An official website of the United States government
In standard models of language production or comprehension, the elements which are retrieved from memory and combined into a syntactic structure are “lemmas” or “lexical items.” Such models implicitly take a “lexicalist” approach, which assumes that lexical items store meaning, syntax, and form together, that syntactic and lexical processes are distinct, and that syntactic structure does not extend below the word level. Across the last several decades, linguistic research examining a typologically diverse set of languages has provided strong evidence against this approach. These findings suggest that syntactic processes apply both above and below the “word” level, and that both meaning and form are partially determined by the syntactic context. This has significant implications for psychological and neurological models of language processing as well as for the way that we understand different types of aphasia and other language disorders. As a consequence of the lexicalist assumptions of these models, many kinds of sentences that speakers produce and comprehend—in a variety of languages, including English—are challenging for them to account for. Here we focus on language production as a case study. In order to move away from lexicalism in psycho- and neuro-linguistics, it is not enough to simply update the syntactic representations of words or phrases; the processing algorithms involved in language production are constrained by the lexicalist representations that they operate on, and thus also need to be reimagined. We provide an overview of the arguments against lexicalism, discuss how lexicalist assumptions are represented in models of language production, and examine the types of phenomena that they struggle to account for as a consequence. We also outline what a non-lexicalist alternative might look like, as a model that does not rely on a lemma representation, but instead represents that knowledge as separate mappings between (a) meaning and syntax and (b) syntax and form, with a single integrated stage for the retrieval and assembly of syntactic structure. By moving away from lexicalist assumptions, this kind of model provides better cross-linguistic coverage and aligns better with contemporary syntactic theory.
more »
« less
This content will become publicly available on August 19, 2027
SYNTAX-GUIDED DIFFUSION LANGUAGE MODELS WITH USER-INTEGRATED PERSONALIZATION
Large language models have achieved remarkable progress in generating human-like text, yet their outputs often lack structural diversity, limiting personalized expression. Recent advances in diffusion models offer new opportunities for text generation, surpassing the limitations of autoregressive paradigms. Building on the idea of hierarchical modeling, we develop a syntax-guided diffusion language model that incorporates latent syntactic variables to enhance text quality, diversity, and personalized conditioning. The proposed framework decomposes the text distribution into a syntactic prior and a conditional text distribution, enabling interpretable control over both syntactic and lexical components. We design a cascaded diffusion process for syntax-to-text generation and extend it to a generalized noncascaded formulation with a unified attention mechanism. To achieve fine-grained personalization, we introduce shared-latent representations that integrate information across users to capture style-specific lexical and syntactic patterns, supporting zero-shot inference. Extensive experiments demonstrate that the proposed approach consistently improves multiple empirical measures of fluency, diversity, and stylistic fidelity. Further qualitative analyses highlight its interpretability and flexibility in learning personalized patterns.
more »
« less
- Award ID(s):
- 2401271
- PAR ID:
- 10707604
- Publisher / Repository:
- Institute of Mathematical Statistics
- Date Published:
- Journal Name:
- Annals of applied statistics
- ISSN:
- 1941-7330
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Learning target side syntactic structure has been shown to improve Neural Machine Translation (NMT). However, incorporating syntax through latent variables introduces additional complexity in inference, as the models need to marginalize over the latent syntactic structures. To avoid this, models often resort to greedy search which only allows them to explore a limited portion of the latent space. In this work, we introduce a new latent variable model, LaSyn, that captures the co-dependence between syntax and semantics, while allowing for effective and efficient inference over the latent space. LaSyn decouples direct dependence between successive latent variables, which allows its decoder to exhaustively search through the latent syntactic choices, while keeping decoding speed proportional to the size of the latent variable vocabulary. We implement LaSyn by modifying a transformer-based NMT system and design a neural expectation maximization algorithm that we regularize with part-of-speech information as the latent sequences. Evaluations on four different MT tasks show that incorporating target side syntax with LaSyn improves both translation quality, and also provides an opportunity to improve diversity.more » « less
-
Despite the growing success of diffusion models in continuous-valued domains (e.g., images), similar efforts for discrete domains such as text have yet to match the performance of autoregressive language models. In this work, we present SSD-LM—a diffusion-based language model with two key design choices. First, SSD-LM is semi-autoregressive, iteratively generating blocks of text, allowing for flexible output length at decoding time while enabling local bidirectional context updates. Second, it is simplex-based, performing diffusion on the natural vocabulary space rather than a learned latent space, allowing us to incorporate classifier guidance and modular control using off-the-shelf classifiers without any adaptation. We evaluate SSD-LM on unconstrained text generation benchmarks, and show that it matches or outperforms strong autoregressive GPT-2 models across standard quality and diversity metrics, while vastly outperforming diffusion-based baselines. On controlled text generation, SSD-LM also outperforms competitive baselines, with an extra advantage in modularity.more » « less
-
Latent variable models for text, when trained successfully, accurately model the data distribution and capture global semantic and syntactic features of sentences. The prominent approach to train such models is variational autoencoders (VAE). It is nevertheless challenging to train and often results in a trivial local optimum where the latent variable is ignored and its posterior collapses into the prior, an issue known as posterior collapse. Various techniques have been proposed to mitigate this issue. Most of them focus on improving the inference model to yield latent codes of higher quality. The present work proposes a short run dynamics for inference. It is initialized from the prior distribution of the latent variable and then runs a small number (e.g., 20) of Langevin dynamics steps guided by its posterior distribution. The major advantage of our method is that it does not require a separate inference model or assume simple geometry of the posterior distribution, thus rendering an automatic, natural and flexible inference engine. We show that the models trained with short run dynamics more accurately model the data, compared to strong language model and VAE baselines, and exhibit no sign of posterior collapse. Analyses of the latent space show that interpolation in the latent space is able to generate coherent sentences with smooth transition and demonstrate improved classification over strong baselines with latent features from unsupervised pretraining. These results together expose a well-structured latent space of our generative model.more » « less
-
We conduct a large-scale, systematic study to evaluate the existing evaluation methods for natural language generation in the context of generating online product reviews. We compare human-based evaluators with a variety of automated evaluation procedures, including discriminative evaluators that measure how well machine-generated text can be distinguished from human-written text, as well as word overlap metrics that assess how similar the generated text compares to human-written references. We determine to what extent these different evaluators agree on the ranking of a dozen of state-of-the-art generators for online product reviews. We find that human evaluators do not correlate well with discriminative evaluators, leaving a bigger question of whether adversarial accuracy is the correct objective for natural language generation. In general, distinguishing machine-generated text is challenging even for human evaluators, and human decisions correlate better with lexical overlaps. We find lexical diversity an intriguing metric that is indicative of the assessments of different evaluators. A post-experiment survey of participants provides insights into how to evaluate and improve the quality of natural language generation systems.more » « less
