Generating Descriptions from Structured Data Using a Bifocal Attention Mechanism and Gated Orthogonalization

Proceedings of NAACL-HLT 2018, pages 1539–1550 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Generating Descriptions from Structured Data Using a Bifocal Attention Mechanism and Gated Orthogonalization Preksha Nema†∗ Shreyas Shetty†∗ Parag Jain•∗ Anirban Laha•† Karthik Sankaranarayanan• Mitesh M. Khapra†‡ †IIT Madras, India •IBM Research ‡ Robert Bosch Center for Data Science and Artificial Intelligence, IIT Madras {preksha,shshett,miteshk}@cse.iitm.ac.in {pajain34,anirlaha,kartsank}@in.ibm.com Abstract In this work, we focus on the task of generat- ing natural language descriptions from a struc- tured table of facts containing fields (such as nationality, occupation, etc) and values (such as Indian, {actor, director}, etc). One sim- ple choice is to treat the table as a sequence of fields and values and then use a standard seq2seq model for this task. However, such a model is too generic and does not exploit task- specific characteristics. For example, while generating descriptions from a table, a hu- man would attend to information at two levels: (i) the fields (macro level) and (ii) the values within the field (micro level). Further, a human would continue attending to a field for a few timesteps till all the information from that field has been rendered and then never return back to this field (because there is nothing left to say about it). To capture this behavior we use (i) a fused bifocal attention mechanism which ex- ploits and combines this micro and macro level information and (ii) a gated orthogonalization mechanism which tries to ensure that a field is remembered for a few time steps and then forgotten. We experiment with a recently re- leased dataset which contains fact tables about people and their corresponding one line bi- ographical descriptions in English. In addi- tion, we also introduce two similar datasets for French and German. Our experiments show that the proposed model gives 21% relative im- provement over a recently proposed state of the art method and 10% relative improvement over basic seq2seq models. The code and the datasets developed as a part of this work are publicly available. 1 work.∗* The first three authors have contributed equally to this 1https://github.com/PrekshaNema25/ StructuredData_To_Descriptions 1 Introduction Rendering natural language descriptions from structured data is required in a wide variety of commercial applications such as generating de- scriptions of products, hotels, furniture, etc., from a corresponding table of facts about the entity. Such a table typically contains {field, value} pairs where the field is a property of the entity (e.g., color) and the value is a set of possible assign- ments to this property (e.g., color = red). Another example of this is the recently introduced task of generating one line biography descriptions from a given Wikipedia infobox (Lebret et al., 2016). The Wikipedia infobox serves as a table of facts about a person and the first sentence from the cor- responding article serves as a one line descrip- tion of the person. Figure 1 illustrates an exam- ple input infobox which contains fields such as Born, Residence, Nationality, Fields, Institutions and …

Content truncated to 500 words