Slot extraction method in multiple rounds of conversations, model training method and storage medium

The invention provides a slot extraction method in multi-round dialogues, a model training method and a storage medium, and the training method comprises the steps: obtaining first training data of a plurality of non-preset scene dialogues, obtaining a plurality of preset scene dialogues, and obtain...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: HAO ZHULIN, PI HUAIYU, JIANG XIAOLIN, LUO CHAO
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention provides a slot extraction method in multi-round dialogues, a model training method and a storage medium, and the training method comprises the steps: obtaining first training data of a plurality of non-preset scene dialogues, obtaining a plurality of preset scene dialogues, and obtaining second training data of slot information corresponding to the preset scene dialogues, the preset large model is trained based on the first training data to obtain a pre-training model, the pre-training model is finely adjusted based on the second training data to obtain a slot position extraction model corresponding to the preset scene, and the slot position extraction model is used for slot position extraction of multiple rounds of conversations. According to the method, the large language model is trained through the non-preset scene dialogue and the preset scene dialogue, the robustness of the model is enhanced, accurate recognition of slot positions in multi-round dialogue scenes is achieved through fine ad