SBIR Phase I: Chart Explainer: Automatically Generating Natural Language Descriptions of Charts and Tables

点击翻译按钮获取中文摘要

This Small Business Innovation Research (SBIR) Phase I project addresses the problem of information accessibility by creating an innovative Chart Explainer, a generic natural language generation program which automatically composes fluent textual summaries and annotations of graphical and tabular information. The introduction of this new tool will enhance access to information for both specialized and general audiences. Self-sufficient textual summaries of charts and tables will make the information displayed in them Web-accessible to visually impaired people via existing text-to-speech tools. For sighted audiences, textual annotations will be slightly different, and serve more to complement what is easily viewed in charts and tables, summarizing and highlighting salient facts and trends. In addition, automatic production of summary text will make graphical and tabular information accessible to all audiences by phone CoGenTex envisions Chart Explainer to have a profound and beneficial impact in the information accessibility area, making graphical and tabular information Web-accessible to visually impaired people. Chapter 508 of the Federal Disability Act mandates that all the graphical and tabular information on the government Web sites be accessible to visually impaired people. Currently, Web accessibility to tables for the visually impaired is implemented by XML tagging of table cells, which are then read by the page reader devices. Web accessibility to charts is implemented using a list of data points on the chart. In both cases, the information is delivered in a very inconvenient, fragmentary way, which makes its integration and understanding difficult. In contrast, Chart Explainer will deliver graphical and tabular information in a much more natural way, conveniently summarizing it and highlighting its salient characteristics. For a general audience, the substantial impact will be in enabling phone accessibility to such information, as well as in enhancing understandability of charts and tables by providing annotations to them, which highlight salient facts and summarize the rest. Both these features will have a substantial impact on the applications in the business intelligence area, in particular, on corporate scorecarding, which commonly uses graphical displays of various kinds.