It is now an evident trend that automated survey using social media as sources complements and will eventually largely replace manual surveys. That is an unstoppable direction as social media are becoming the major outlets of public opinions.  The technology is ready too.

Automated survey, or auto poll, refers to the use of computers to collect the public opinions and sentiments on a topic.  The data sources are social media big data where people are discussing most every topic all the time.  The technology is a parser that reads social media posts and mines salient information (facts, evaluations and emotions) about any topic.  More specially, deep information extraction and sentiment analysis are the required and mature text mining technology that can be enabled by an underlying parser.  This is the part of Artificial Intelligence that is proven to work and has been serve the clients in the business world (e.g. our customer insight products).  

Polls can provide quantitative information for decision-making in government, businesses and the general public, enjoying an extremely wide range of applications for many years. The presidential election is a prominent example, polls are conducted from time to time during the election to inform the voters as well as the president candidates how the public feels about the race so voters can make an educated choice and the candidate president teams can adjust  their policies and campaign strategy to enhance their public image.  Product launch is an example of the enterprise, feedback collected from customer surveys can help businesses to detect issues and to address them.

Auto-poll is dong the same, just that it is doing it much faster, more comprehensive, in a larger scale and is less costly.  Compared with the traditional manual questionnaires or polls, auto-poll has the following salient features.

Real time.  No need to go through a series of traditional survey process, designing the questionnaire, distributing them or by telephone interviews or street interviews, collecting and summarizing the results, with all steps carried out manually.  It often takes days or even weeks to complete a serious survey.  But auto-polls are instant, you get results as soon as you enter your topic.  As long as there are people discussing it , the insights will mined out of the text sea.  For any topic, using automated survey is as easy as using a search engine with the same response time but much more accurate results,  Our deep parser reads social media day and night to feed our storage just as a search engine indexes the Internet in their storage.  

Low cost.  Manual surveys are constantly struggling between the required costs and the scale of surveys (bigger scale reduces the error margin to be more reliable and convincing).  They often have to compromise the sample size given the budget.  Auto-poll is done fully automatically by the system, and the same system can serve a variety of different customers in different topics, each poll is inexpensive, costing just a fraction of the traditional poll.  The sample size can easily be magnitudes higher than that of manual surveys (often millions of data points vs. several thousand data points), way beyond the reach of most traditional polls.  

Objectivity. Traditional polls or surveys need to design a questionnaire, which may intentionally or unintentionally introduce subjective bias or implied suggestions.  Auto poll is bottom-up data analysis and mining, hence more objective by nature.  The public opinions are collected from the natural comments people make on topics, not as a response to a designed specific question. Moreover, in order to collect sufficient number of survey responses, the investigators administering the surveys sometimes need to offer incentives, which introduce a possible bias because some customers who answer the surveys too quickly to be honest, mainly do so to gain rewards, not to really air their opinions, causing the return of low-quality or polluted results.

Multi-topic comparison.  This is particularly important, because almost for any topic, we need a competitor or industry as background to figure out the real image in public mind.  For example, the poll on Obama's presidential campaign's effectiveness is of little sense if it is not contrasted to his rival Romney. Likewise, customer surveys on AT&T's cellular network service is inseparable from comparisons of its competitors like Verizon.  Ideally, a full picture will be clear on one brand once it is in comparison with all leading brands in the same industry.  In theory, manual surveys can also perform multi-brand comparisons, but in practice, the costs and time required to investigate many brands at the same time are often beyond feasible.  Investigators have had to reduce or sacrifice on the front of investigation of competitors, and use the limited resources on their own brand. Automated survey is different, multi-topic survey and comparison of these products is designed as a feature in these systems, it is just as easy as surveying one brand in this fully automated environment.  BPI (Brand Passion Index) in our products is one such feature that instantly surveys multiple brands in one industry and compar them in three dimensions, buzz (size of the bubble), popularity (up or down in the graph), passion intensity (right or left in the graph: the more right, the more intense).  For example, the illustration of BPI for the US retail stores gives a clear picture of the landscape and where each brand stands in its space..

In short, we are entering a big data age with no short of information on any topics you may need to study.  With mobile-web and social media in everyone's hands, public opinions and sentiments are buried in the big data calling for deep technology to mine.  Thus, there is absolutely no doubt that automated survey will become the direction of polls as the mainstream. Its supporting technologies are mature, large-scale multi-lingual text mining system that parses and reads big data around the world is just around the corner.

Related posts in my original Chinese blog:

【立委科普：自动民调】

奥巴马赢了昨晚辩论吗？舆情自动检测告诉你
社会媒体舆情自动分析：马英九 vs 陈水扁
舆情自动分析表明，谷歌的社会评价度高出百度一倍
【置顶：立委科学网博客NLP博文一览（定期更新版）】