In the text analytics industry, sensitivity is a measure of how well a text analytics system can identify relevant information in a given text. The higher the sensitivity, the more likely it is that the system will correctly identify relevant information. Sensitivity is often measured as a percentage, with 100% indicating that the system always correctly identifies relevant information.
Outside of this industry, sensitivity may be used to refer to several different things, including the ability to detect small changes or the ability to respond quickly to stimuli. However, in the context of text analytics, sensitivity refers specifically to the ability of a system to identify relevant information.
Sensitivity is often measured as a percentage, with 100% indicating that the system always correctly identifies relevant information.
To calculate sensitivity, analysts will take the number of true positives – which is to say, the number of relevant items that are correctly identified by the system – and divide it by the total number of relevant items. This will give them a percentage that indicates how sensitive the system is.
Sensitivity vs. Specificity
Sensitivity and specificity are two measures that are often used together to evaluate the performance of a text analytics system. Sensitivity measures the ability of the system to identify relevant information, while specificity measures the ability of the system to identify irrelevant information.
Both sensitivity and specificity are important when evaluating a text analytics system. A system with high sensitivity but low specificity may correctly identify a lot of relevant information, but it will also incorrectly identify a lot of irrelevant information. This can lead to problems such as false positives, where the system identifies something as relevant when it is actually not. On the other hand, a system with high specificity but low sensitivity may correctly identify a lot of irrelevant information, but it will also miss a lot of relevant information. This can lead to problems such as false negatives, where the system fails to identify something as relevant when it actually is.