Syntax Controlled Knowledge Graph-to-Text Generation with Order and Semantic Consistency

Abstract

The knowledge graph (KG) stores a large amount of structural knowledge, while it is not easy for direct human understanding. Knowledge graph-to-text (KG-to-text) generation aims to generate easy-to-understand sentences from the KG, and at the same time, maintains semantic consistency between generated sentences and the KG. Existing KG-to-text generation methods phrase this task as a sequence-tosequence generation task with linearized KG as input and consider the consistency issue of the generated texts and KG through a simple selection between decoded sentence word and KG node word at each time step. However, the linearized KG order is commonly obtained through a heuristic search without data-driven optimization. In this paper, we optimize the knowledge description order prediction under the order supervision extracted from the caption and further enhance the consistency of the generated sentences and KG through syntactic and semantic regularization. We incorporate the Part-of-Speech (POS) syntactic tags to constrain the positions to copy words from the KG and employ a semantic context scoring function to evaluate the semantic fitness for each word in its local context when decoding each word in the generated sentence. Extensive experiments are conducted on two datasets, WebNLG and DART, and achieve state-of-the-art performances. Our code is now public available 1 .

Structure-aware Knowledge Graph-to-text Generation with Planning Selection and Similarity Distinction

“…Graformer (Schmitt et al, 2020) introduces a model that combines relative position information to compute self-attention. Other approaches (Wang et al, 2021b;Liu et al, 2022;Guo et al, 2020;Ribeiro et al, 2020a) first linearize KG into sequences and then feed them into the sequence-to-sequence (Seq2Seq) model for generating desired texts. In this paper, we employ the Seq2Seq model with a planning selector to control linearized sequence orders.…”

“…Existing works (Zhao et al, 2020) have shown that the linearized order of the given triples has an effect on the generated text's quality. Previous works mainly use graph traversal or multistep prediction (Su et al, 2021;Liu et al, 2022;Zhao et al, 2020) methods for triple order generation. uses the relation-biased BFS (RBFS) strategy to traverse and linearize KGs into sequences.…”

“…Seven other models are compared with our model. ‡ indicates that the results are reprinted from(Liu et al, 2022).…”