Probabilistic reasoning in intelligent systems: networks of plausible inference
Probabilistic reasoning in intelligent systems: networks of plausible inference
Machine Learning - Special issue on learning with probabilistic representations
Studying the language and structure in non-programmers' solutions to programming problems
International Journal of Human-Computer Studies
Machine Learning
Programming in natural language: “NLC” as a prototype
ACM '79 Proceedings of the 1979 annual conference
Metafor: visualizing stories as code
Proceedings of the 10th international conference on Intelligent user interfaces
Learning to transform natural to formal languages
AAAI'05 Proceedings of the 20th national conference on Artificial intelligence - Volume 3
A global view of automatic programming
IJCAI'73 Proceedings of the 3rd international joint conference on Artificial intelligence
On natural language based computer systems
IBM Journal of Research and Development
| Hi-index | 0.00 |
Learning a programming language is a painstaking process, as it requires knowledge of its syntax, apart from knowing the basic process of representing logical sequences to programming stages. This fact deteriorates the coding process and expels most users from programming. Particularly for novice users or persons with vision problems, learning of how to program and tracing the syntax errors could be improved dramatically by using the most natural of all interfaces, i.e. natural language. Towards this orientation, we suggest a wider framework for allowing programming using natural language. The framework can be easily extended to support different object-oriented programming languages such as C, C++, Visual Basic or Java. Our suggested model is named "Language Oriented Basic" and it concerns an intelligent interface that supports code creation, modification and control in Visual Basic. Users can use simple-structured Greek sentences in natural language and the system can output the corresponding syntactic tree. When users declare end of input, the system transforms the syntactic trees to source code. Throughout the whole interaction process, users can check the under-development code in order to verify its correspondence to their expectations. Due to the fact that using natural language can cause a great degree of ambiguity, Bayesian networks and learning from examples have been utilized as an attempt to reason on the most probable programming representation, given a natural language input sentence. In order to enhance the classifier, we propose a novel variation of Bayesian networks that favor the classification process. Experimental results have depicted precision and recall measures in a range of 73% and 70% respectively.