Pre-trained Language Models recently gained traction in the Natural Language
Processing (NLP) domain for text summarization, generation and
question-answering tasks. This stems from the innovation introduced in
Transformer models and their overwhelming performance compared with Recurrent
Neural Network Models (Long Short Term Memory (LSTM)). In this paper, we
leverage the attention mechanism of pre-trained causal language models such as
Transformer model for the downstream task of generating Failure Analysis
Triplets (FATs) - a sequence of steps for analyzing defected components in the
semiconductor industry. We compare different transformer models for this
generative task and observe that Generative Pre-trained Transformer 2 (GPT2)
outperformed other transformer model for the failure analysis triplet
generation (FATG) task. In particular, we observe that GPT2 (trained on 1.5B
parameters) outperforms pre-trained BERT, BART and GPT3 by a large margin on
ROUGE. Furthermore, we introduce Levenshstein Sequential Evaluation metric
(LESE) for better evaluation of the structured FAT data and show that it
compares exactly with human judgment than existing metrics.

本文研究了利用 Transformer 模型的注意机制为下游任务生成半导体行业中的故障分析三元组（FATs）。发现使用 1.5B 参数进行训练的 GPT2 模型在 ROUGE 上的表现显著优于 BERT、BART 和 GPT3 等其他 Transformer 模型，在人类评估和结构化 FAT 数据方面引入了 Levenshstein 顺序评估度量（LESE）。