Most existing methods of Out-of-Domain (OOD) intent classification rely on extensive auxiliary OOD corpora or specific training paradigms. However, they are underdeveloped in the underlying principle that the models should have differentiated confidence in In- and Out-of-domain intent. In this work, we shed light on the fundamental cause of model overconfidence on OOD and demonstrate that calibrated subnetworks can be uncovered by pruning the overparameterized model. Calibrated confidence provided by the subnetwork can better distinguish In- and Out-of-domain, which can be a benefit for almost all post hoc methods. In addition to bringing fundamental insights, we also extend the Lottery Ticket Hypothesis to open-world scenarios. We conduct extensive experiments on four real-world datasets to demonstrate our approach can establish consistent improvements compared with a suite of competitive baselines.
Meta-Learning has emerged as a research direction to better transfer knowledge from related tasks to unseen but related tasks. However, Meta-Learning requires many training tasks to learn representations that transfer well to unseen tasks; otherwise, it leads to overfitting, and the performance degenerates to worse than Multi-task Learning. We show that a state-of-the-art data augmentation method worsens this problem of overfitting when the task diversity is low. We propose a simple method, TaskMix, which synthesizes new tasks by linearly interpolating existing tasks. We compare TaskMix against many baselines on an in-house multilingual intent classification dataset of N-Best ASR hypotheses derived from real-life human-machine telephony utterances and two datasets derived from MTOP. We show that TaskMix outperforms baselines, alleviates overfitting when task diversity is low, and does not degrade performance even when it is high.
Few-shot Intent Classification (FSIC) is one of the key challenges in modular task-oriented dialog systems. While advanced FSIC methods are similar in using pretrained language models to encode texts and nearest neighbour-based inference for classification, these methods differ in details. They start from different pretrained text encoders, use different encoding architectures with varying similarity functions, and adopt different training regimes. Coupling these mostly independent design decisions and the lack of accompanying ablation studies are big obstacle to identify the factors that drive the reported FSIC performance. We study these details across three key dimensions: (1) Encoding architectures: Cross-Encoder vs Bi-Encoders; (2) Similarity function: Parameterized (i.e., trainable) functions vs non-parameterized function; (3) Training regimes: Episodic meta-learning vs the straightforward (i.e., non-episodic) training. Our experimental results on seven FSIC benchmarks reveal three important findings. First, the unexplored combination of the cross-encoder architecture (with parameterized similarity scoring function) and episodic meta-learning consistently yields the best FSIC performance. Second, Episodic training yields a more robust FSIC classifier than non-episodic one. Third, in meta-learning methods, splitting an episode to support and query sets is not a must. Our findings paves the way for conducting state-of-the-art research in FSIC and more importantly raise the community's attention to details of FSIC methods. We release our code and data publicly.
Teacher-student knowledge distillation is a popular technique for compressing today's prevailing large language models into manageable sizes that fit low-latency downstream applications. Both the teacher and the choice of transfer set used for distillation are crucial ingredients in creating a high quality student. Yet, the generic corpora used to pretrain the teacher and the corpora associated with the downstream target domain are often significantly different, which raises a natural question: should the student be distilled over the generic corpora, so as to learn from high-quality teacher predictions, or over the downstream task corpora to align with finetuning? Our study investigates this trade-off using Domain Classification (DC) and Intent Classification/Named Entity Recognition (ICNER) as downstream tasks. We distill several multilingual students from a larger multilingual LM with varying proportions of generic and task-specific datasets, and report their performance after finetuning on DC and ICNER. We observe significant improvements across tasks and test sets when only task-specific corpora is used. We also report on how the impact of adding task-specific data to the transfer set correlates with the similarity between generic and task-specific data. Our results clearly indicate that, while distillation from a generic LM benefits downstream tasks, students learn better using target domain data even if it comes at the price of noisier teacher predictions. In other words, target domain data still trumps teacher knowledge.
To proactively offer social media users a safe online experience, there is a need for systems that can detect harmful posts and promptly alert platform moderators. In order to guarantee the enforcement of a consistent policy, moderators are provided with detailed guidelines. In contrast, most state-of-the-art models learn what abuse is from labelled examples and as a result base their predictions on spurious cues, such as the presence of group identifiers, which can be unreliable. In this work we introduce the concept of policy-aware abuse detection, abandoning the unrealistic expectation that systems can reliably learn which phenomena constitute abuse from inspecting the data alone. We propose a machine-friendly representation of the policy that moderators wish to enforce, by breaking it down into a collection of intents and slots. We collect and annotate a dataset of 3,535 English posts with such slots, and show how architectures for intent classification and slot filling can be used for abuse detection, while providing a rationale for model decisions.
