Integration, Use And Analysis Of Crowdsourcing Documents

Software components are typically reused via Application Programming Interfaces(APIs).Learning new APIs is a major obstacle faced by software developers.One of the reasons for such a difficulty is that API documentation is often dispersed over many sources,including official documents,Q&A websites,wikis,blogs,etc.,and the missing pieces are sometime hard to find.One solution to this problem is to glean scattered documents for an API and organize them in a way that facilitates authoring,inquiry,and sharing.In general,most existing technologies for this problem are based on linguistic and textual features and attempt to create an automatic solution.Because of the unstructured and imprecise nature of human languages,it is difficult to accurately collect and stitch fragmented information into a human-readable document.In recent years,there has been a great interest in integrating crowdsourced API documents that are often dispersed across multiple places.Because of the complexity of natural language,however,automatically synthesized documents fall short on quality and completeness compared to those authored by human experts.We investigate the benefits and limitations of integrated documents in the context of programming education.We created a new “human-as-sensor” approach to document integration and experimented with a prototype on 13 novice programmers in a programming task.When a programmer faces a problem with an API,she often takes a sequence of actions to find the answer,including reading official documents for the API,conducting a Web search for the question,reading the search results for the query,and(implicitly)confirming the relevance of the result webpages(answers)to the search query.Our idea is implemented by a prototype system called Crowdsourced Online FAQs(COFAQ).COFAQ captures the context of an API learning session,and based on that context automatically creates a new FAQ section and insert it into the existing official document for the concerned API.Our results suggest: despite many integrated documents are rated as helpful,programmers still spend a significant amount of time digesting the information in the context of their own work.We also find that,often lack of information seeking skills,novice programmers use several strategies to seek and utilize API knowledge.