Hierarchical Encoder with Auxiliary Supervision for Neural Table-to-Text Generation: Learning Better Representation for Tables

Abstract

Generating natural language descriptions for the structured tables which consist of multiple attribute-value tuples is a convenient way to help people to understand the tables. Most neural table-to-text models are based on the encoder-decoder framework. However, it is hard for a vanilla encoder to learn the accurate semantic representation of a complex table. The challenges are two-fold: firstly, the table-to-text datasets often contain large number of attributes across different domains, thus it is hard for the encoder to incorporate these heterogeneous resources. Secondly, the single encoder also has difficulties in modeling the complex attribute-value structure of the tables. To this end, we first propose a two-level hierarchical encoder with coarse-to-fine attention to handle the attribute-value structure of the tables. Furthermore, to capture the accurate semantic representations of the tables, we propose 3 joint tasks apart from the prime encoder-decoder learning, namely auxiliary sequence labeling task, text autoencoder and multi-labeling classification, as the auxiliary supervisions for the table encoder. We test our models on the widely used dataset WIKIBIO which contains Wikipedia infoboxes and related descriptions. The dataset contains complex tables as well as large number of attributes across different domains. We achieve the state-of-the-art performance on both automatic and human evaluation metrics.

A Hierarchical Model for Data-to-Text Generation

“…It consists in conditioning the decoder on entity representations that are updated during inference at each decoding step. On the other hand, Liu et al [18,17] rather focus on introducing structure into the encoder. For instance, they propose a dual encoder [17] which encodes separately the sequence of element names and the sequence of element values.…”

“…On the other hand, Liu et al [18,17] rather focus on introducing structure into the encoder. For instance, they propose a dual encoder [17] which encodes separately the sequence of element names and the sequence of element values. These approaches are however designed for single-entity data structures and do not account for delimitation between entities.…”

Natural Language Generation for Non-Expert Users

“…In GF, NP, VV, V2, VP, and Cl stand for noun phrase, verb-phrase-complement verb, two-place verb, verb phrase and clause, respectively. Note that although the set of GF grammatical rules can be used to construct a constituency-based parse tree 12 , the reverse direction is not always true. To the best of our knowledge, there exists no algorithm for converting a constituency-based parse tree to a set GF grammar rules.…”

“…A more advanced NLG system in this direction is described in [16], which works with ontologies annotated using the Attempto language and can generate a natural language description for workflows created by the systems built in the Phylotastic project 3 . The applications targeted by these systems are significantly different from NLG systems, whose main purpose is to generate high-quality natural language description of objects or reports, such as those reported in the recent AAAI conference [12,6,15].…”

The Rare Word Issue in Natural Language Generation: A Character-Based Solution

“…Neural methods for DTT have shown the ability to produce fluent texts conditioned on input data in several domains [7,8], without relying on heavy manual work from field experts. They successfully combined ideas from machine translation, such as encoderdecoder architectures [9] and attention mechanisms [10,11], and from extractive summarization, such as the copy mechanism [12,13], often explicitly modeling the key-value structure of the input table [14][15][16][17][18][19]. However a common issue to deal with is that most of the commonly used models share the word-by-word representation of data in both input sequences and generated utterances; such schemes cannot be effective without a special, non-neural delexicalization phase that handles rare or unknown words, such as proper names, telephone numbers, or foreign words [20].…”