This interface has been developed as part of a research initiative aimed at collecting authentic human typing errors in the Maltese language. The primary objective is to support the development of a data-driven spell-checking system tailored to the linguistic and orthographic characteristics of Maltese.
Previous spell checkers were trained on cleaner corpora, which fail to capture the natural variability and error patterns present in real-world typing. To address this limitation, this tool facilitates the collection of unedited user input, enabling the creation of more robust models for error detection and correction in Maltese.
Participants are asked to:
- Listen to an audio recording of a spoken Maltese sentence.
- Transcribe the sentence as they hear it using the provided text field.
A key feature of this interface is that the backspace key is disabled, thereby preventing users from correcting their mistakes during the typing process. This constraint ensures that all initial typing errors are preserved, resulting in a more authentic dataset that reflects actual human typing behavior.
The data collected through this interface can be used to:
- Fine-tune AI models for spelling correction and error prediction in Maltese.
- Study common error patterns in Maltese orthography and keyboard input.
- Support broader efforts in natural language processing (NLP) and digital resource development for under-resourced languages.