DiaBiz
DiaBiz corpus is a dialog corpus of Polish comprising recordings and annotated transcriptions of phone-based customer-agent interactions in several key business domains.
TryHow to use this tool?
1
4036 call-center dialogues with a total length of nearly 410 hours (over 3.2 million words) based on 251 real-world interaction scenarios.
2
DiaBiz includes data from 9 business domains, i.e. banking, energy services, telecommunications, insurance, medical care, debt collection, tourism, car rental and retail, with high commercial demand for conversation analysis and automation solutions.
3
All transcripts are enhanced by punctuation and truecasing. A subset of DiaBiz transcriptions were annotated with intentions of customers and agents that are appropriate to the business domain.
Features of the tool
Product feature
The corpus data were collected between May 2021 and April 2022. The phone-call interactions between 5 agents with professional experience in call center settings and 191 participants assuming the role of customers were recorded via the Genesys PureCloud platform.
Product feature
DiaBiz can serve as a source of training and evaluation data for a wide range of tasks, such as speech recognition and transcript formatting, speaker diarization, conversational analytics, modelling of dialog systems, and many more.
Product feature
The transcripts are also punctuated in order to enhance their clarity. They were first punctuated automatically and subsequently manually corrected by, for which purpose punctuation guidelines that reflect the characteristics of spoken language were developed.
Product feature
The data was automatically automatically transcribed and time-aligned and subsequently manually corrected and annotated.
Product feature
The corpus was developed to boost the development of third-party speech recognition engines, dialog systems and conversational intelligence tools for Polish.
Lorem ipsum dolor sit amet consectetur.
Lorem ipsum dolor sit amet consectetur. Lorem ipsum dolor sit amet consectetur. Lorem ipsum dolor sit amet consectetur. Lorem ipsum dolor sit amet consectetur.
Try