Context:

Programming by Example (PBE) is increasingly assisting human users by recognizing and executing repetitive tasks, such as text editing and spreadsheet manipulation. Yet, existing work falls short on dealing with rich-formatted documents like PowerPoint (PPT) files, when examples are few and collecting them is intrusive.

Objective:

This article presents TIDY, a PBE-based framework, to assist automated entity transformations for their layout and style consistency in rich-formatted documents like PowerPoint, in a way adaptive to entity contexts and flexible with user selections.

Methods:

TIDY achieves this by examining entities’ operation histories, and proposes a two-stage framework to first identify user intentions behind histories and then make wise next-operation recommendations for users, in order to maintain the entity consistency for rich-formatted documents.

Results:

We implemented TIDY as a prototype tool and integrated it into PowerPoint as a plug-in module. We experimentally evaluated TIDY with real-world user operation data. The evaluation reports that TIDY achieved promising effectiveness with a hit rate of 77.3% on average, which was stably holding for a variety of editing tasks. Besides, TIDY took only marginal time overhead, costing several to several tens of milliseconds, to complete each recommendation.

Conclusion:

TIDY assists users to complete repetitive tasks in rich-formatted documents by non-intrusive user intention recognition and smart next-operation recommendations, which is effective and practically useful.

中文翻译：

---

TIDY：基于PBE的框架，支持智能转换以实现PowerPoint中的实体一致性

语境：

通过示例编程（PBE），通过识别和执行重复性任务（例如文本编辑和电子表格操作），越来越多地为人类用户提供帮助。但是，现有的工作不足以处理诸如PowerPoint（PPT）文件之类的格式丰富的文档，而示例很少且收集起来具有干扰性。

客观的：

本文介绍了TIDY（一种基于PBE的框架），它以适应实体上下文和灵活地用户选择的方式，协助自动实体转换，以在诸如PowerPoint之类的格式丰富的文档中实现布局和样式的一致性。

方法：

TIDY通过检查实体的操作历史记录来实现此目的，并提出了一个两阶段的框架来首先识别历史记录背后的用户意图，然后为用户提出明智的下一步操作建议，以保持格式丰富的文档的实体一致性。

结果：

我们将TIDY用作原型工具，并将其作为插件模块集成到PowerPoint中。我们使用真实的用户操作数据对TIDY进行了实验评估。评估报告指出，TIDY取得了可喜的效果，平均命中率为77.3％，在各种编辑任务中均保持稳定。此外，TIDY仅花费了很少的时间开销，即花费了几到几十毫秒来完成每个建议。

结论：

TIDY通过非侵入式用户意图识别和智能的下一个操作建议来帮助用户完成格式丰富的文档中的重复任务，这是有效且实用的。