Open-Domain Conversational Agents: Current Progress, Open Problems, and Future Directions

Roller, Stephen; Boureau, Y-Lan; Weston, Jason; Bordes, Antoine; Dinan, Emily; Fan, Angela; Gunning, David; Ju, Da; Li, Margaret; Poff, Spencer; Ringshia, Pratik; Shuster, Kurt; Smith, Eric Michael; Szlam, Arthur; Urbanek, Jack; Williamson, Mary

doi:10.48550/arxiv.2006.12442

Cited by 13 publications

(20 citation statements)

References 95 publications

(192 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar trends are observed on the biased set (i.e., 93.75 in bold), which indicates that incorporating the other party's personalized information in the decoding process is beneficial to generate more personalized and more bilateral personaconsistent responses. (5) The proposed different personaaware weights (i.e., α, β, and γ) can be used to control the persona presented in the generated response. The results of the two test sets show that under different context settings, it will improve the effect of personalized response generation with different persona-aware weights.…”

Section: B Results and Analysismentioning

confidence: 99%

See 1 more Smart Citation

Bilateral Personalized Dialogue Generation with Contrastive Learning

Li¹,

Deng²

2021

Preprint

View full text Add to dashboard Cite

Generating personalized responses is one of the major challenges in natural human-robot interaction. Current researches in this field mainly focus on generating responses consistent with the robot's pre-assigned persona, while ignoring the user's persona. Such responses may be inappropriate or even offensive, which may lead to the bad user experience. Therefore, we propose a bilateral personalized dialogue generation (BPDG) method with dynamic persona-aware fusion via multi-task transfer learning to generate responses consistent with both personas. The proposed method aims to accomplish three learning tasks: 1) an encoder is trained with dialogue utterances added with corresponded personalized attributes and relative position (language model task), 2) a dynamic persona-aware fusion module predicts the persona presence to adaptively fuse the contextual and bilateral personas encodings (persona prediction task) and 3) a decoder generates natural, fluent and personalized responses (dialogue generation task). To make the generated responses more personalized and bilateral persona-consistent, the Conditional Mutual Information Maximum (CMIM) criterion is adopted to select the final response from the generated candidates. The experimental results show that the proposed method outperforms several state-of-the-art methods in terms of both automatic and manual evaluations.

show abstract

Section: B Results and Analysismentioning

confidence: 99%

“…Currently, personalized dialogue agents have been widely applied in various human-robot interaction scenarios, such as intelligent personal assistants [2], public service robots [3], wearable devices [4], etc. The agents with personalization are considered reliable and trustworthy, and can gain the user's confidence and trust [5].…”

Section: Introductionmentioning

confidence: 99%

Bilateral Personalized Dialogue Generation with Contrastive Learning

Li¹,

Deng²

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…While the term "hallucination" seems to have newly emerged in the field, a related behavior of neural models has been widely discussed. The behavior commonly known as "inconsistency" has been pointed out as a shortcoming of generation-based approaches for open-domain chatbot [74,109,151]. Two possible types of inconsistency occur in open-domain dialogue generation: (1) inconsistency among the system utterances such as when the system contradicts its previous utterance; (2) inconsistency with some external source, such as factually incorrect utterances.…”

Section: Open-domain Dialogue Generationmentioning

confidence: 99%

“…Under this setting, the study of extrinsic hallucination is of great value but still poorly investigated. [151]. As of now, there is no standard metric.…”

Section: External Consistencymentioning

confidence: 99%

See 1 more Smart Citation

Survey of Hallucination in Natural Language Generation

Ji¹,

Lee²,

Frieske³

et al. 2022

Preprint

View full text Add to dashboard Cite

Natural Language Generation (NLG) has improved exponentially in recent years thanks to the development of sequence-to-sequence deep learning technologies such as Transformer-based language models. This advancement has led to more fluent and coherent natural language generation, leading to improved development in downstream tasks such as abstractive summarization, dialogue generation and data-to-text generation. However, it is also apparent that deep learning based generation is prone to hallucinate unintended texts, which degrades the system performance and fail to meet user expectations in many real-world scenarios. In order to address this issue, there have been studies in measuring and mitigating hallucinated texts. However there has not been a comprehensive review of the state-of-the-art in hallucination detection and mitigation.In this survey, we provide a broad overview of the research progress and challenges in the hallucination problem of NLG. The survey is organized into two parts: (1) a general overview of metrics, mitigation methods, and future directions; (2) an overview of task-specific research progress for hallucinations in a large set of downstream tasks, namely abstractive summarization, dialogue generation, generative question answering, data-to-text generation, and machine translation. This survey serves to facilitate collaborative efforts among researchers in tackling the challenge of hallucinated texts in NLG.

show abstract

Commentary: Future thoughts on ChatGPT and its applications in education

Hasnain

2023

JRIT

View full text Add to dashboard Cite

intelligence" AI application ChatGPT and other forms of generative AI are establishing themselves and are going to revolutionize the way that schooling is now conducted. Proponents of AI technology have begun creating tools to identify AI-generated messages. While opponents of AI technology have urged for the eradication of ChatGPT in universities, some have offered advice that professors may discourage pupils from utilizing ChatGPT when completing dissertations and schoolwork (Baidoo-Anu and Owusu Ansah, 2023). According to NBC news, ChatGPT is currently restricted on educational networked computers by the City of New York's Education Division (NYC), making it unavailable to pupils as well as instructors (Rosenblatt, 2023). Although some of these different approaches might be effective in the short term, they might not hold up in the long run when competing with more advanced forms of generative AI language models. This has been demonstrated by a research preprint in a repository that AI-generated text sensors are ineffective when used with modern, complex natural language processing models (Roller et al., 2020). Learners can modify the content produced by these sensors to make it undetected, so we shouldn't forget that they've got permission to use them as well. Instead of such, it was past time that we started to acknowledge the quickly changing nature of educational practices and incorporated these adjustments into our existing pedagogical practices. Furthermore, ChatGPT will soon become standard due to Microsoft's efforts to compete it by launching bard.ai application with several new features of ChatGPT (Rudolph et al., 2023). By the time institutions of learning reconsider their rules and procedures in order to assist and direct the pupils they teach to utilize ChatGPT effectively, it might have been too late.According to the article published in the Central European Journal of Education Research, the evaluation of students has garnered increased interest and has emerged as a significant topic of debate in academia (Cs epes, 2021). Although it is too early to draw any conclusions, instructors will likely want to reconsider how pupils are evaluated. The way evaluations are carried out right now might have to adapt to accommodate more creative tests. Existing research has shown that instructors' abilities and knowledge to participate in excellent evaluation practices that improve education are constrained. Instead, teachers continue to encourage instructors to develop the skills required to engage students in excellent evaluation practices.Educators can acquire the necessary expertise to utilize ChatGPT and other generative AI applications, enabling them to participate in excellent evaluation practices that enhance students' achievement through professional capacity enhancement. Integrating generative AI technologies into educational environments and instructing students to effectively and responsibly utilize them would also contribute to their success in an AI-dominated world.Consequently, instructors migh...

show abstract

Open-Domain Conversational Agents: Current Progress, Open Problems, and Future Directions

Cited by 13 publications

References 95 publications

Bilateral Personalized Dialogue Generation with Contrastive Learning

Bilateral Personalized Dialogue Generation with Contrastive Learning

Survey of Hallucination in Natural Language Generation

Commentary: Future thoughts on ChatGPT and its applications in education

Contact Info

Product

Resources

About