We present LINGUIST, a method for generating annotated data for Intent Classification and Slot Tagging (IC+ST), via fine-tuning AlexaTM 5B, a 5-billion-parameter multilingual sequence-to-sequence (seq2seq) model, on a flexible instruction prompt. In a 10-shot novel intent setting for the SNIPS dataset, LINGUIST surpasses state-of-the-art approaches (Back-Translation and Example Extrapolation) by a wide margin, showing absolute improvement for the target intents of +1.9 points on IC Recall and +2.5 points on ST F1 Score. In the zero-shot cross-lingual setting of the mATIS++ dataset, LINGUIST out-performs a strong baseline of Machine Translation with Slot Alignment by +4.14 points absolute on ST F1 Score across 6 languages, while matching performance on IC. Finally, we verify our results on an internal large-scale multilingual dataset for conversational agent IC+ST and show significant improvements over a baseline which uses Back-Translation, Paraphrasing and Slot Catalog Resampling. To our knowledge, we are the first to demonstrate instruction fine-tuning of a large-scale seq2seq model to control the outputs of multilingual intent- and slot-labeled data generation.
Traditional intent classification models are based on a pre-defined intent set and only recognize limited in-domain (IND) intent classes. But users may input out-of-domain (OOD) queries in a practical dialogue system. Such OOD queries can provide directions for future improvement. In this paper, we define a new task, Generalized Intent Discovery (GID), which aims to extend an IND intent classifier to an open-world intent set including IND and OOD intents. We hope to simultaneously classify a set of labeled IND intent classes while discovering and recognizing new unlabeled OOD types incrementally. We construct three public datasets for different application scenarios and propose two kinds of frameworks, pipeline-based and end-to-end for future work. Further, we conduct exhaustive experiments and qualitative analysis to comprehend key challenges and provide new guidance for future GID research.
The use of deep neural networks in electromyogram (EMG) based prostheses control provides a promising alternative to the hand-crafted features by automatically learning muscle activation patterns from the EMG signals. Meanwhile, the use of raw EMG signals as input to convolution neural networks (CNN) offers a simple, fast, and ideal scheme for effective control of prostheses. Therefore, this study investigates the relationship between window length and overlap, which may influence the generation of robust raw EMG 2-dimensional (2D) signals for application in CNN. And a rule of thumb for a proper combination of these parameters that could guarantee optimal network performance was derived. Moreover, we investigate the relationship between the CNN receptive window size and the raw EMG signal size. Experimental results show that the performance of the CNN increases with the increase in overlap within the generated signals, with the highest improvement of 9.49% accuracy and 23.33% F1-score realized when the overlap is 75% of the window length. Similarly, the network performance increases with the increase in receptive window (kernel) size. Findings from this study suggest that a combination of 75% overlap in 2D EMG signals and wider network kernels may provide ideal motor intents classification for adequate EMG-CNN based prostheses control scheme.
Intent discovery is a crucial task in natural language processing, and it is increasingly relevant for various of industrial applications. Identifying novel, unseen intents from user inputs remains one of the biggest challenges in this field. Herein, we propose Zero-Shot-BERT-Adapters, a two-stage method for multilingual intent discovery relying on a Transformer architecture, fine-tuned with Adapters. We train the model for Natural Language Inference (NLI) and later perform unknown intent classification in a zero-shot setting for multiple languages. In our evaluation, we first analyze the quality of the model after adaptive fine-tuning on known classes. Secondly, we evaluate its performance in casting intent classification as an NLI task. Lastly, we test the zero-shot performance of the model on unseen classes, showing how Zero-Shot-BERT-Adapters can effectively perform intent discovery by generating semantically similar intents, if not equal, to the ground-truth ones. Our experiments show how Zero-Shot-BERT-Adapters outperforms various baselines in two zero-shot settings: known intent classification and unseen intent discovery. The proposed pipeline holds the potential for broad application in customer care. It enables automated dynamic triage using a lightweight model that can be easily deployed and scaled in various business scenarios, unlike large language models. Zero-Shot-BERT-Adapters represents an innovative multi-language approach for intent discovery, enabling the online generation of novel intents. A Python package implementing the pipeline and the new datasets we compiled are available at the following link: https://github.com/GT4SD/zero-shot-bert-adapters.
