Natural Language Processing (NLP) is a very essential component in the data science domain, it’s vastly applicable across various industries in the world. Most times humans get an easy experience to understand different languages but for machines, it’s otherwise.
Therefore, Natural Language Processing is the technique that enables machines to interpret and understand the ways humans communicate. In the latest episode of the African Experts Podcast, “Natural Language Processing” host Vickie Remoe is in conversation with Language Processing expert Paul Azunre.
Remoe & Azunre explores the cost of big data, the dos and don’ts of AI, and how he is using
transfer learning to process Ghanaian languages so that other Africans can communicate with computers in Zulu, Lingala, or Madingo with the same ease and level of access that will allow speakers of resource-rich languages like French, English & Spanish to do today.
Azunre was first asked if is it possible to have machine learning language processing models that are free of bias?
He responded by saying that yes, it is possible to have a machine learning language process that is bias-free, it’s not something that will happen on its own, it has to be an active process where people from different backgrounds have to participate in the process making sure that these systems don’t have a bias. But if there is a bias, the one way to make people aware and use it at their own risk is to put warning labels on everything.
He was further asked about how AI ethics leaders like Timnit Gebru can be so easily pushed out where they could potentially do the most good? Where will the checks and balances come from as big data that goes beyond the resources available to check for these kinds of biases and inaccuracies?
According to Azunre, there are a lot of opinions on that, but he thinks that there are checks and balances. He later states that within the context of Western society there is a power struggle between community leaders, like Timnit Gebru and people believe that the way things are done right now can continue for years to come. So one way of doing things the right way is getting people involved in the political & development processes, also people need to build enterprises that are going to look out for ways that could help show how things can be improved in society than how it is presently.
The conversation was then moved on to what inequities did he notice that made him launch Ghanaian language processing?
Azunre said that he was inspired by the lack of local Ghanaian languages on Google Translate and that Africa’s presence is indispensable in the digital space.
“If you look at Google Translate, you will find that it doesn’t have any Ghanaian languages and our presence on the digital space today is indispensable,” he stated.
So in short people from Africa have to participate in the process, their culture and who they are has to be preserved, it has to be digitised and it has to be secured for the future. And that is what the Ghana NLP is all about.
Get more insight about machine learning, Ai and Natural Language Processing by listening to the podcast now: