[Paper Review] Analysis of Titles and Readers For Title Generation Centered on the Readers

Yasuko Senda, Yasusi Sinohara|ArXiv.org|Jul 2, 2002

Advanced Text Analysis Techniques3 references4 citations

TL;DR

This paper proposes a reader-centered approach to title generation by analyzing how content and wording in titles affect reader comprehension, interest, and positive feelings. It identifies that expressing the purpose of technology (pattern 2.0) most effectively attracts general readers, while expert readers are less influenced by title structure, suggesting tailored title generation strategies based on target audience expertise.

ABSTRACT

The title of a document has two roles, to give a compact summary and to lead the reader to read the document. Conventional title generation focuses on finding key expressions from the author's wording in the document to give a compact summary and pays little attention to the reader's interest. To make the title play its second role properly, it is indispensable to clarify the content (``what to say'') and wording (``how to say'') of titles that are effective to attract the target reader's interest. In this article, we first identify typical content and wording of titles aimed at general readers in a comparative study between titles of technical papers and headlines rewritten for newspapers. Next, we describe the results of a questionnaire survey on the effects of the content and wording of titles on the reader's interest. The survey of general and knowledgeable readers shows both common and different tendencies in interest.

Motivation & Objective

• To shift title generation from author-centered to reader-centered approaches by analyzing reader engagement with titles.
• To identify typical content and wording features in titles that attract general readers, based on a comparison between technical paper titles and newspaper headlines.
• To empirically evaluate how different title expression patterns affect comprehension, positive feelings, and interest across diverse reader groups.
• To provide data-driven guidelines for generating titles that enhance reader engagement beyond mere summarization.
• To inform future title generation systems by identifying effective expression patterns for specific reader profiles.

Proposed method

• Conducted a comparative syntactic analysis of 1464 technical paper titles and 313 newspaper headlines, tagging components by function (e.g., behavior, object, purpose, method).
• Classified titles into expression patterns based on obligatory component structures, such as 'purpose of development' (pattern 2.0), 'behavior' (pattern 1.0), and 'method for realization' (pattern 1.1).
• Administered a web-based questionnaire to 1,500+ respondents across four reader groups: commoner, unconcerned, engineer, and researcher.
• Measured reader impressions using three metrics: comprehensibility, evoked positive feelings, and interest in the technology.
• Applied chi-square tests and Cramer’s V to assess the strength and significance of associations between expression patterns and reader impressions.
• Used contingency tables to analyze the relationship between title structure and reader response across different expertise levels.

Experimental results

Research questions

• RQ1 Which title expression patterns most effectively enhance comprehension among general readers?
• RQ2 How do different title structures influence the level of positive emotional response in readers of varying technical expertise?
• RQ3 To what extent does the reader’s expertise level moderate the impact of title wording and content on interest and engagement?
• RQ4 Which components (e.g., purpose, method, technology type) are most influential in attracting non-expert readers to technical content?
• RQ5 How do newspaper-style headlines differ from technical paper titles in structure and reader appeal, and what can be learned from this for title generation?

Key findings

• Expression pattern 2.0—explicitly stating the purpose of the technology—was the most effective in enhancing comprehensibility, positive feelings, and interest among general readers and commoners.
• The association between title structure and reader impression was strongest for non-expert readers (commoner and unconcerned), with Cramer’s V values of 0.62 and 0.59 for comprehensibility, respectively.
• For expert readers (engineers and researchers), the influence of title structure was weaker, with Cramer’s V values below 0.30, indicating lower sensitivity to wording patterns.
• All chi-square tests showed significance levels below 1%, confirming a statistically significant relationship between title expression patterns and reader impressions across all reader groups.
• Titles expressing the purpose of development (pattern 2.0) consistently outperformed other patterns in all three evaluation dimensions (comprehensibility, positive feelings, interest) for non-expert audiences.
• The results suggest that expert readers are less swayed by title structure, implying that title generation systems should adapt to the target reader’s expertise level.