hooks to change the model’s behavior in predictable ways. We train a 170M-parameter Backpack language model on OpenWebText, matching the loss of a GPT-2 small (124Mparameter) Transformer. On lexical similarity evaluations, we find that Backpack sense vectors outperform even a 6B-parameter Transformer LM’s word embeddings. Finally, we present simple algorithms that intervene on sense vectors to perform controllable text generation and debiasing. For example, we can edit the sense vocabulary to tend more towards a topic, or localize a source of gender bias to a sense vector and globally suppress that sense.

An increasingly large percentage of natural language processing (NLP) tasks center around the generation of text from probabilistic language models. Despite this trend, techniques for improving or specifying preferences in these generated texts rely mostly on intuition-based heuristics. Further, there lacks a unified presentation of their motivations, practical implementation, successes and pitfalls. Practitioners must, therefore, choose somewhat blindly between generation algorithms—like top-p sampling or beam search—which can lead to wildly different results. At the same time, language generation research continues to criticize and improve the standard toolboxes, further adding entropy to the state of the field. In this tutorial, we will provide a centralized and cohesive discussion of critical considerations when choosing how to generate from a language model. We will cover a wide range of empirically-observed problems (like degradation, hallucination, repetition) and their corresponding proposed algorithmic solutions from recent research (like top-p sampling and its successors). We will then discuss a subset of these algorithms under a unified light; most stochastic generation strategies can be framed as locally adapting the probabilities of a model to avoid failure cases. Finally, we will then cover methods in controlled generation, that go beyond just ensuring coherence to ensure text exhibits specific desired properties. We aim for NLP practitioners and researchers to leave our tutorial with a unified framework which they can use to evaluate and contribute to the latest research in language generation.

The Backpack is a Transformer alternative shown to improve interpretability in English language modeling by decomposing predictions into a weighted sum of token sense components. However, Backpacks’ reliance on token-defined meaning raises questions as to their potential for languages other than English, a language for which subword tokenization provides a reasonable approximation for lexical items. In this work, we train, evaluate, interpret, and control Backpack language models in character-tokenized Chinese, in which words are often composed of many characters. We find that our (134M parameter) Chinese Backpack language model performs comparably to a (104M parameter) Transformer, and learns rich character-level meanings that log-additively compose to form word meanings. In SimLex-style lexical semantic evaluations, simple averages of Backpack character senses outperform input embeddings from a Transformer. We find that complex multi-character meanings are often formed by using the same per-character sense weights consistently across context. Exploring interpretability-through control, we show that we can localize a source of gender bias in our Backpacks to specific character senses and intervene to reduce the bias.

Long samples of text from neural language models can be of poor quality. Truncation sampling algorithms–like top-p or top-k—address this by setting some words’ probabilities to zero at each step. This work investigates why these methods are important, and how to improve them. We propose thinking of a neural language model as a mixture of a true distribution and a smoothing distribution that avoids infinite perplexity. In this light, truncation algorithms aim to perform desmoothing, estimating a subset of the support of the true distribution. Finding a good subset is crucial: we show that top-p unnecessarily truncates high-probability words, for example causing it to truncate all words but Trump for a document that starts with Donald. We introduce eta-sampling, which truncates words below an entropy-dependent probability threshold. Compared to previous algorithms, our eta-sampling generates more plausible long documents according to humans, is better at breaking out of repetition, and behaves more reasonably on a battery of test distributions.

JamPatoisNLI provides the first dataset for natural language inference in a creole language, Jamaican Patois.Many of the most-spoken low-resource languages are creoles. These languages commonly have a lexicon derived from a major world language and a distinctive grammar reflecting the languages of the original speakers and the process of language birth by creolization. This gives them a distinctive place in exploring the effectiveness of transfer from large monolingual or multilingual pretrained models. While our work, along with previous work, shows that transfer from these models to low-resource languages that are unrelated to languages in their training set is not very effective, we would expect stronger results from transfer to creoles. Indeed, our experiments show considerably better results from few-shot learning of JamPatoisNLI than for such unrelated languages, and help us begin to understand how the unique relationship between creoles and their high-resource base languages affect cross-lingual transfer. JamPatoisNLI, which consists of naturally-occurring premises and expert-written hypotheses, is a step towards steering research into a traditionally underserved language and a useful benchmark for understanding cross-lingual NLP.

