Recently, large pretrained language models have demonstrated strong language understanding capabilities. This is particularly reflected in their zero-shot and in-context learning abilities on downstream tasks through prompting. To assess their impact on spoken language understanding (SLU), we evaluate several such models like ChatGPT and OPT of different sizes on multiple benchmarks. We verify the emergent ability unique to the largest models as they can reach intent classification accuracy close to that of supervised models with zero or few shots on various languages given oracle transcripts. By contrast, the results for smaller models fitting a single GPU fall far behind. We note that the error cases often arise from the annotation scheme of the dataset; responses from ChatGPT are still reasonable. We show, however, that the model is worse at slot filling, and its performance is sensitive to ASR errors, suggesting serious challenges for the application of those textual models on SLU.
End-to-end spoken language understanding (SLU) remains elusive even with current large pretrained language models on text and speech, especially in multilingual cases. Machine translation has been established as a powerful pretraining objective on text as it enables the model to capture high-level semantics of the input utterance and associations between different languages, which is desired for speech models that work on lower-level acoustic frames. Motivated particularly by the task of cross-lingual SLU, we demonstrate that the task of speech translation (ST) is a good means of pretraining speech models for end-to-end SLU on both monolingual and cross-lingual scenarios. By introducing ST, our models give higher performance over current baselines on monolingual and multilingual intent classification as well as spoken question answering using SLURP, MINDS-14, and NMSQA benchmarks. To verify the effectiveness of our methods, we also release two new benchmark datasets from both synthetic and real sources, for the tasks of abstractive summarization from speech and low-resource or zero-shot transfer from English to French. We further show the value of preserving knowledge from the pretraining task, and explore Bayesian transfer learning on pretrained speech models based on continual learning regularizers for that.
Conceptualization, or viewing entities and situations as instances of abstract concepts in mind and making inferences based on that, is a vital component in human intelligence for commonsense reasoning. Although recent artificial intelligence has made progress in acquiring and modelling commonsense, attributed to large neural language models and commonsense knowledge graphs (CKGs), conceptualization is yet to thoroughly be introduced, making current approaches ineffective to cover knowledge about countless diverse entities and situations in the real world. To address the problem, we thoroughly study the possible role of conceptualization in commonsense reasoning, and formulate a framework to replicate human conceptual induction from acquiring abstract knowledge about abstract concepts. Aided by the taxonomy Probase, we develop tools for contextualized conceptualization on ATOMIC, a large-scale human annotated CKG. We annotate a dataset for the validity of conceptualizations for ATOMIC on both event and triple level, develop a series of heuristic rules based on linguistic features, and train a set of neural models, so as to generate and verify abstract knowledge. Based on these components, a pipeline to acquire abstract knowledge is built. A large abstract CKG upon ATOMIC is then induced, ready to be instantiated to infer about unseen entities or situations. Furthermore, experiments find directly augmenting data with abstract triples to be helpful in commonsense modelling.
End-to-end TTS suffers from high data requirements as it is difficult for both costly speech corpora to cover all necessary knowledge and neural models to learn the knowledge, hence additional knowledge needs to be injected manually. For example, to capture pronunciation knowledge on languages without regular orthography, a complicated grapheme-to-phoneme pipeline needs to be built based on a structured, large pronunciation lexicon, leading to extra, sometimes high, costs to extend neural TTS to such languages. In this paper, we propose a framework to learn to extract knowledge from unstructured external resources using Token2Knowledge attention modules. The framework is applied to build a novel end-to-end TTS model named Neural Lexicon Reader that extracts pronunciations from raw lexicon texts. Experiments support the potential of our framework that the model significantly reduces pronunciation errors in low-resource, end-to-end Chinese TTS, and the lexicon-reading capability can be transferred to other languages with a smaller amount of data.
We present a multilingual end-to-end Text-To-Speech framework that maps byte inputs to spectrograms, thus allowing arbitrary input scripts. Besides strong results on 40+ languages, the framework demonstrates capabilities to adapt to various new languages under extreme low-resource and even few-shot scenarios of merely 40s transcribed recording without the need of lexicon, extra corpus, auxiliary models, or particular linguistic expertise, while retains satisfactory intelligibility and naturalness matching rich-resource models. Exhaustive comparative studies are performed to reveal the potential of the framework for low-resource application and the impact of various factors contributory to adaptation. Furthermore, we propose a novel method to extract language-specific sub-networks for a better understanding of the mechanism of multilingual models.
Commonsense knowledge graphs (CKGs) like Atomic and ASER are substantially different from conventional KGs as they consist of much larger number of nodes formed by loosely-structured text, which, though, enables them to handle highly diverse queries in natural language related to commonsense, leads to unique challenges for automatic KG construction methods. Besides identifying relations absent from the KG between nodes, such methods are also expected to explore absent nodes represented by text, in which different real-world things, or entities, may appear. To deal with the innumerable entities involved with commonsense in the real world, we introduce to CKG construction methods conceptualization, i.e., to view entities mentioned in text as instances of specific concepts or vice versa. We build synthetic triples by conceptualization, and further formulate the task as triple classification, handled by a discriminatory model with knowledge transferred from pretrained language models and fine-tuned by negative sampling. Experiments demonstrate that our methods can effectively identify plausible triples and expand the KG by triples of both new nodes and edges of high diversity and novelty.
In this paper, we study a new graph learning problem: learning to count subgraph isomorphisms. Although the learning based approach is inexact, we are able to generalize to count large patterns and data graphs in polynomial time compared to the exponential time of the original NP-complete problem. Different from other traditional graph learning problems such as node classification and link prediction, subgraph isomorphism counting requires more global inference to oversee the whole graph. To tackle this problem, we propose a dynamic intermedium attention memory network (DIAMNet) which augments different representation learning architectures and iteratively attends pattern and target data graphs to memorize different subgraph isomorphisms for the global counting. We develop both small graphs (<= 1,024 subgraph isomorphisms in each) and large graphs (<= 4,096 subgraph isomorphisms in each) sets to evaluate different models. Experimental results show that learning based subgraph isomorphism counting can help reduce the time complexity with acceptable accuracy. Our DIAMNet can further improve existing representation learning models for this more global problem.
Neural TTS has demonstrated strong capabilities to generate human-like speech with high quality and naturalness, while its generalization to out-of-domain texts is still a challenging task, with regard to the design of attention-based sequence-to-sequence acoustic modeling. Various errors occur in those texts with unseen context, including attention collapse, skipping, repeating, etc., which limits the broader applications. In this paper, we propose a novel stepwise monotonic attention method in sequence-to-sequence acoustic modeling to improve the robustness on out-of-domain texts. The method utilizes the strict monotonic property in TTS with extra constraints on monotonic attention that the alignments between inputs and outputs sequence must be not only monotonic but also allowing no skipping on the inputs. In inference, soft attention could be used to evade mismatch between training and test in monotonic hard attention. The experimental results show that the proposed method could achieve significant improvements in robustness on various out-of-domain scenarios, without any regression on the in-domain test set.