Language model prompt optimization research has shown that semantically and grammatically well-formed manually crafted prompts are routinely outperformed by automatically generated token sequences with no apparent meaning or syntactic structure, including sequences of vectors from a model’s embedding space. We use machine-generated prompts to probe how models respond to input that is not composed of natural language expressions. We study the behavior of models of different sizes in multiple semantic tasks in response to both continuous and discrete machine-generated prompts, and compare it to the behavior in response to human-generated natural-language prompts. Even when producing a similar output, machine-generated and human prompts trigger different response patterns through the network processing pathways, including different perplexities, different attention and output entropy distributions, and different unit activation profiles. We provide preliminary insight into the nature of the units activated by different prompt types, suggesting that only natural language prompts recruit a genuinely linguistic circuit.

For a language model (LM) to faithfully model human language, it must compress vast, potentially infinite information into relatively few dimensions. We propose analyzing compression in (pre-trained) LMs from two points of view: geometric and information-theoretic. We demonstrate that the two views are highly correlated, such that the intrinsic geometric dimension of linguistic data predicts their coding length under the LM. We then show that, in turn, high compression of a linguistic dataset predicts rapid adaptation to that dataset, confirming that being able to compress linguistic information is an important part of successful LM performance. As a practical byproduct of our analysis, we evaluate a battery of intrinsic dimension estimators for the first time on linguistic data, showing that only some encapsulate the relationship between information-theoretic compression, geometric compression, and ease-of-adaptation.

We compare the 0-shot performance of a neural caption-based image retriever when given as input either human-produced captions or captions generated by a neural captioner. We conduct this comparison on the recently introduced ImageCoDe data-set (Krojer et al. 2022), which contains hard distractors nearly identical to the images to be retrieved. We find that the neural retriever has much higher performance when fed neural rather than human captions, despite the fact that the former, unlike the latter, were generated without awareness of the distractors that make the task hard. Even more remarkably, when the same neural captions are given to human subjects, their retrieval performance is almost at chance level. Our results thus add to the growing body of evidence that, even when the “language” of neural models resembles English, this superficial resemblance might be deeply misleading.

Large pre-trained deep networks are the standard building blocks of modern AI applications. This raises fundamental questions about how to control their behaviour and how to make them efficiently interact with each other. Deep net emergent communication tackles these challenges by studying how to induce communication protocols between neural network agents, and how to include humans in the communication loop. Traditionally, this research had focussed on relatively small-scale experiments where two networks had to develop a discrete code from scratch for referential communication. However, with the rise of large pre-trained language models that can work well on many tasks, the emphasis is now shifting on how to let these models interact through a language-like channel to engage in more complex behaviors. By reviewing several representative papers, we will provide an introduction to deep net emergent communication, we will cover various central topics from the present and recent past, as well as discussing current shortcomings and suggest future directions. The presentation is complemented by a hands-on section where participants will implement and analyze two emergent communications setups from the literature. The tutorial should be of interest to researchers wanting to develop more flexible AI systems, but also to cognitive scientists and linguists interested in the evolution of communication systems.

In recent years, the NLP community has shown increasing interest in analysing how deep learning models work. Given that large models trained on complex tasks are difficult to inspect, some of this work has focused on controlled tasks that emulate specific aspects of language. We propose a new set of such controlled tasks to explore a crucial aspect of natural language processing that has not received enough attention: the need to retrieve discrete information from sequences. We also study model behavior on the tasks with simple instantiations of Transformers and LSTMs. Our results highlight the beneficial role of decoder attention and its sometimes unexpected interaction with other components. Moreover, we show that, for most of the tasks, these simple models still show significant difficulties. We hope that the community will take up the analysis possibilities that our tasks afford, and that a clearer understanding of model behavior on the tasks will lead to better and more transparent models.

Natural language allows us to refer to novel composite concepts by combining expressions denoting their parts according to systematic rules, a property known as compositionality. In this paper, we study whether the language emerging in deep multi-agent simulations possesses a similar ability to refer to novel primitive combinations, and whether it accomplishes this feat by strategies akin to human-language compositionality. Equipped with new ways to measure compositionality in emergent languages inspired by disentanglement in representation learning, we establish three main results: First, given sufficiently large input spaces, the emergent language will naturally develop the ability to refer to novel composite concepts. Second, there is no correlation between the degree of compositionality of an emergent language and its ability to generalize. Third, while compositionality is not necessary for generalization, it provides an advantage in terms of language transmission: The more compositional a language is, the more easily it will be picked up by new learners, even when the latter differ in architecture from the original agents. We conclude that compositionality does not arise from simple generalization pressure, but if an emergent language does chance upon it, it will be more likely to survive and thrive.

