Controlling contents in data-to-document generation with human-designed topic labels

Abstract

We propose a data-to-document generator that can easily control the contents of output texts based on a neural language model. Conventional data-to-text model is useful when a reader seeks a global summary of data because it has only to describe an important part that has been extracted beforehand. However, since it differs from users to users what they are interested in, it is necessary to develop a method to generate various summaries according to users’ requests. We develop a model to generate various summaries and to control their contents by providing the explicit targets for a reference to the model as controllable factors. In the experiments, we used five-minute or one-hour charts of 9 indicators (e.g., Nikkei 225), as time-series data, and daily summaries of Nikkei Quick News as textual data. We conducted comparative experiments using two pieces of information: human-designed topic labels indicating the contents of a sentence and automatically extracted keywords as the referential information for generation. Experiments show both models using additional information of target document achieved higher performance in terms of BLEU and human evaluation. We found that human-designed topic labels are superior to extracted keywords in terms of controllability.