Improving Neural Response Diversity with Frequency-Aware Cross-Entropy Loss

Jiang, Shaojie; Ren, Pengjie; Monz, Christof; Rijke, Maarten de

doi:10.1145/3308558.3313415

Cited by 57 publications

(72 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, such methods tend to generate non-informative answers such as “Thank you” and “I have no idea”. To generate more informative answers, researches introduced external knowledge [ 21 ] or adjusted the objective function [ 22 ].…”

Section: Related Workmentioning

confidence: 99%

Integrate Candidate Answer Extraction with Re-Ranking for Chinese Machine Reading Comprehension

Zeng

Sun

Zhang

et al. 2021

Entropy

View full text Add to dashboard Cite

Machine Reading Comprehension (MRC) research concerns how to endow machines with the ability to understand given passages and answer questions, which is a challenging problem in the field of natural language processing. To solve the Chinese MRC task efficiently, this paper proposes an Improved Extraction-based Reading Comprehension method with Answer Re-ranking (IERC-AR), consisting of a candidate answer extraction module and a re-ranking module. The candidate answer extraction module uses an improved pre-training language model, RoBERTa-WWM, to generate precise word representations, which can solve the problem of polysemy and is good for capturing Chinese word-level features. The re-ranking module re-evaluates candidate answers based on a self-attention mechanism, which can improve the accuracy of predicting answers. Traditional machine-reading methods generally integrate different modules into a pipeline system, which leads to re-encoding problems and inconsistent data distribution between the training and testing phases; therefore, this paper proposes an end-to-end model architecture for IERC-AR to reasonably integrate the candidate answer extraction and re-ranking modules. The experimental results on the Les MMRC dataset show that IERC-AR outperforms state-of-the-art MRC approaches.

show abstract

Section: Related Workmentioning

confidence: 99%

Integrate Candidate Answer Extraction with Re-Ranking for Chinese Machine Reading Comprehension

Zeng

Sun

Zhang

et al. 2021

Entropy

View full text Add to dashboard Cite

show abstract

“…To discourage the generation of frequently occurring words, [68] proposed a frequency aware cross-entropy function (24) whereby they incorporated a weighting mechanism conditioned on the token frequency whereby frequent words will get lower weights and vice-versa. w i is the weight for y t which is calculated based on frequency in the training set.…”

Section: ) Alternative Loss Function Learningmentioning

confidence: 99%

Enhancements to the Sequence-to-Sequence-Based Natural Answer Generation Models

et al. 2020

View full text Add to dashboard Cite

There is a great interest shown by academic researchers to continuously improve the sequence-to-sequence (Seq2Seq) model for natural answer generation (NAG) in chatbots. The Seq2Seq model shows a weakness whereby the model tends to generate answers that are generic, meaningless and inconsistent with the questions. However, a comprehensive literature review on the factors contributing to the weakness and potential solutions are still missing. Therefore, this review article fills the gap by reviewing Seq2Seq based natural answer generation-based literature to identify those factors and proposed methods to address the weakness. This literature review identified several factors such as input question is not sufficient to determine a meaningful output, usage of cross-entropy function as the loss function during training, infrequent words in training data, language model influence which generates answers not relevant to the question, utilization of teacher forcing method during training which results in exposure bias, long sentences and inability to consider dialogue history as the factors. Additionally, this literature review also identified and reviewed the methods proposed to address the weakness such as utilizing additional embedding and encoders, using different loss functions and training approaches, as well as utilizing other mechanisms like copying source word(s) and paying attention to a certain portion of the input. For discussion, these methods are categorized into four broad categories which are Structural Modifications, Augmented Learning, Beam Search and Complementary Mechanisms. Additionally, the paper highlights unexplored areas in Seq2Seq modeling and proposes potential future works for natural answer generation. INDEX TERMS Seq2Seq, natural answer generation, natural language processing, dialogue generation, chatbot.

show abstract

“…therefore, focused on making the responses diverse [10,11]. Our current work is greatly motivated from these prior works, but our focus is on building a multimodal dialogue system.…”

Section: Plos Onementioning

confidence: 99%

“…In [32], the authors proposed a reinforcement learning-based approach which considers a set of responses jointly and generates multiple diverse responses simultaneously. The authors in [11] propose a Frequency-Aware Cross-Entropy (FACE) loss function for generating diverse responses by incorporating a weighting mechanism conditioned on token frequency. In [33], the authors proposed an easyto-extend learning framework named MEMD (Multi-Encoder to Multi-Decoder), in which an auxiliary encoder and an auxiliary decoder are introduced to provide essential training guidance for generating diverse responses.…”

Section: Unimodal Dialogue Systemsmentioning

confidence: 99%

More to diverse: Generating diversified responses in a task oriented multimodal dialog system

2020

View full text Add to dashboard Cite

Multimodal dialogue system, due to its many-fold applications, has gained much attention to the researchers and developers in recent times. With the release of large-scale multimodal dialog dataset Saha et al. 2018 on the fashion domain, it has been possible to investigate the dialogue systems having both textual and visual modalities. Response generation is an essential aspect of every dialogue system, and making the responses diverse is an important problem. For any goal-oriented conversational agent, the system’s responses must be informative, diverse and polite, that may lead to better user experiences. In this paper, we propose an end-to-end neural framework for generating varied responses in a multimodal dialogue setup capturing information from both the text and image. Multimodal encoder with co-attention between the text and image is used for focusing on the different modalities to obtain better contextual information. For effective information sharing across the modalities, we combine the information of text and images using the BLOCK fusion technique that helps in learning an improved multimodal representation. We employ stochastic beam search with Gumble Top K-tricks to achieve diversified responses while preserving the content and politeness in the responses. Experimental results show that our proposed approach performs significantly better compared to the existing and baseline methods in terms of distinct metrics, and thereby generates more diverse responses that are informative, interesting and polite without any loss of information. Empirical evaluation also reveals that images, while used along with the text, improve the efficiency of the model in generating diversified responses.

show abstract

Improving Neural Response Diversity with Frequency-Aware Cross-Entropy Loss

Cited by 57 publications

References 18 publications

Integrate Candidate Answer Extraction with Re-Ranking for Chinese Machine Reading Comprehension

Integrate Candidate Answer Extraction with Re-Ranking for Chinese Machine Reading Comprehension

Enhancements to the Sequence-to-Sequence-Based Natural Answer Generation Models

More to diverse: Generating diversified responses in a task oriented multimodal dialog system

Contact Info

Product

Resources

About