| |
|
|
| |
 |
|
| |
| |
|
|
| |
Considering the sheer multiplicity of African languages, as a first step in the process, we have chosen one language; Yoruba, with which we are now running a pilot programme. Within this pilot programme we have been involved in the following projects:
Corpus Development and Language Modelling
We have developed a modest corpus of computer readable texts in standard Yoruba orthography and as a result we have started to build a statistical model of the language. Recognising the centrality of a Statistical Language Model (SLM) as a basic prerequisite in many human language technologies, we continue to update this model by continually enhancing the Alt-I corpus of computer readable Yoruba texts.
Yoruba Keyboard and Word processor
We have designed an efficient and ergonomic layout for the Yoruba keyboard which has been acclaimed to be quite efficient. It is gaining popularity among Yoruba typesetters as well as undergraduates and post-graduate students of Yoruba studies in Nigerian universities. We have also developed a 20,000-word Yoruba lexicon, which can be used for spelling correction in various stardard word processors. Work is also on-gong on Yoruba grammar correction.
|
|
| |
|
|
| |
 |
|
| |
Alt-i Yoruba Keyboard Layout 1, based on
minimal rearrangement of the QWERTY layout |
|
| |
 |
|
| |
Alt-i Yoruba Keyboard Layout 2, based on
statistical distribution
of characters in Yoruba text for efficiency and ergonomics |
|
| |
|
|
| |
Seminars and Workshops
As part of our advocacy efforts, we run seminars and workshops aimed at developing interests in HLT research in universities and research centres. In appreciation of the high level of interdisciplinary scholarship required for these projects, we employ these seminars and workshops to encourage 'bridge building' between relevant departments and faculties in universities and to facilitate curricular review to accommodate developments in HLT.
Speech recognition
In accordance to our development plan, we are working actively on a Yoruba Speech Recogniser. Starting life as merely a laboratory novelty a few years ago, the recogniser is now gradually becoming a usable product, capable of transforming a sound file of a Yoruba utterance in the .wav format into text in standard Yoruba orthography. We continue to work on improving both the accuracy and speed of the recogniser.
Speech Synthesis
We have also developed a modest Yoruba text-to-speech system, which is capable of reading any electronic version of Yoruba texts in standard orthography. We continue to colaborate with Dr. Odutunji Odejobi of the Department of Electronics and Computer Engineering of the Obafemi Awolowo University in his work on Yoruba Speech Synthesis.
Addressing other Nigerian Languages
We have an active working relationship with the National Institute for Nigerian Languages (NINLAN), in Aba, Nigeria. We employ this relationship to keep in close contact with top scholars of Nigerian languages. |
|
| |
|
|
|
|
| |
|
|
| |
 |
|
| |
| |
|
|
| |
Projects in the pipeline
Apart from the above projects that we are actively pursuing at present, we are also examining other areas of HLT needs and are developing relevant projects and resources to meet these needs. Despite funding constraints, we continue to keep our interests in machine translation and natural language understanding active.
|
|
| |
|
|
|
|
| |
|
|
