A Mixed Hierarchical Attention based Encoder-Decoder Approach for Standard Table Summarization

TL;DR

This paper introduces a mixed hierarchical attention encoder-decoder model for summarizing structured table data, significantly improving BLEU scores on the WEATHERGOV dataset by leveraging table structure and content.

Contribution

The paper presents a novel mixed hierarchical attention model specifically designed for standard table summarization, enhancing the ability to utilize table structure and content.

Findings

Achieved approximately 18 BLEU (~30%) improvement over state-of-the-art.

Demonstrated effectiveness on WEATHERGOV dataset.

Validated the model's ability to leverage table structure for better summaries.

Abstract

Structured data summarization involves generation of natural language summaries from structured input data. In this work, we consider summarizing structured data occurring in the form of tables as they are prevalent across a wide variety of domains. We formulate the standard table summarization problem, which deals with tables conforming to a single predefined schema. To this end, we propose a mixed hierarchical attention based encoder-decoder model which is able to leverage the structure in addition to the content of the tables. Our experiments on the publicly available WEATHERGOV dataset show around 18 BLEU (~ 30%) improvement over the current state-of-the-art.