<p>In the context of enhancing the production efficiency of semiconductor chips during mass production, the utilization of automated test machines (ATE) becomes imperative for assessing chip quality. Test data sets generated by certain automated test machines are typically presented in text format and exhibit characteristics such as high dimensionality, heterogeneity, and substantial volume. This study takes the Ultraflex automated test machine, manufactured by Teradyne as an illustrative example. The output test results from this machine are formatted as text, possessing poor readability, thereby posing challenges in the analysis of test data. Addressing the complexities associated with the analysis of high-dimensional, heterogeneous, and massive text test datasets produced by Ultraflex test chips, this paper introduces a visualization method utilizing word embedding based on Natural Language Processing (NLP) and quality feature spectrum. Initially, NLP is employed to transform semi-structured data into a structured format, facilitating the extraction of pertinent information from the test data set. Subsequently, leveraging the structured test data, the quality spectrum of the chip is established, enabling a quantitative analysis of the chip's quality. The proposed method's efficacy is ultimately validated through the application to a set of ADC test data.</p>