Google

The T5 text-to-text framework can be applied to question-answering tasks by formatting both the input and the target output as text strings. For instance, a question is presented with a prefix like 'Answer:', and the model is trained to generate the corresponding answer. A complete example of this format is: `[CLS] Answer: when was Albert Einstein born? → ⟨s⟩ He was born on March 14, 1879.`

Example of a T5 Question-Answering Sample

A data scientist is preparing a dataset to train a text-to-text model for a question-answering task. One of the data points is the question 'What is the chemical symbol for gold?' and its corresponding answer 'Au'. Construct the complete, single-line training sample for this data point, following the specific format required by the model. Your answer should include the standard prefix for this task, the question, the separator, the start-of-sequence token, and the answer.

Formatting a Training Sample for a Text-to-Text Model

A machine learning engineer is preparing a dataset for a text-to-text model to perform a question-answering task. They have created the following training sample: `[CLS] what is the capital of France? → ⟨s⟩ The capital is Paris.` Identify the primary error in this sample's format and explain why it would likely cause the model to perform poorly.

Analyzing an Incorrect T5 Question-Answering Sample

A researcher is preparing data for a unified text-to-text model. To have the model perform a question-answering task, the input data must be structured as a single text string that begins with a specific instruction, followed by the question itself. Based on this requirement, which of the following examples is formatted correctly to serve as the model's input?

Learn Before

Related