| dc.description.abstract | Dictation is the process of converting language from spoken format to
written format. This process is usually carried out manually, that is, a
human being listens to speech and then either handwrites or types the
equivalent written form. Dictation systems are computer based
automatic speech recognition systems that are meant to automate this
process. Ideally a dictation system should improve significantly on the
speed of the transcription and also reduce the number of errors that
occur.
Dictation systems are language specific. This is because each language
has different words, which are used to form the sentences in that
language. In spite of this, the techniques used to implement dictation
systems for different languages are largelythe same.
The purpose of this report is to describe the work done in developing a
prototype dictation system for the Kiswahili language. The
implementations of the speech recognition front-end, the acoustic model,
the Viterbi algorithm for recognition, and an evaluation component are
described. The report also highlights the relevant literature on the
techniques used to build dictation systems, the design of the dictation
system, the results obtained from the prototype system that is built, and
finally gives suggestions on how this system can be improved. | en |