Interest in dialog systems has grown substantially in the past decade. By extension, so too has interest in developing and improving intent classification and slot-filling models, which are two components that are commonly used in task-oriented dialog systems. Moreover, good evaluation benchmarks are important in helping to compare and analyze systems that incorporate such models. Unfortunately, much of the literature in the field is limited to analysis of relatively few benchmark datasets. In an effort to promote more robust analyses of task-oriented dialog systems, we have conducted a survey of publicly available datasets for the tasks of intent classification and slot-filling. We catalog the important characteristics of each dataset, and offer discussion on the applicability, strengths, and weaknesses of each. Our goal is that this survey aids in increasing the accessibility of these datasets, which we hope will enable their use in future evaluations of intent classification and slot-filling models for task-oriented dialog systems.
Task-oriented dialogue systems often employ a Dialogue State Tracker (DST) to successfully complete conversations. Recent state-of-the-art DST implementations rely on schemata of diverse services to improve model robustness and handle zero-shot generalization to new domains [1], however such methods [2, 3] typically require multiple large scale transformer models and long input sequences to perform well. We propose a single multi-task BERT-based model that jointly solves the three DST tasks of intent prediction, requested slot prediction and slot filling. Moreover, we propose an efficient and parsimonious encoding of the dialogue history and service schemata that is shown to further improve performance. Evaluation on the SGD dataset shows that our approach outperforms the baseline SGP-DST by a large margin and performs well compared to the state-of-the-art, while being significantly more computationally efficient. Extensive ablation studies are performed to examine the contributing factors to the success of our model.
In Spoken Language Understanding (SLU) the task is to extract important information from audio commands, like the intent of what a user wants the system to do and special entities like locations or numbers. This paper presents a simple method for embedding intents and entities into Finite State Transducers, and, in combination with a pretrained general-purpose Speech-to-Text model, allows building SLU-models without any additional training. Building those models is very fast and only takes a few seconds. It is also completely language independent. With a comparison on different benchmarks it is shown that this method can outperform multiple other, more resource demanding SLU approaches.
We present results from a large-scale experiment on pretraining encoders with non-embedding parameter counts ranging from 700M to 9.3B, their subsequent distillation into smaller models ranging from 17M-170M parameters, and their application to the Natural Language Understanding (NLU) component of a virtual assistant system. Though we train using 70% spoken-form data, our teacher models perform comparably to XLM-R and mT5 when evaluated on the written-form Cross-lingual Natural Language Inference (XNLI) corpus. We perform a second stage of pretraining on our teacher models using in-domain data from our system, improving error rates by 3.86% relative for intent classification and 7.01% relative for slot filling. We find that even a 170M-parameter model distilled from our Stage 2 teacher model has 2.88% better intent classification and 7.69% better slot filling error rates when compared to the 2.3B-parameter teacher trained only on public data (Stage 1), emphasizing the importance of in-domain data for pretraining. When evaluated offline using labeled NLU data, our 17M-parameter Stage 2 distilled model outperforms both XLM-R Base (85M params) and DistillBERT (42M params) by 4.23% to 6.14%, respectively. Finally, we present results from a full virtual assistant experimentation platform, where we find that models trained using our pretraining and distillation pipeline outperform models distilled from 85M-parameter teachers by 3.74%-4.91% on an automatic measurement of full-system user dissatisfaction.
Training accurate intent classifiers requires labeled data, which can be costly to obtain. Data augmentation methods may ameliorate this issue, but the quality of the generated data varies significantly across techniques. We study the process of systematically producing pseudo-labeled data given a small seed set using a wide variety of data augmentation techniques, including mixing methods together. We find that while certain methods dramatically improve qualitative and quantitative performance, other methods have minimal or even negative impact. We also analyze key considerations when implementing data augmentation methods in production.
Zero-shot intent classification is a vital and challenging task in dialogue systems, which aims to deal with numerous fast-emerging unacquainted intents without annotated training data. To obtain more satisfactory performance, the crucial points lie in two aspects: extracting better utterance features and strengthening the model generalization ability. In this paper, we propose a simple yet effective meta-learning paradigm for zero-shot intent classification. To learn better semantic representations for utterances, we introduce a new mixture attention mechanism, which encodes the pertinent word occurrence patterns by leveraging the distributional signature attention and multi-layer perceptron attention simultaneously. To strengthen the transfer ability of the model from seen classes to unseen classes, we reformulate zero-shot intent classification with a meta-learning strategy, which trains the model by simulating multiple zero-shot classification tasks on seen categories, and promotes the model generalization ability with a meta-adapting procedure on mimic unseen categories. Extensive experiments on two real-world dialogue datasets in different languages show that our model outperforms other strong baselines on both standard and generalized zero-shot intent classification tasks.
