Dictionary of terms

On this page the definitions of the five (5) met terms in the presentation of the use cases are found.

Please, click on the term you would like to see the definition of:

Quantitative measures used to asses the performance and effectiveness of the GenAI models. They were utilized in order to evaluate the code generated from the two (2) different GenAI models, GenAI model 1 and GenAI model 2 in correspondence with the human-written code (used as reference).

CodeBLEU: An evaluation metric that elevates the BLUE score metric (used for natural language processing), specifically for generated code evaluation. It utilizes the strengths of BLEU in n-gram (token) matching and takes into consideration code syntax and code semantics.
CodeBERTScore: An evaluation metric tailored for generated code evaluation. Instead of only taking into consideration the reference (human-written) code it also encodes the natural language input (prompt) used to generate the code, modeling the consistency between them.

A natural language text describing the task that the Generative Artificial Intelligence (GenAI) system should perform.

The Gen AI models were categorized on 3 different difficulty types of prompts, based on the complexity (how elaborative they are) of their used commands:

1: For low difficulty level

2: For medium difficulty level

3: For high difficulty level

In our questionnaire we differentiate between the prompt levels using the terminology: low – medium – high.

A rudimentary working model of an information system built for demonstration purposes or as part of the development process.

Specific prototyping ideas developed in the digital.auto playground. They utilize APIs provided by Vehicle Signal Specification (VSS), an open standard initiated by the Connected Vehicle Systems Alliance (COVESA) to define a syntax and a catalog for vehicle signals. The examples demonstrated in the questionnaire are:

Smart Wipers
Anti Kinetosis
Passenger Welcome
Driver Distraction Detection
Battery Management