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 verb’s conjugation. The method evaluates whether language 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 model’s syntactic knowledge: given a sentence, can it conjugate arbitrary verbs correctly? Second, evaluating a model’s 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.

Probing experiments investigate the extent to which neural representations make properties—like part-of-speech—predictable. One suggests that a representation encodes a property if probing that representation produces higher accuracy than probing a baseline representation like non-contextual word embeddings. Instead of using baselines as a point of comparison, we’re interested in measuring information that is contained in the representation but not in the baseline. For example, current methods can detect when a representation is more useful than the word identity (a baseline) for predicting part-of-speech; however, they cannot detect when the representation is predictive of just the aspects of part-of-speech not explainable by the word identity. In this work, we extend a theory of usable information called V-information and propose conditional probing, which explicitly conditions on the information in the baseline. In a case study, we find that after conditioning on non-contextual word embeddings, properties like part-of-speech are accessible at deeper layers of a network than previously thought.

Recent work has found evidence that Multilingual BERT (mBERT), a transformer-based multilingual masked language model, is capable of zero-shot cross-lingual transfer, suggesting that some aspects of its representations are shared cross-lingually. To better understand this overlap, we extend recent work on finding syntactic trees in neural networks’ internal representations to the multilingual setting. We show that subspaces of mBERT representations recover syntactic tree distances in languages other than English, and that these subspaces are approximately shared across languages. Motivated by these results, we present an unsupervised analysis method that provides evidence mBERT learns representations of syntactic dependency labels, in the form of clusters which largely agree with the Universal Dependencies taxonomy. This evidence suggests that even without explicit supervision, multilingual masked language models learn certain linguistic universals.

Extrapolation to unseen sequence lengths is a challenge for neural generative models of language. In this work, we characterize the effect on length extrapolation of a modeling decision often overlooked: predicting the end of the generative process through the use of a special end-of-sequence (EOS) vocabulary item. We study an oracle setting - forcing models to generate to the correct sequence length at test time - to compare the length-extrapolative behavior of networks trained to predict EOS (+EOS) with networks not trained to (-EOS). We find that -EOS substantially outperforms +EOS, for example extrapolating well to lengths 10 times longer than those seen at training time in a bracket closing task, as well as achieving a 40% improvement over +EOS in the difficult SCAN dataset length generalization task. By comparing the hidden states and dynamics of -EOS and +EOS models, we observe that +EOS models fail to generalize because they (1) unnecessarily stratify their hidden states by their linear position is a sequence (structures we call length manifolds) or (2) get stuck in clusters (which we refer to as length attractors) once the EOS token is the highest-probability prediction.

Recurrent neural networks empirically generate natural language with high syntactic fidelity. However, their success is not well-understood theoretically. We provide theoretical insight into this success, proving in a finite-precision setting that RNNs can efficiently generate bounded hierarchical languages that reflect the scaffolding of natural language syntax. We introduce Dyck-(k,m), the language of well-nested brackets (of k types) and m-bounded nesting depth, reflecting the bounded memory needs and long-distance dependencies of natural language syntax. The best known results use O(k^m⁄₂) memory (hidden units) to generate these languages. We prove that an RNN with O(m log k) hidden units suffices, an exponential reduction in memory, by an explicit construction. Finally, we show that no algorithm, even with unbounded computation, can suffice with o(m log k) hidden units.

Recent work has improved our ability to detect linguistic knowledge in word representations. However, current methods for detecting syntactic knowledge do not test whether syntax trees are represented in their entirety. In this work, we propose a structural probe, which evaluates whether syntax trees are embedded in a linear transformation of a neural network’s word representation space. The probe identifies a linear transformation under which squared L2 distance encodes the distance between words in the parse tree, and one in which squared L2 norm encodes depth in the parse tree. Using our probe, we show that such transformations exist for both ELMo and BERT but not in baselines, providing evidence that entire syntax trees are embedded implicitly in deep models’ vector geometry.

