CGF: Constrained Generation Framework for Query Rewriting in Conversational AI

In conversational AI agents, Query Rewriting 001 (QR) plays a crucial role in reducing users fric- 002 tions and satisfying their daily demands. Users 003 frictions are caused by various reasons, such 004 as errors in the spoken dialogue system, users’ 005 accent or their abridged language. In this work, 006 we present a novel Constrained Generation 007 Framework (CGF) for query rewriting at both 008 global and personalized level. The proposed 009 framework is based on the encoder-decoder 010 framework and consists of a context-enhanced 011 encoding and constrained generation decoding 012 phrases. The model takes the query and its 013 previous dialogue context information as the 014 encoder input, then the decoder relies on the 015 pre-defined global or personalized constrained 016 decoding space to generate the rewrites. Ex- 017 tensive offline and online A/B experimental re- 018 sults show that the proposed CGF significantly 019 boosts the query rewriting performance. 020

[1]  Nikit Begwani,et al.  Diversity driven Query Rewriting in Search Advertising , 2021, KDD.

[2]  Kentaro Inui,et al.  Encoder-Decoder Models Can Benefit from Pre-trained Masked Language Models in Grammatical Error Correction , 2020, ACL.

[3]  S. Longpre,et al.  Question Rewriting for Conversational Question Answering , 2020, WSDM.

[4]  Danqi Chen,et al.  Dense Passage Retrieval for Open-Domain Question Answering , 2020, EMNLP.

[5]  Lambert Mathias,et al.  Pre-Training for Query Rewriting in a Spoken Language Understanding System , 2020, ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[6]  Ruhi Sarikaya,et al.  Feedback-Based Self-Learning in Large-Scale Conversational AI Agents , 2019, AAAI.

[7]  Omer Levy,et al.  BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension , 2019, ACL.

[8]  Wei Zhao,et al.  Denoising based Sequence-to-Sequence Pre-training for Text Generation , 2019, EMNLP.

[9]  Arpit Gupta,et al.  Scaling Multi-Domain Dialogue State Tracking via Query Reformulation , 2019, NAACL.

[10]  Wei Zhao,et al.  Improving Grammatical Error Correction via Pre-Training a Copy-Augmented Architecture with Unlabeled Data , 2019, NAACL.

[11]  Xianchao Wu,et al.  Dialog Generation Using Multi-Turn Reasoning Neural Networks , 2018, NAACL.

[12]  Lukasz Kaiser,et al.  Attention is All you Need , 2017, NIPS.

[13]  Andy Way,et al.  Exploiting Cross-Sentence Context for Neural Machine Translation , 2017, EMNLP.

[14]  Qun Liu,et al.  Lexically Constrained Decoding for Sequence Generation Using Grid Beam Search , 2017, ACL.

[15]  Jeff Johnson,et al.  Billion-Scale Similarity Search with GPUs , 2017, IEEE Transactions on Big Data.

[16]  Dingcheng Li,et al.  Contextual Rephrase Detection for Reducing Friction in Dialogue Systems , 2021, EMNLP.

[17]  Xing Fan,et al.  Search based Self-LearningQuery Rewrite System in Conversational AI , 2021 .

[18]  Eunah Cho,et al.  Personalized Search-based Query Rewrite System for Conversational AI , 2021, NLP4CONVAI.

[19]  Chenlei Guo,et al.  RoBERTaIQ: An Efficient Framework for Automatic Interaction Quality Estimation of Dialogue Systems , 2021 .

[20]  Zhaopeng Tu,et al.  RAST: Domain-Robust Dialogue Rewriting as Sequence Tagging , 2021, EMNLP.

[21]  Ming-Wei Chang,et al.  BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , 2019, NAACL.

[22]  Pushpendre Rastogi,et al.  Scaling Multi-Domain Dialogue State Tracking via Query Reformulation , 2019, NAACL.