Bitte verwenden Sie diesen Link, um diese Publikation zu zitieren, oder auf sie als Internetquelle zu verweisen: https://hdl.handle.net/10419/324733 
Erscheinungsjahr: 
2023
Quellenangabe: 
[Journal:] Research Journal for Applied Management (RJAM) [ISSN:] 2701-6633 [Volume:] 4 [Issue:] 1 [Year:] 2023 [Pages:] 133-158
Verlag: 
BoD – Books on Demand GmbH, Norderstedt
Zusammenfassung: 
This research is a feasibility study to design a Natural Language Processing (NLP) system within a Ques-tion-and-Answering (Q&A) environment for internal IT help desk ticketing operations. The proceedings will be used to develop a conversational agent for an IT consultancy company. Tests with few-shot learning algorithms were performed by calibrating two GPT-2 language models. Another benchmark model was tested on the most recent GPT-3 standard. The input data stems from software license release requests of an internal IT help desk. The final model will be a hybrid approach of first guidance by an automated agent and a human expert intervention for more complicated IT problem. The agent will improve itself by constantly evaluating user feedback.
Schlagwörter: 
Natural Language Processing
Question Answering
GPT-2
GPT-3
Few-shot learning
Natural Language Generation
Conversational Agent Implementation
JEL: 
C67
C88
M15
O31
O32
Dokumentart: 
Article

Datei(en):
Datei
Größe





Publikationen in EconStor sind urheberrechtlich geschützt.