Perceptions are the experience which people gain about the world via the sensory system [1]. Perception is a key step of the human perceptual process. Followings are the steps belong to the human perceptual process,
These are the inputs for the human perceptual process. These inputs are collected from the human sensory system.
Converting the stimulus in to an electrical signal is transduction. These electrical signals are generated in the human nerve system.
The electrical signals generated by the transduction are propagated through series of interconnected neurons to the brain. The path obtained for the propagation of the signal is dependent to the type of the signal. Auditory signal go through one path, visual signals go through another path etc. The processing happens between the propagation of the signal and at the brain is known as neural processing.
The point where the human is consciously aware about the stimulus is known as the perception. This happens after the neural processing.
Categorizing and Interpreting stimulus is known as the recognition. This happens inside the human brain.
The response given to the identified environment stimulus is known as the action. This is the final step of human perpetual process.
The perception step, recognition step and the action step of the human perceptual process are heavily dependent on the previous experience and the knowledge of the human. Importantly those indicate the internal characteristics of the human mind. For an example consider a situation where a person is watching a movie. The liking about the movie is totally dependent to his previous experiences. His comments and criticisms about the movie will be aligning with his mindset. These comments and criticism provides a perfect reflection of human mind.
Careful identification and interpretation of human perceptions, recognitions and actions can reveal human mind. That is the basic of perception analysis. Basically perception analysis is identifying peoples mind by analyzing the data related to perceptions. Within the past few decades researchers have developed many mechanisms to do perception analysis. There have been many motivations and advantages behind these research. All these motivations will be discussed in the next sections.
The most important thing in perception analysis is finding human perceptions related data. In the past, people have developed explicit mechanisms to collect perceptions related data. Surveys, questioners, interviews and think aloud [2] are such mechanisms. When these mechanisms are in action people have to participate explicitly to provide data related to their perceptions. That participation can be answering to a Likert Scale [3] based questioner, writing answers to questions and speaking etc.One of the previous sections of this literature survey has comprehensively discussed the surveying techniques, their advantages and weaknesses.
The recent trend in collecting perceptions related data is using passive mechanisms. Development of information technology sector was the main reason behind this trend. In the current context, the regular life of a person has been changed a lot with respect to the past. People who used to read books, newspapers to explore the world, now use handheld gateways to access to the internet. A mobile phone, a tablet or any other smart device allows them to access diversified content all around the world. That was enabled by the invention of World Wide Web (WWW) in 1989. But the earliest web; i.e. web 1.0 restricted content generation and sharing to site authors. With those restrictions, only a specific sector including authors, journalists and critics were given the chance to express themselves publicly. But the development of technology has facilitated each and every person to express their thoughts using media like blogs, social networks and video sharing sites such as YouTube. Ultimately a regular person has been evolved to an active writer and a viewer with the transition of web 1.0 to web 2.0. In web 2.0 the end users of WWW actively participate in content generation which was dominated by server administrators in web1.0.
With the movement of content creation capability to general public, people started to create content in different forms such as blog articles, videos, tweets, Facebook status updates, comments etc. That crowdsourced contents reflect various psychological aspects of a human being having such sensations, opinions and attitudes. Also these content are available in public. Thus researchers developed mechanisms to use these publicly available data in perception analysis. Such perception analysis mechanisms are passive since those data have been created for other purposes. Sentiment Analysis is the most popular passive mechanism for perception analysis. The previous section on that topic provides a comprehensive discussion on this.
Accurate analytics was the main advantage of the explicit mechanisms which were used in the earlier stages. The main reason behind that accuracy was the collected data were indented to highlight perception. Also these mechanisms guaranteed the privacy of the participants. The main disadvantage of this is the effort needed in the process. Passive mechanisms provide much effective and effortless ways of perception analysis. But the accuracy and privacy of user have some issues.
One of the first motivations of perception analysis is physiological research. Physiologists wanted to understand behavior of the mind. That interest provided the birth to this research area. There are many research done in this area. For an example, Munmun, Scott and Michael from Microsoft Research have done a research one human emotional status in social media [3]. They have analyzed the human emotions related data in different angles. Those angles are like this,
By looking at the above list, anyone can understand how compressively they have analyzed the social media and the behavior of people. To do this kind of analytics of human emotions, perception analysis on social media is mandatory.
Some researchers wanted to develop cognitive tools to simulate the human mind inside the computer. “HiTEC” is such a cognitive tool which used to simulate human computer interaction and reactions [4]. The researchers propose this “HiTEC” cognitive tool as an alternative for human subject in user interfaces testing. They have developed this cognitive tool by analyzing perceptions and responses of human computer interactions.