Studies of discrete languages emerging when neural agents communicate to solve a joint task often look for evidence of compositional structure. This stems for the expectation that such a structure would allow languages to be acquired faster by the agents and enable them to generalize better. We argue that these beneficial properties are only loosely connected to compositionality. In two experiments, we demonstrate that, depending on the task, non-compositional languages might show equal, or better, generalization performance and acquisition speed than compositional ones. Further research in the area should be clearer about what benefits are expected from compositionality, and how the latter would lead to them.

Recent work has shown that LSTMs trained on a generic language modeling objective capture syntax-sensitive generalizations such as long-distance number agreement. We have however no mechanistic understanding of how they accomplish this remarkable feat. Some have conjectured it depends on heuristics that do not truly take hierarchical structure into account. We present here a detailed study of the inner mechanics of number tracking in LSTMs at the single neuron level. We discover that long-distance number information is largely managed by two “number units”. Importantly, the behaviour of these units is partially controlled by other units independently shown to track syntactic structure. We conclude that LSTMs are, to some extent, implementing genuinely syntactic processing mechanisms, paving the way to a more general understanding of grammatical encoding in LSTMs.

There is renewed interest in simulating language emergence among deep neural agents that communicate to jointly solve a task, spurred by the practical aim to develop language-enabled interactive AIs, as well as by theoretical questions about the evolution of human language. However, optimizing deep architectures connected by a discrete communication channel (such as that in which language emerges) is technically challenging. We introduce EGG, a toolkit that greatly simplifies the implementation of emergent-language communication games. EGG’s modular design provides a set of building blocks that the user can combine to create new games, easily navigating the optimization and architecture space. We hope that the tool will lower the technical barrier, and encourage researchers from various backgrounds to do original work in this exciting area.

Recent research studies communication emergence in communities of deep network agents assigned a joint task, hoping to gain insights on human language evolution. We propose here a new task capturing crucial aspects of the human environment, such as natural object affordances, and of human conversation, such as full symmetry among the participants. By conducting a thorough pragmatic and semantic analysis of the emergent protocol, we show that the agents solve the shared task through genuine bilateral, referential communication. However, the agents develop multiple idiolects, which makes us conclude that full symmetry is not a sufficient condition for a common language to emerge.

Lake and Baroni (2018) introduced the SCAN dataset probing the ability of seq2seq models to capture compositional generalizations, such as inferring the meaning of “jump around” 0-shot from the component words. Recurrent networks (RNNs) were found to completely fail the most challenging generalization cases. We test here a convolutional network (CNN) on these tasks, reporting hugely improved performance with respect to RNNs. Despite the big improvement, the CNN has however not induced systematic rules, suggesting that the difference between compositional and non-compositional behaviour is not clear-cut.

Embedding a clause inside another (“the girl [who likes cars [that run fast]] has arrived”) is a fundamental resource that has been argued to be a key driver of linguistic expressiveness. As such, it plays a central role in fundamental debates on what makes human language unique, and how they might have evolved. Empirical evidence on the prevalence and the limits of embeddings has however been based on either laboratory setups or corpus data of relatively limited size. We introduce here a collection of large, dependency-parsed written corpora in 17 languages, that allow us, for the first time, to capture clausal embedding through dependency graphs and assess their distribution. Our results indicate that there is no evidence for hard constraints on embedding depth: the tail of depth distributions is heavy. Moreover, although deeply embedded clauses tend to be shorter, suggesting processing load issues, complex sentences with many embeddings do not display a bias towards less deep embeddings. Taken together, the results suggest that deep embeddings are not disfavoured in written language. More generally, our study illustrates how resources and methods from latest-generation big-data NLP can provide new perspectives on fundamental questions in theoretical linguistics.

Sequence-processing neural networks led to remarkable progress on many NLP tasks. As a consequence, there has been increasing interest in understanding to what extent they process language as humans do. We aim here to uncover which biases such models display with respect to “natural” word-order constraints. We train models to communicate about paths in a simple gridworld, using miniature languages that reflect or violate various natural language trends, such as the tendency to avoid redundancy or to minimize long-distance dependencies. We study how the controlled characteristics of our miniature languages affect individual learning and their stability across multiple network generations. The results draw a mixed picture. On the one hand, neural networks show a strong tendency to avoid long-distance dependencies. On the other hand, there is no clear preference for the efficient, non-redundant encoding of information that is widely attested in natural language. We thus suggest inoculating a notion of “effort” into neural networks, as a possible way to make their linguistic behavior more human-like.

