A free fortnightly newsletter on Taxonomy, Thesauri & Ontology and Semantic Publishing
Enabling machine learning to handle ambiguity
It is tough to preempt simple factors such as uncertainty when using machine learning (ML) and artificial intelligence (AI). For example, in image classification, if the image features in the data are not captured accurately, the output will be vague. While writing in the Analytics India Magazine, Abhishek Sharma states that this ambiguity is common to ML and gives a few instances where it was handled convincingly.
Two factors cause ambiguity. Ambiguity occurs because of the goal or the data assigned to the ML system. For instance, if the goal is vague or data is of poor quality, the ML system will process it ambiguously. As a result, it will lead to open interpretations. To overcome the resulting uncertainty, researchers have used innovative methods and reduced ambiguity in natural language, improved DNA sequencing, and enhanced image-word matching in image classification.
The instances mentioned above cover only the text aspect of ML. However, ML works on many other types of data such as images, videos, codes, among other examples. Therefore, to cut ambiguity, care should be taken that only high-quality data is used. In addition, the goal of the project should be precise and should follow the requirements of the ML project.
Click here to read the original article.