Google

This example illustrates a masked language modeling task. The model is given an input sentence with missing words, structured as `[C] The kitten [M] playing the [M] .`, where `[C]` denotes the context and `[M]` represents the masked tokens. The objective is for the model to predict the masked words to complete the sentence, for example, 'The kitten is playing with the toy'.

Example of Masked Language Modeling: Kitten Playing

A language model is tasked with filling in the missing words, marked as `[M]`, in the following sentence: "The kitten `[M]` playing with the `[M]`." Based on common linguistic patterns and world knowledge, which pair of words is the most probable prediction for the two `[M]` slots, respectively?

Consider the sentence: `The kitten [M] playing with the [M] .` where `[M]` represents a missing word. Provide a grammatically correct and contextually plausible word for each `[M]` and briefly explain why your choices are suitable based on the surrounding words.

Analyzing Context for Word Prediction

A language model is presented with the sentence `The kitten [M] playing with the [M] .` where `[M]` represents a missing word. The model generates three potential completed sentences. Your task is to evaluate these outputs. Identify the best completion and explain why it is superior to the others, considering both grammatical structure and real-world plausibility.

Learn Before

Related