Personalizing Dialogue Agents for Russian: Retrieve and Refine

Posokhov, Pavel; Apanasovich, Kirill; Matveeva, Anastasia; Makhnytkina, Olesia; Matveev, Anton

doi:10.23919/fruct54823.2022.9770895

Cited by 6 publications

(1 citation statement)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Refining methods have been studied to improve the generation results when they are not satisfactory or not in the desired intention. (He, 2021;Geng et al, 2021;Sun et al, 2022;Bao and Zhang, 2023) Especially in dialogue systems, the researches aim to improve the quality of response by applying the refining method have been proposed (Tran and Nguyen, 2018;Posokhov et al, 2022;. To tackle the limitation of generating uninformative or not engaging utterances (Shen et al, 2018), retrieves the utterance and refines it by regenerating the retrieved utterance.…”

Section: Utterance Refining Methodsmentioning

confidence: 99%

Post-hoc Utterance Refining Method by Entity Mining for Faithful Knowledge Grounded Conversations

Jang,

Son,

Lee

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Despite the striking advances in recent language generation performance, modelgenerated responses have suffered from the chronic problem of hallucinations that are either untrue or unfaithful to a given source. Especially in the task of knowledge grounded conversation, the models are required to generate informative responses, but hallucinated utterances lead to miscommunication. In particular, entity-level hallucination that causes critical misinformation and undesirable conversation is one of the major concerns. To address this issue, we propose a post-hoc refinement method called REM. It aims to enhance the quality and faithfulness of hallucinated utterances by refining them based on the source knowledge. If the generated utterance has a low source-faithfulness score with the given knowledge, REM mines the key entities in the knowledge and implicitly uses them for refining the utterances. We verify that our method reduces entity hallucination in the utterance. Also, we show the adaptability and efficacy of REM with extensive experiments and generative results. Our code is available at https://github.com/YOONNAJANG/REM.

show abstract