Zero-shot multi-label text classification (ZMTC) is a fundamental task in natural language processing with applications in the cold start problem of recommendation systems. Ideally, one would learn an expressive representation of both input text and label features so that ZMTC is transformed into a nearest neighbor search problem. However, the existing representation learning approaches for ZMTC struggle with accuracy as well as poor training efficiency. Firstly, the input text is structural, consisting of both short title sentences and long content paragraphs. It is challenging to model the correlation between short label descriptions and long structural input documents. Secondly, the enormous label space in ZMTC forces the existing approaches to perform multi-stage learning with label engineering. As a result, the training overhead is significant. In this paper, we address both problems by introducing an end-to-end structural contrastive representation learning approach. We propose a randomized text segmentation (RTS) technique to generate high-quality contrastive pairs. This RTS technique allows us to model title-content correlation. Additionally, we simplify the multi-stage ZMTC learning strategy by avoiding label engineering. Extensive experiments demonstrate that our approach leads to up to 2.33% improvement in precision@1 and 5.94x speedup in training time on publicly available datasets. Our code is available publicly.

Natural language interaction is a promising direction for democratizing 3D shape design. However, existing methods for text-driven 3D shape editing face challenges in producing decoupled, local edits to 3D shapes. We address this problem by learning disentangled latent representations that ground language in 3D geometry. To this end, we propose a complementary tool set including a novel network architecture, a disentanglement loss, and a new editing procedure. Additionally, to measure edit locality, we define a new metric that we call part-wise edit precision. We show that our method outperforms existing SOTA methods by 20% in terms of edit locality, and up to 6.6% in terms of language reference resolution accuracy. Human evaluations additionally show that compared to the existing SOTA, our method produces shape edits that are more local, more semantically accurate, and more visually obvious. Our work suggests that by solely disentangling language representations, downstream 3D shape editing can become more local to relevant parts, even if the model was never given explicit part-based supervision.

This paper introduces the shared task of summrizing documents in several creative domains, namely literary texts, movie scripts, and television scripts. Summarizing these creative documents requires making complex literary interpretations, as well as understanding non-trivial temporal dependencies in texts containing varied styles of plot development and narrative structure. This poses unique challenges and is yet underexplored for text summarization systems. In this shared task, we introduce four sub-tasks and their corresponding datasets, focusing on summarizing books, movie scripts, primetime television scripts, and daytime soap opera scripts. We detail the process of curating these datasets for the task, as well as the metrics used for the evaluation of the submissions. As part of the CREATIVESUMM workshop at COLING 2022, the shared task attracted 18 submissions in total. We discuss the submissions and the baselines for each sub-task in this paper, along with directions for facilitating future work.

We study how masking and predicting tokens in an unsupervised fashion can give rise to linguistic structures and downstream performance gains. Recent theories have suggested that pretrained language models acquire useful inductive biases through masks that implicitly act as cloze reductions for downstream tasks. While appealing, we show that the success of the random masking strategy used in practice cannot be explained by such cloze-like masks alone. We construct cloze-like masks using task-specific lexicons for three different classification datasets and show that the majority of pretrained performance gains come from generic masks that are not associated with the lexicon. To explain the empirical success of these generic masks, we demonstrate a correspondence between the Masked Language Model (MLM) objective and existing methods for learning statistical dependencies in graphical models. Using this, we derive a method for extracting these learned statistical dependencies in MLMs and show that these dependencies encode useful inductive biases in the form of syntactic structures. In an unsupervised parsing evaluation, simply forming a minimum spanning tree on the implied statistical dependence structure outperforms a classic method for unsupervised parsing (58.74 vs. 55.91 UUAS).