Probes, supervised models trained to predict properties (like parts-of-speech) from representations (like ELMo), have achieved high accuracy on a range of linguistic tasks. But does this mean that the representations encode linguistic structure or just that the probe has learned the linguistic task? In this paper, we propose control tasks, which associate word types with random outputs, to complement linguistic tasks. By construction, these tasks can only be learned by the probe itself. So a good probe, (one that reflects the representation), should be selective, achieving high linguistic task accuracy and low control task accuracy. The selectivity of a probe puts linguistic task accuracy in context with the probe’s capacity to memorize from word types. We construct control tasks for English part-of-speech tagging and dependency edge prediction, and show that popular probes on ELMo representations are not selective. We also find that dropout, commonly used to control probe complexity, is ineffective for improving selectivity of MLPs, but that other forms of regularization are effective. Finally, we find that while probes on the first layer of ELMo yield slightly better part-of-speech tagging accuracy than the second, probes on the second layer are substantially more selective, which raises the question of which layer better represents parts-of-speech.

With the advent of conversational assistants, like Amazon Alexa, Google Now, etc., dialogue systems are gaining a lot of traction, especially in industrial setting. These systems typically consist of Spoken Language understanding component which, in turn, consists of two tasks - Intent Classification (IC) and Slot Labeling (SL). Generally, these two tasks are modeled together jointly to achieve best performance. However, this joint modeling adds to model obfuscation. In this work, we first design framework for a modularization of joint IC-SL task to enhance architecture transparency. Then, we explore a number of self-attention, convolutional, and recurrent models, contributing a large-scale analysis of modeling paradigms for IC+SL across two datasets. Finally, using this framework, we propose a class of ‘label-recurrent’ models that otherwise non-recurrent, with a 10-dimensional representation of the label history, and show that our proposed systems are easy to interpret, highly accurate (achieving over 30% error reduction in SL over the state-of-the-art on the Snips dataset), as well as fast, at 2x the inference and 2/3 to 1/2 the training time of comparable recurrent models, thus giving an edge in critical real-world systems.

Modeling derivational morphology to generate words with particular semantics is useful in many text generation tasks, such as machine translation or abstractive question answering. In this work, we tackle the task of derived word generation. That is, we attempt to generate the word “runner” for “someone who runs.” We identify two key problems in generating derived words from root words and transformations. We contribute a novel aggregation model of derived word generation that learns derivational transformations both as orthographic functions using sequence-to-sequence models and as functions in distributional word embedding space. The model then learns to choose between the hypothesis of each system. We also present two ways of incorporating corpus information into derived word generation.

We conduct the most comprehensive study to date into translating words via images. To facilitate research on the task, we introduce a large-scale multilingual corpus of images, each labeled with the word it represents. Past datasets have been limited to only a few high-resource languages and unrealistically easy translation settings. In contrast, we have collected by far the largest available dataset for this task, with images for approximately 10,000 words in each of 100 languages. We run experiments on a dozen high resource languages and 20 low resources languages, demonstrating the effect of word concreteness and part-of-speech on translation quality. %We find that while image features work best for concrete nouns, they are sometimes effective on other parts of speech. To improve image-based translation, we introduce a novel method of predicting word concreteness from images, which improves on a previous state-of-the-art unsupervised technique. This allows us to predict when image-based translation may be effective, enabling consistent improvements to a state-of-the-art text-based word translation system. Our code and the Massively Multilingual Image Dataset (MMID) are available at http://multilingual-images.org/.

Bilingual Lexicon Induction is the task of learning word translations without bilingual parallel corpora. We model this task as a matrix completion problem, and present an effective and extendable framework for completing the matrix. This method harnesses diverse bilingual and monolingual signals, each of which may be incomplete or noisy. Our model achieves state-of-the-art performance for both high and low resource languages.

Statistical Machine Translation (SMT) of highly inflected, low-resource languages suffers from the problem of low bitext availability, which is exacerbated by large inflectional paradigms. When translating into English, rich source inflections have a high chance of being poorly estimated or out-of-vocabulary (OOV). We present a source language-agnostic system for automatically constructing phrase pairs from foreign-language inflections and their morphological analyses using manually constructed datasets, including Wiktionary. We then demonstrate the utility of these phrase tables in improving translation into English from Finnish, Czech, and Turkish in simulated low-resource settings, finding substantial gains in translation quality. We report up to +2.58 BLEU in a simulated low-resource setting and +1.65 BLEU in a moderateresource setting. We release our morphologically-motivated translation models, with tens of thousands of inflections in each of 8 languages.