Select and Attend: Towards Controllable Content Selection in Text Generation
This work addresses the need for explicit content control in text generation for applications like automated reporting or summarization, though it is incremental in building on existing encoder-decoder models.
The paper tackles the problem of controlling content selection in text generation by decoupling it from surface realization, enabling human-interpretable manipulation without annotations, and achieves promising results in data-to-text and headline generation tasks.
Many text generation tasks naturally contain two steps: content selection and surface realization. Current neural encoder-decoder models conflate both steps into a black-box architecture. As a result, the content to be described in the text cannot be explicitly controlled. This paper tackles this problem by decoupling content selection from the decoder. The decoupled content selection is human interpretable, whose value can be manually manipulated to control the content of generated text. The model can be trained end-to-end without human annotations by maximizing a lower bound of the marginal likelihood. We further propose an effective way to trade-off between performance and controllability with a single adjustable hyperparameter. In both data-to-text and headline generation tasks, our model achieves promising results, paving the way for controllable content selection in text generation.