Marc, Micheline and Koen from Ghent University, Belgium have developed a computerized music analyzing platform called “IPEM Toolbox” [5]. This toolbox is used to mimic the perceptual process of human for music subjects. They have used the toolbox to categories and score music without using human subjects for evaluations. For the development of tools box they have done enormous amount of perception analysis.
Likewise to model human mind, sophisticated perception analysis is mandatory. Thus researchers have positive influence for their research in perception analysis.
Sales and sector is driven mainly one the requirements of users and customers. Their main focus is to supply what customer wants and emphasis the importance of company’s products to customers. To identify the customer behavior and their needs, sales and marketing people uses many workaround methods. “Rightwave” is a cloud service which provides marketing as a service [6]. They provide methods to track customer’s behavior in a company. Also “Google Analytics” provides web based analytical methods to find details about viewers of websites [7].
The requirement of understanding thoughts, feelings of customers and users motivated the research in perception analysis. Because accurate perception analysis can provide valuable data related to customers. The current development of perception analysis has positively influenced in marketing and sales. Marketers and product managers can align their products and services according to trends and preferences of consumers through accessing perceptual data which would impact positively in profit maximization. Those influences provide more motivation to the research to find more and more effective and accurate methods.
Individual or organizational level decision making process can be supported with crowdsourced perception analytics results. Often, people search through online product review sites such as GSMArena [8] before purchasing a product. Their decision is heavily influenced with the reviews and comments by the users who are already consuming the product. Organizational decisions such as introducing a new product or a service are also affected by perceptions of crowd regarding trends and their favors. Individual discussion like purchasing, travelling destinations selection, educational pathway selection can also be supported by perception analysis results.
Other major concern of individuals and organizations is improving reputation. It is more effective to concern about social trends, opinions in managing reputation. Those trends and opinions provide clues to work on to maximize reputation. Individuals and organization can use methodologies like surveys to obtain details about trends and opinions of the society.
Usability and acceptance checking is another major motivation for perception analysis research. The current business domain is more customer oriented, which mainly focuses on customer acceptance and satisfaction.
User interface design for software products is highly dependent on user’s acceptance and their comment on usability. Because of this importance, people have developed methods like “Think Aloud” to get feed backs from users [9]. This method uses colleting verbalization of thoughts came across the user’s mind when they use the user interface [2]. The collected data will be analyzed later to find important results. Also researchers further improve these methods by tracking the eye movement, tacking the cursor movement of user to find the acceptance of user [2].
Human perceptions rapidly change when they are engaging in activities like watching TV shows and movies, lectures, conferences etc. Being aware of those rapidly changing perceptions is invaluable to the lecturer, speaker, producer or the performer as a measure of success of their product and to understand the desires of the audience.
The real time analyses of user’s perceptions are important for many parties and they can use the result in many ways in maximizing their works. For an example if there is a way to analyse perceptions in real time, presenters can adapt on the requirement of audience [10]. For an example this concept was used in the 2012 presidential election in USA. Limited number of audience was given the chance to give their rating to live presidential debate [11]. The analytical data were published in both real-time and non-real-time. So debaters can align their arguments and speech watching the real-time feedback. Another example would be TV program production. They value the user’s feedback to the content they telecast. Focusing on that importance, researchers have developed interactive TV concept to get users feedback. Some cable TV operators provide mechanisms to viewer to rate TV programs. “iTV” is such an interactive TV concept which allows user to give real-time feedback in many areas like content, camera angle etc. [12].
If we think about the first focus of the process, we have to find data where people express their thoughts and opinions. From the ancient time people were readers and writers. So the first focused data category was text. People have used text to express their opinions in places like Twitter, Facebook, Blogs, Reviews in shopping sites etc. Those text were public most of the time. So researchers developed ways to identify perception of the writer of that text content by processing those text. This process involved processing of natural language statements. This process has specially named as sentiment analysis. We have talked about that process much in this article also. But here we will highlight some drawbacks of this particular method where researchers have tried finding more ways to do perception analysis.
The accuracy of understanding the perceptions of people using the text written by them is less. The reasons are existing sentiment analysis have problems in named entity recognition (what the person actually talking about). Also parsing sentences is difficult. (What are subject and the object of the sentence etc.) Identifying and analyzing sarcasms, tracking abbreviations, spellings and punctuations is bit complex. Also poor grammars of text make sentiment analysis even hard. Also the write-ups are relative to the culture of the people. Different sentence can mean different things in different cultures.
Most of the time existing implementation can do only polarizing. That means they say the sentence is positive on the subject or negative or neutral on this subject. That is not sufficient to have a good understanding of perceptions.
