Google has open sourced LaserTagger, an AI model to tackle text generation in a less error-prone manner.
Instead of generating the output text from scratch, LaserTagger produces output by tagging words with predicted edit operations that are then applied to the input words in a separate realization step. According to research team at Google, “This is a less error-prone way of tackling text generation, which can be handled by an easier to train and faster to execute model architecture.”
The team said it evaluated LaserTagger on four tasks: sentence fusion, split and rephrase, abstractive summarization, and grammar correction. LaserTagger, across these 4 tasks, performed comparably to a strong BERT-based seq2seq baseline that used a large number of training examples. Moreover, it outperformed this baseline when the number of training examples was limited.
“The advantages of LaserTagger become even more pronounced when applied at large scale, such as improving the formulation of voice answers in some services by reducing the length of the responses and making them less repetitive,” the team added.
Google has open sourced the code for LaserTagger through its GitHub repo.