Learning high-quality dialogue representations is essential for solving a variety of dialogue-oriented tasks, especially considering that dialogue systems often suffer from data scarcity. In this paper, we introduce Dialogue Sentence Embedding (DSE), a self-supervised contrastive learning method that learns effective dialogue representations suitable for a wide range of dialogue tasks. DSE learns from dialogues by taking consecutive utterances of the same dialogue as positive pairs for contrastive learning. Despite its simplicity, DSE achieves significantly better representation capability than other dialogue representation and universal sentence representation models. We evaluate DSE on five downstream dialogue tasks that examine dialogue representation at different semantic granularities. Experiments in few-shot and zero-shot settings show that DSE outperforms baselines by a large margin. For example, it achieves 13% average performance improvement over the strongest unsupervised baseline in 1-shot intent classification on 6 datasets. We also provide analyses on the benefits and limitations of our model.
Spoken dialog systems are slowly becoming and integral part of the human experience due to their various advantages over textual interfaces. Spoken language understanding (SLU) systems are fundamental building blocks of spoken dialog systems. But creating SLU systems for low resourced languages is still a challenge. In a large number of low resourced language, we don't have access to enough data to build automatic speech recognition (ASR) technologies, which are fundamental to any SLU system. Also, ASR based SLU systems do not generalize to unwritten languages. In this paper, we present a series of experiments to explore extremely low-resourced settings where we perform intent classification with systems trained on as low as one data-point per intent and with only one speaker in the dataset. We also work in a low-resourced setting where we do not use language specific ASR systems to transcribe input speech, which compounds the challenge of building SLU systems to simulate a true low-resourced setting. We test our system on Belgian Dutch (Flemish) and English and find that using phonetic transcriptions to make intent classification systems in such low-resourced setting performs significantly better than using speech features. Specifically, when using a phonetic transcription based system over a feature based system, we see average improvements of 12.37% and 13.08% for binary and four-class classification problems respectively, when averaged over 49 different experimental settings.
This work explores the intrinsic limitations of the popular one-hot encoding method in classification of intents when detection of out-of-scope (OOS) inputs is required. Although recent work has shown that there can be significant improvements in OOS detection when the intent classes are represented as dense-vectors based on domain specific knowledge, we argue in this paper that such gains are more likely due to advantages of dense-vector to one-hot encoding methods in representing the complexity of the OOS space. We start by showing how dense-vector encodings can create OOS spaces with much richer topologies than one-hot encoding methods. We then demonstrate empirically, using four standard intent classification datasets, that knowledge-free, randomly generated dense-vector encodings of intent classes can yield massive, over 20% gains over one-hot encodings, and also outperform the previous, domain knowledge-based, SOTA of one of the datasets. We finish by describing a novel algorithm to search for good dense-vector encodings and present initial but promising experimental results of its use.
A law practitioner has to go through a lot of long legal case proceedings. To understand the motivation behind the actions of different parties/individuals in a legal case, it is essential that the parts of the document that express an intent corresponding to the case be clearly understood. In this paper, we introduce a dataset of 93 legal documents, belonging to the case categories of either Murder, Land Dispute, Robbery, or Corruption, where phrases expressing intent same as the category of the document are annotated. Also, we annotate fine-grained intents for each such phrase to enable a deeper understanding of the case for a reader. Finally, we analyze the performance of several transformer-based models in automating the process of extracting intent phrases (both at a coarse and a fine-grained level), and classifying a document into one of the possible 4 categories, and observe that, our dataset is challenging, especially in the case of fine-grained intent classification.
User intent classification is an important task in information retrieval. In this work, we introduce a revised taxonomy of user intent. We take the widely used differentiation between navigational, transactional and informational queries as a starting point, and identify three different sub-classes for the informational queries: instrumental, factual and abstain. The resulting classification of user queries is more fine-grained, reaches a high level of consistency between annotators, and can serve as the basis for an effective automatic classification process. The newly introduced categories help distinguish between types of queries that a retrieval system could act upon, for example by prioritizing different types of results in the ranking.We have used a weak supervision approach based on Snorkel to annotate the ORCAS dataset according to our new user intent taxonomy, utilising established heuristics and keywords to construct rules for the prediction of the intent category. We then present a series of experiments with a variety of machine learning models, using the labels from the weak supervision stage as training data, but find that the results produced by Snorkel are not outperformed by these competing approaches and can be considered state-of-the-art. The advantage of a rule-based approach like Snorkel's is its efficient deployment in an actual system, where intent classification would be executed for every query issued. The resource released with this paper is the ORCAS-I dataset: a labelled version of the ORCAS click-based dataset of Web queries, which provides 18 million connections to 10 million distinct queries.