Recurrent neural networks (RNNs) have reached striking performance in many natural language processing tasks. This has renewed interest in whether these generic sequence processing devices are inducing genuine linguistic knowledge. Nearly all current analytical studies, however, initialize the RNNs with a vocabulary of known words, and feed them tokenized input during training. We present a multi-lingual study of the linguistic knowledge encoded in RNNs trained as character-level language models, on input data with word boundaries removed. These networks face a tougher and more cognitively realistic task, having to discover any useful linguistic unit from scratch based on input statistics. The results show that our “near tabula rasa” RNNs are mostly able to solve morphological, syntactic and semantic tasks that intuitively presuppose word-level knowledge, and indeed they learned, to some extent, to track word boundaries. Our study opens the door to speculations about the necessity of an explicit, rigid word lexicon in language learning and usage.

Lake and Baroni (2018) recently introduced the SCAN data set, which consists of simple commands paired with action sequences and is intended to test the strong generalization abilities of recurrent sequence-to-sequence models. Their initial experiments suggested that such models may fail because they lack the ability to extract systematic rules. Here, we take a closer look at SCAN and show that it does not always capture the kind of generalization that it was designed for. To mitigate this we propose a complementary dataset, which requires mapping actions back to the original commands, called NACS. We show that models that do well on SCAN do not necessarily do well on NACS, and that NACS exhibits properties more closely aligned with realistic use-cases for sequence-to-sequence models.

Systematic compositionality is the ability to recombine meaningful units with regular and predictable outcomes, and it’s seen as key to the human capacity for generalization in language. Recent work (Lake and Baroni, 2018) has studied systematic compositionality in modern seq2seq models using generalization to novel navigation instructions in a grounded environment as a probing tool. Lake and Baroni’s main experiment required the models to quickly bootstrap the meaning of new words. We extend this framework here to settings where the model needs only to recombine well-trained functional words (such as “around” and “right”) in novel contexts. Our findings confirm and strengthen the earlier ones: seq2seq models can be impressively good at generalizing to novel combinations of previously-seen input, but only when they receive extensive training on the specific pattern to be generalized (e.g., generalizing from many examples of “X around right” to “jump around right”), while failing when generalization requires novel application of compositional rules (e.g., inferring the meaning of “around right” from those of “right” and “around”).

There is growing interest in the language developed by agents interacting in emergent-communication settings. Earlier studies have focused on the agents’ symbol usage, rather than on their representation of visual input. In this paper, we consider the referential games of Lazaridou et al. (2017), and investigate the representations the agents develop during their evolving interaction. We find that the agents establish successful communication by inducing visual representations that almost perfectly align with each other, but, surprisingly, do not capture the conceptual properties of the objects depicted in the input images. We conclude that, if we care about developing language-like communication systems, we must pay more attention to the visual semantics agents associate to the symbols they use.

Recurrent neural networks (RNNs) achieved impressive results in a variety of linguistic processing tasks, suggesting that they can induce non-trivial properties of language. We investigate to what extent RNNs learn to track abstract hierarchical syntactic structure. We test whether RNNs trained with a generic language modeling objective in four languages (Italian, English, Hebrew, Russian) can predict long-distance number agreement in various constructions. We include in our evaluation nonsensical sentences where RNNs cannot rely on semantic or lexical cues (“The colorless green ideas I ate with the chair sleep furiously”), and, for Italian, we compare model performance to human intuitions. Our language-model-trained RNNs make reliable predictions about long-distance agreement, and do not lag much behind human performance. We thus bring support to the hypothesis that RNNs are not just shallow-pattern extractors, but they also acquire deeper grammatical competence.

Although much effort has recently been devoted to training high-quality sentence embeddings, we still have a poor understanding of what they are capturing. “Downstream” tasks, often based on sentence classification, are commonly used to evaluate the quality of sentence representations. The complexity of the tasks makes it however difficult to infer what kind of information is present in the representations. We introduce here 10 probing tasks designed to capture simple linguistic features of sentences, and we use them to study embeddings generated by three different encoders trained in eight distinct ways, uncovering intriguing properties of both encoders and training methods.

Distributional semantics models are known to struggle with small data. It is generally accepted that in order to learn ‘a good vector’ for a word, a model must have sufficient examples of its usage. This contradicts the fact that humans can guess the meaning of a word from a few occurrences only. In this paper, we show that a neural language model such as Word2Vec only necessitates minor modifications to its standard architecture to learn new terms from tiny data, using background knowledge from a previously learnt semantic space. We test our model on word definitions and on a nonce task involving 2-6 sentences’ worth of context, showing a large increase in performance over state-of-the-art models on the definitional task.

