ALOHA: Artificial Learning of Human Attributes for Dialogue Agents

Abstract

For conversational AI and virtual assistants to communicate with humans in a realistic way, they must exhibit human characteristics such as expression of emotion and personality. Current attempts toward constructing human-like dialogue agents have presented significant difficulties. We propose Human Level Attributes (HLAs) based on tropes as the basis of a method for learning dialogue agents that can imitate the personalities of fictional characters. Tropes are characteristics of fictional personalities that are observed recurrently and determined by viewers' impressions. By combining detailed HLA data with dialogue data for specific characters, we present a dataset, HLA-Chat, that models character profiles and gives dialogue agents the ability to learn characters' language styles through their HLAs. We then introduce a three-component system, ALOHA (which stands for Artificial Learning of Human Attributes), that combines character space mapping, character community detection, and language style retrieval to build a character (or personality) specific language model. Our preliminary experiments demonstrate that two variations of ALOHA, combined with our proposed dataset, can outperform baseline models at identifying the correct dialogue responses of chosen target characters, and are stable regardless of the character's identity, the genre of the show, and the context of the dialogue.

Coffee With a Hint of Data: Towards Using Data-Driven Approaches in Personalised Long-Term Interactions

“…Another solution could be to combine a data-driven model with another approach to compensate for the deficiencies in the models, such as combining a generative model (e.g., Sequence-to-Sequence) with a Memory Network ( Madotto et al, 2018 ; Zhang B. et al, 2020 ) or with transformers ( Vaswani et al, 2017 ), such as in the work of Roller et al (2020) , Generative Pre-trained Transformer (GPT) ( Radford et al, 2018 , 2019 ; Brown et al, 2020 ; Zhang Y. et al, 2020 ), Bidirectional Encoder Representations from Transformers (BERT) ( Devlin et al, 2019 ; Song et al, 2021 ), and Poly-encoders ( Humeau et al, 2020 ; Li et al, 2020 ). Data-driven models can also be combined with graphical models ( Zhou et al, 2020 ; Song et al, 2019 ; Moon et al, 2019 ; Shi et al, 2020 ; Wu B. et al, 2020 ; Xu et al, 2020 ), rule-based or slot-filling systems ( Tammewar et al, 2018 ; Zhang Z. et al, 2019 ), a knowledge-base ( Ganhotra and Polymenakos, 2018 ; Ghazvininejad et al, 2018 ; Luo et al, 2019 ; Yavuz et al, 2019 ; Moon et al, 2019 ; Wu et al, 2019 ; Lian et al, 2019 ; Zhang B. et al, 2020 ; Majumder et al, 2020 ; Tuan et al, 2021 ) or with automatic extraction of attributes from dialogue ( Tigunova et al, 2019 , 2020 ; Wu C.-S. et al, 2020 , 2021 ; Ma et al, 2021 ) to improve the personalised entity selection in responses.…”

Stylistically User-specific Response Generation

“…In the second category, Li et al (2020) introduced an approach to construct human-level attributes from movie character tropes and used them in a response selection task. They learned the language styles of movie characters associated with several traits, and then retrieved the suitable response associated with the same traits as the target character.…”

“…As our intention is to capture the characteristics of the users, that is, the persons who respond to the given utterance, we argue that using user-related information would be useful to achieve that objective. In contrast to Li et al (2020), which classified movie characters into several associated traits, we wanted the model to learn the style at the user level. Additionally, because we adopt LSTM in our model, we argue that using a simpler mechanism would be more effective.…”

“…To address the interestingness and response diversity issues, some studies have focused on improving the style or response characteristics to resemble human-produced ones. One approach is to integrate user-specific information/features (Li et al 2016b;Bak and Oh 2019;Wu et al 2020), or some specific persona characteristics (Chu et al 2018;Li et al 2020) to establish the association between the responses and the corresponding persona. Another approach is to transfer the specific style or latent information from additional texts to the responses (Herzig et al 2017;Zhang et al 2018;Gao et al 2019).…”

Extracting and Inferring Personal Attributes from Dialogue

“…One such task is to ground open-domain chit-chat dialogue agents to minimize inconsistencies in their language use (e.g., I like cabbage →(next turn) →Cabbage is disgusting) and make them engaging to talk with (Li et al 2016;Zhang et al 2018;Mazaré et al 2018;Qian et al 2018;Zheng et al 2020a,b;Li et al 2020;Majumder et al 2020). Thus far, personalization in chit-chat has made use of dense embeddings and natural language sentences.…”