Scholarlink

<p>One of the most important aspects of Human-Machine Interaction (HMI) is the recognition of human emotions through the analysis of texts. As a subset of sentiment analysis, the process of affect analysis examines and identifies various human emotions such as <i>love</i>,<i> hate</i>,<i> excitement</i>,<i> anxiety</i>, etc. through automatic analysis of texts. Conventional feature extraction methods for converting unstructured textual data to structured tabular data each deal with a special aspect of textual information. In this paper, we will focus on specific types of textual features derived from the texts<i>’</i> writing style. During this study, we will show that the characteristics derived from the writing style, the style of using the part-of-speeches, and the semantic style in a text, provide suitable clues to identify different emotions such as <i>joy</i>,<i> sadness</i>,<i> anger</i>,<i> satisfaction</i>, etc. These style-based features provide significant distinguishing power in identifying emotional dichotomies in textual data in very low-dimensional vectors, and each reflects an aspect of the text<i>’</i>s writing style. In this study, a developed set of style-based features is introduced in three separate categories <i>stylometric</i>,<i> stylo-syntactic</i>, and <i>stylo-semantic</i>, and their effectiveness will be examined using dissociation histograms, manifold-based mappings, and hypothesis testing.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A study on discriminative power of stylistic features in affect analysis

  • 期刊论文

  • 发表日期：

    2025年8月19日

查看全文
  • Emran Younesi,
  • Ebrahim Akbari,
  • Homayun Motameni

摘要

One of the most important aspects of Human-Machine Interaction (HMI) is the recognition of human emotions through the analysis of texts. As a subset of sentiment analysis, the process of affect analysis examines and identifies various human emotions such as love, hate, excitement, anxiety, etc. through automatic analysis of texts. Conventional feature extraction methods for converting unstructured textual data to structured tabular data each deal with a special aspect of textual information. In this paper, we will focus on specific types of textual features derived from the texts’ writing style. During this study, we will show that the characteristics derived from the writing style, the style of using the part-of-speeches, and the semantic style in a text, provide suitable clues to identify different emotions such as joy, sadness, anger, satisfaction, etc. These style-based features provide significant distinguishing power in identifying emotional dichotomies in textual data in very low-dimensional vectors, and each reflects an aspect of the text’s writing style. In this study, a developed set of style-based features is introduced in three separate categories stylometric, stylo-syntactic, and stylo-semantic, and their effectiveness will be examined using dissociation histograms, manifold-based mappings, and hypothesis testing.