As automated image analysis progresses, there is increasing interest in richer linguistic annotation of pictures, with attributes of objects (e.g., furry, brown…) attracting most attention. By building on the recent “zero-shot learning” approach, and paying attention to the linguistic nature of attributes as noun modifiers, and specifically adjectives, we show that it is possible to tag images with attribute-denoting adjectives even when no training data containing the relevant annotation are available. Our approach relies on two key observations. First, objects can be seen as bundles of attributes, typically expressed as adjectival modifiers (a dog is something furry, brown, etc.), and thus a function trained to map visual representations of objects to nominal labels can implicitly learn to map attributes to adjectives. Second, objects and attributes come together in pictures (the same thing is a dog and it is brown). We can thus achieve better attribute (and object) label retrieval by treating images as “visual phrases”, and decomposing their linguistic representation into an attribute-denoting adjective and an object-denoting noun. Our approach performs comparably to a method exploiting manual attribute annotation, it out-performs various competitive alternatives in both attribute and object annotation, and it automatically constructs attribute-centric representations that significantly improve performance in supervised object recognition.

Corpus-based distributional semantic models capture degrees of semantic relatedness among the words of very large vocabularies, but have problems with logical phenomena such as entailment, that are instead elegantly handled by model-theoretic approaches, which, in turn, do not scale up. We combine the advantages of the two views by inducing a mapping from distributional vectors of words (or sentences) into a Boolean structure of the kind in which natural language terms are assumed to denote. We evaluate this Boolean Distributional Semantic Model (BDSM) on recognizing entailment between words and sentences. The method achieves results comparable to a state-of-the-art SVM, degrades more gracefully when less training data are available and displays interesting qualitative properties.

The lexicon of any natural language encodes a huge number of distinct word meanings. Just to understand this article, you will need to know what thousands of words mean. The space of possible sentential meanings is infinite: In this article alone, you will encounter many sentences that express ideas you have never heard before, we hope. Statistical semantics has addressed the issue of the vastness of word meaning by proposing methods to harvest meaning automatically from large collections of text (corpora). Formal semantics in the Fregean tradition has developed methods to account for the infinity of sentential meaning based on the crucial insight of compositionality, the idea that meaning of sentences is built incrementally by combining the meanings of their constituents. This article sketches a new approach to semantics that brings together ideas from statistical and formal semantics to account, in parallel, for the richness of lexical meaning and the combinatorial power of sentential semantics. We adopt, in particular, the idea that word meaning can be approximated by the patterns of co-occurrence of words in corpora from statistical semantics, and the idea that compositionality can be captured in terms of a syntax-driven calculus of function application from formal semantics.

Shared and internationally recognized benchmarks are fundamental for the development of any computational system. We aim to help the research community working on compositional distributional semantic models (CDSMs) by providing SICK (Sentences Involving Compositional Knowldedge), a large size English benchmark tailored for them. SICK consists of about 10,000 English sentence pairs that include many examples of the lexical, syntactic and semantic phenomena that CDSMs are expected to account for, but do not require dealing with other aspects of existing sentential data sets (idiomatic multiword expressions, named entities, telegraphic language) that are not within the scope of CDSMs. By means of crowdsourcing techniques, each pair was annotated for two crucial semantic tasks: relatedness in meaning (with a 5-point rating scale as gold score) and entailment relation between the two elements (with three possible gold labels: entailment, contradiction, and neutral). The SICK data set was used in SemEval-2014 Task 1, and it freely available for research purposes.

There is by now widespread agreement that the most realistic way to construct the large-scale commonsense knowledge repositories required by natural language and artificial intelligence applications is by letting machines learn such knowledge from large quantities of data, like humans do. A lot of attention has consequently been paid to the development of increasingly sophisticated machine learning algorithms for knowledge extraction. However, the nature of the input that humans are exposed to while learning commonsense knowledge has received much less attention. The BabyExp project is collecting very dense audio and video recordings of the first 3 years of life of a baby. The corpus constructed in this way will be transcribed with automated techniques and made available to the research community. Moreover, techniques to extract commonsense conceptual knowledge incrementally from these multimodal data are also being explored within the project. The current paper describes BabyExp in general, and presents pilot studies on the feasibility of the automated audio and video transcriptions.

Cleaneval is a shared task and competitive evaluation on the topic of cleaning arbitrary web pages, with the goal of preparing web data for use as a corpus for linguistic and language technology research and development. The first exercise took place in 2007. We describe how it was set up, results, and lessons learnt