Integrating LLMs into the DIARC Cognitive Architecture to Resolve Past Temporal References

Valerio Farriciello; Frank  Foerster

Integrating LLMs into the DIARC Cognitive Architecture to Resolve Past Temporal References

Valerio Farriciello
, Frank Foerster

Research output: Contribution to conference › Paper › peer-review

Abstract

We report on our efforts to resolve past temporal references – past temporal deictic and past discourse deictic expressions – as well as pronominal anaphora, present in instructions given to robots controlled by the DIARC cognitive architecture. Instructions given to a robot, are sent to a large language model (LLM) to firstly determine whether these contain such references, and, in the affirmative case, to rewrite the instructions such that they comply with the format accepted by the parser of DIARC’s natural language understanding component (NLU). The preliminary results are promising.

Original language	English
Number of pages	3
Publication status	Published - Sept 2025
Event	Semdial 2025 - Bialogue, the 29th Workshop on the Semantics and Pragmatics of Dialogue: Bialogue - Germany, Bielefeld, Germany Duration: 3 Sept 2025 → 5 Sept 2025 Conference number: 2025 https://semdial2025.github.io/

Workshop

Workshop	Semdial 2025 - Bialogue, the 29th Workshop on the Semantics and Pragmatics of Dialogue
Abbreviated title	SemDial
Country/Territory	Germany
City	Bielefeld
Period	3/09/25 → 5/09/25
Internet address	https://semdial2025.github.io/

Keywords

dialogue
dialogue systems

Access to Document

http://semdial.org/anthology/Z25-Farriciello_semdial_4430.pdfLicence: CC BY

Cite this

@conference{c4dd70205caa47a29b025e76540344dd,

title = "Integrating LLMs into the DIARC Cognitive Architecture to Resolve Past Temporal References",

abstract = "We report on our efforts to resolve past temporal references – past temporal deictic and past discourse deictic expressions – as well as pronominal anaphora, present in instructions given to robots controlled by the DIARC cognitive architecture. Instructions given to a robot, are sent to a large language model (LLM) to firstly determine whether these contain such references, and, in the affirmative case, to rewrite the instructions such that they comply with the format accepted by the parser of DIARC{\textquoteright}s natural language understanding component (NLU). The preliminary results are promising.",

keywords = "dialogue, dialogue systems",

author = "Valerio Farriciello and Frank Foerster",

year = "2025",

month = sep,

language = "English",

note = "Semdial 2025 - Bialogue, the 29th Workshop on the Semantics and Pragmatics of Dialogue : Bialogue, SemDial ; Conference date: 03-09-2025 Through 05-09-2025",

url = "https://semdial2025.github.io/",

}

TY - CONF

T1 - Integrating LLMs into the DIARC Cognitive Architecture to Resolve Past Temporal References

AU - Farriciello, Valerio

AU - Foerster, Frank

N1 - Conference code: 2025

PY - 2025/9

Y1 - 2025/9

N2 - We report on our efforts to resolve past temporal references – past temporal deictic and past discourse deictic expressions – as well as pronominal anaphora, present in instructions given to robots controlled by the DIARC cognitive architecture. Instructions given to a robot, are sent to a large language model (LLM) to firstly determine whether these contain such references, and, in the affirmative case, to rewrite the instructions such that they comply with the format accepted by the parser of DIARC’s natural language understanding component (NLU). The preliminary results are promising.

AB - We report on our efforts to resolve past temporal references – past temporal deictic and past discourse deictic expressions – as well as pronominal anaphora, present in instructions given to robots controlled by the DIARC cognitive architecture. Instructions given to a robot, are sent to a large language model (LLM) to firstly determine whether these contain such references, and, in the affirmative case, to rewrite the instructions such that they comply with the format accepted by the parser of DIARC’s natural language understanding component (NLU). The preliminary results are promising.

KW - dialogue

KW - dialogue systems

M3 - Paper

T2 - Semdial 2025 - Bialogue, the 29th Workshop on the Semantics and Pragmatics of Dialogue

Y2 - 3 September 2025 through 5 September 2025

ER -

Integrating LLMs into the DIARC Cognitive Architecture to Resolve Past Temporal References

Abstract

Workshop

Keywords

Access to Document

Fingerprint

Cite this