PROTAUGMENT: Unsupervised diverse short-texts paraphrasing for intent detection meta-learning

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 2454–2466 August 1–6, 2021. ©2021 Association for Computational Linguistics 2454 PROTAUGMENT: Unsupervised diverse short-texts paraphrasing for intent detection meta-learning Thomas Dopierre12 and Christophe Gravier1 and and Wilfried Logerais2 1Laboratoire Hubert Curien UMR CNRS 5516 Universit´e Jean Monnet 2Meetic Paris, France Saint-´Etienne, France firstname.lastname@univ-st-etienne.fr {t.dopierre,w.logerais}@meetic-corp.com Abstract Recent research considers few-shot intent de- tection as a meta-learning problem: the model is learning to learn from a consecutive set of small tasks named episodes. In this work, we propose PROTAUGMENT, a meta-learning algorithm for short texts classification ap- plied to the intent detection task. PROTAUG- MENT is a novel extension of Prototypical Net- works (Snell et al., 2017) that limits over-fitting on the bias introduced by the few-shots classi- fication objective at each episode. It relies on diverse paraphrasing: a conditional language model is first fine-tuned for paraphrasing, and diversity is later introduced at the decoding stage at each meta-learning episode. The di- verse paraphrasing is unsupervised as it is ap- plied to unlabelled data and then fueled to the Prototypical Network training objective as a consistency loss. PROTAUGMENT is the state- of-the-art method for intent detection meta- learning, at no extra labeling efforts and with- out the need to fine-tune a conditional language model on a given application domain. 1 Introduction Intent detection, a sub-field of text classification, involves classifying user-generated short-texts into intent classes, usually for conversational agents applications (Casanueva et al., 2020). Since con- versational agent applications are domain-specific, intent detection is a challenging task because of labeled data scarcity and the number of classes (in- tents) it usually involves (Dopierre et al., 2020). As a consequence, recent research (Snell et al., 2017; Ren et al., 2018) considers few-shot intent detection as a meta-learning problem: the model is trained to classify user utterances from a con- secutive set of small tasks named episodes. Each episode contains a limited number of C classes alongside a limited number of K labeled data for each of the C classes – this is usually referred to as a C-way K-shots setup. At test time, the algorithm is evaluated on classes that were not seen during training. That is the reason why meta-learning is sometimes referred to as learning to learn: it mimics human abilities to learn iteratively from different and small tasks. Meta-learning has suc- cessfully been applied to a wide set of NLP tasks: hypernym detection (Yu et al., 2020), low resource machine translation (Gu et al., 2018), machine un- derstanding tasks (Dou et al., 2019) or structured query generation (Huang et al., 2018). Most meta- learning algorithms (Section 2) were developed in the course of the last 5 years. It has recently been empirically demonstrated that comparative studies in follow-up papers of (Snell et al., 2017) are debat- able – for short texts classification – because of the two following main issues (Dopierre et al., 2021). First, comparative studies …

Content truncated to 500 words