Dialogue text generation device, dialogue text generation method, and program

摘要

An object of the present invention is to provide a dialogue text generation technology that generates utterances that reflect the context of the utterances. A dialogue system 1 includes a learning section 10 and a generation section 20. The learning section inputs the uttered text with attributes of the speaker, extracts words from the previous uttered text and the next uttered text, and uses multi-view attention to control access to other tokens for each token type. Input the words and attributes of the previous utterance text as conditions and the words of the next utterance text as output results into the Transformer that has introduced the mechanism, and create an objective function representing the prediction accuracy of the attributes, the previous utterance text, and the next utterance. The Transformer is trained to minimize the objective function representing the degree of engagement with the text. The generation unit inputs the immediately preceding utterance text and attributes of the speaker into a trained Transformer, connects words recursively output from the Transformer, and generates a response text that becomes the next utterance of the speaker. [Selection diagram] Figure 2