Collection of interviews with experts from end-user groups and key value chains of agricultural waste in the food industry. The interviews were conducted as part of the Agriloop project, which aims to transform unused agricultural residues into valuable products, making the process both economically profitable and environmentally sustainable. The primary purpose of the interviews is to investigate the market readiness of Agriloop innovations for the B2B (business-to-business) sector. Specifically, this includes identifying practical opportunities (promising residue streams, regional focus areas, relevant applications, and B2B market assessments) and understanding the current state (existing valorization, stakeholders' expectations for PHA, PHA applications and microbial proteins, and potential challenges and limitations).

The interviews will be further analyzed using the Delphi method (Linstone et al., 1975). This collection includes interviews from the first round (Delphi I) and the second round (Delphi II) of a series of planned rounds. The data from this round comes from 17 participants in the Agriloop project and is organized into four questionnaires adapted to different target groups: suppliers of residual streams, users of end-product PHA applications, converters of the end-product PHA, feed traders and processors of end-product extracted microbial proteins. Each questionnaire is divided into blocks: "Introduction and Icebreaker", "Strengths and Challenges", "Current Structures", "Expectations", "Outlook and Perspectives". The exact blocks vary depending on the questionnaire.

Following the Delphi method practices, the interview responses were annotated according to a specially developed code system. The code system consists of 145 codes (Delphi I) and 296 codes (Delphi II) organized hierarchically, with up to five levels. The files included in this dataset are:

Transcripts_Delphi(n)_Agriloop.zip: An archive containing transcripts of interviews from Delphi(n) in CSV (Comma-separated values) format. Coded_segments_Delphi(n)_AgriLoop.zip: An archive containing Delphi(n) interview fragments annotated with the complete code system in CSV format. Processed_data_Delphi1_Delphi2_AgriLoop.zip: An archive with processed data, containing paragraphs and sentences annotated using both 7 and 15 classes (corresponding to the first level and first with the second level of the code system hierarchy, with the classes "Company" and "Experts" merged together). The processed data is particularly useful for developing an automatic annotation system. These data is available in both CSV and JSON formats. Synthetic_data.zip: An archive with synthetic data, containing annotated paragraphs and sentences generated using the Retrieval Augmented Generation (RAG) method. This data is particularly useful in combination of processed data for developing an automatic annotation system. These data is available in both CSV and JSON formats. Agriloop_Delphi_Codesystem.xlsx: The complete code system used for interview annotation, provided in MS Excel format.

The original interview responses contained brand names and personal information of the respondents. All the data were pseudonymized with the respondents' permission. Any resemblance to real brands or individuals, if present, is purely coincidental.