Planning and Generating Natural and Diverse Disfluent Texts as Augmentation for Disfluency Detection

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 1450–1460, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 1450 Planning and Generating Natural and Diverse Disfluent Texts as Augmentation for Disfluency Detection Jingfeng Yang, Diyi Yang, Zhaoran Ma Georgia Institute of Technology {jyang690, makima, dyang888}@gatech.edu Abstract Existing approaches to disfluency detection heavily depend on human-annotated data. Numbers of data augmentation methods have been proposed to alleviate the dependence on labeled data. However, current augmentation approaches such as random insertion or repetition fail to resemble training corpus well and usually resulted in unnatural and limited types of disfluencies. In this work, we propose a simple Planner-Generator based disfluency generation model to generate natural and diverse disfluent texts as aug- mented data, where the Planner decides on where to insert disfluent segments and the Generator follows the prediction to generate corresponding disfluent segments. We further utilize this augmented data for pretraining and leverage it for the task of disfluency detection. Experiments demonstrated that our two-stage disfluency generation model outperforms existing baselines; those disflu- ent sentences generated significantly aided the task of disfluency detection and led to state-of-the-art performance on Switchboard corpus. We have publicly released our code at https://github.com/GT-SALT/ Disfluency-Generation-and-Detection. 1 Introduction Disfluency is a para-linguistic concept defining the interruption to the flow of speech (Kowal, 2009). As shown in Figure 1, a standard annotation of the disfluency structure indicates the reparandum (the region to repair), an optional interregnum (filled pauses, discourse cue words, etc.) and the associated repair (corrected linguistic materials) (Nakatani and Hirschberg, 1994). Disfluency de- tection (Lou et al., 2018; Wang et al., 2019) mainly deals with identifying and removing reparandum1, 1We use reparandum and disfluent segments interchange- ably in this paper. I want a flight [ to Boston + {um} to Denver ] reparandum interregnum repair Figure 1: Example of an annotated disfluent sentence. Type Example Repetition they they learn to share. Deletion this is just happened yesterday. Substitution it’s nothing but wood up here down here. Table 1: Different types of reparandum. since interregnum can be easily detected in that they belong to a closed set of words and phrases, e.g. “uh” “I mean” “you know” etc. The output flu- ent sentences from disfluency detection can serve as clean inputs for most downstream NLP tasks, like dialogue systems, question answering, and ma- chine translation (Wang et al., 2010). Reparandum in disfluency can be categorized as repetition, deletion and substitution (McDougall and Duckworth, 2017), as shown in Table 1. Repe- tition occurs when linguistic materials repeat, usu- ally in form of partial words, words or short phrases. Substitution occurs when linguistic materials are replaced in order to clarify a concept or idea. Dele- tion, also known as false restart, refers to aban- doned linguistics materials. Neural models have achieved reasonable perfor- mance in disfluency detection on English Switch- board (SWBD) corpus (Godfrey et al., 1992). Such models involve applying pretrained models to con- duct disfluency detection as sequence tagging or seq2seq tasks (Wang et …

Content truncated to 500 words