Spanning tree approaches for statistical sentence generation
Empirical methods in natural language generation
| Hi-index | 0.00 |
Text-to-text generation aims to produce a coherent text by extracting, combining and rewriting information given in input texts. Examples of its applications include summarization, answer fusion in question-answering and text simplification. At first glance, text-to-text generation seems a much easier task than the traditional generation set-up where the input consists of a non-linguistic representation. Research in summarization over the last decade proved that the opposite is true --- texts generated by these methods rarely match the quality of those written by humans. One of the key reasons is the lack of coherence in the generated text.