Other way of capturing perception is getting feedback using a questionnaire. Most of the time place where service offered, this methodology is used. A well organized questionnaire with Likert items is given to people to provide there feed backs. Collecting big number of such feedback can be used to do a good perception analysis. Regardless of the accuracy of this method, this makes trouble to people. Because they are forced sometimes to do this while this process takes much time. So this has some negatives.
Some people have used videos, voice to provide the feedbacks and their opinions. So researchers tried to analyses those videos, voice using AI, Image processing and Sound Processing. But that task also a painful hard task.
There are other ways of capturing perceptions, opinions of people by observing them. By observing the visited sites, viewed videos we can understand the perceptions. But this method has some privacy issues also.
Also there are many others out of the box ways to capture perception of people. Some researchers trying to develop a game kind of a thing where user have to organize some mixed set of pictures or objects. By classifying the organizing patterns researchers trying understand the opinion or the perception of the people. Also some researchers have tried to build a neural network to identify the sensations of cloth to its wearer. By identifying that sensation they are trying to understand the perceptions.
This literature review highlights that the perception analysis is a broad area with multiple use cases in many domains. The perception analysis started from active (direct) methods such as survey techniques and has been evolved towards passive (indirect) methods such as sentiment analysis. All these methods have inherent drawbacks which have been discussed in previous sections. The current trends of perception analysis which focuses on overcoming these drawbacks will be discussed as the future research direction of this survey.
The current research on perception analysis have following architectural concerns,
One of the main architectural concerns of current perception analysis research is to crowdsource the perception analysis process. This requires people to actively participate in the process of perception analysis by expressing their emotions, thoughts, opinions etc. Many research have done to introduce new explicit perception capturing/sharing mechanisms, replacing existing passive perception capturing mechanisms. The main objectives behind these research are to achieve higher accuracy and validity in perception analysis while guaranteeing privacy [15]. In addition to analytics there are many psychological, health and social advantage of expressing/sharing emotions, opinions and thoughts [16].
There are many research and prototype level implementations done on the concept of explicit perception sharing and crowdsourced perception analysis. “Mobimood” [15], “Aurora” [16], “Mappiness” [13], “Got a feeling” [14] are such prototype mobile applications which have been used to share emotion within a group of people. In those applications, people can explicitly express their feelings using colours, pictures and texts. People who use the application can view the perceptions and feelings of the other people and give feedbacks, ideas, opinions on those perceptions.
This explicit perception sharing &crowdosurced perception analysis concept has created a network of perceptions. Some researchers have suggested social networks on top of this concept. “IRENE” [17] and “EmoSoNet” [18] are such conceptual social networks of emotions and perception sharing.
Crowsdsourced perception analysis process has to be facilitated with perception capturing mechanisms, where the users can convey their exact perceptions in a simplest way with minimum clicks, readings and navigations. That objective is closely related with psychometrics in which we quantify the psychological variables. The GUI should support to convey perceptions with minimum effort. Facebook ‘like button’ is only single click, but it has an enormous value addition for whole Facebook echo system. It enables expressing users’ positive feeling on some content in a very easy manner and now it has become a part of users’ lifestyle. The introduction of new mechanism for expressing feelings has to guarantee that it will be a part of user’s lives without being inconvenient because the success of this kind of system mainly depends on how it attaches with users’ lives and how extensively it is used.
The perception capturing mechanisms also should be diversified in order to avoid restricting perception sharing by the location, the tasks they are involving, hardware limitations etc.
The future perception analysis process will be highly dependent on crowdsourced perception data from unlimited end users. Thus the rate of events generated by end user applications is extremely high and there might have spikes in event stream which cannot be anticipated. The events which are generated from users have to be categorized based on event generator or the user, the action related with the event, event generating location etc. for the convenience of data analysis. Some of the events are related with other events and tracking those relations is significant. Thus capturing, processing, and analysis of user generating events is one of the major concerns in perception analysis. Crowdsourced perception analysis process have to be concerned with real time complex event processing programming model for large clusters of events that provide consistency, efficient fault recovery, efficient query processing of real time events and powerful integration with batch systems.
In a corwdsoruced perception analysis process, perception related data sets will be collected in huge amounts, in various complex forms. Since this is a crowdsourced system, amount of generated data will show an exponential growth with the number of users. Also there will be very high data generating rates and analysing the stored data either in real-time or as batch processes is essential. Therefore traditional data analysing methodologies and database management tools are not appropriate and the usage has to be limited for the new system. In this kind of corwdsourced perception analysis approach, the data from various media, many formats and types has to be analysed to give an integrated analytical results, hence managing the heterogeneity of data is one of the other concern. Due to these reasons the storage system related with perception analysis process has to be coped with big data management and storage mechanisms based on the programming models like Map-Reduce for large scale analysis [19] [20], and distributed database with NO- SQL concepts [21].