Sentiment Analysis and “Personal Influence”. A Theoretical and Methodological Comparison

In the digital age there are ever more areas in which the footprints and the traces we leave (voluntarily and not) take on particular relevance for the use (legitimate and not) that may be made of them; this creating new wide scenarios of analysis in different areas of interest: politics; equity markets; communication and marketing; sports; medical and natural sciences. In this context, one of the most widely diffused analysis tools with the massive extension of social media is the «sentiment analysis» (also known as «social media analysis» or «web sentiment analysis») (SA), which is now being studied both academically and commercially. Specifically, it consists in extracting from the web platforms the expression of users’ judgments and in the computational analysis of feelings and opinions expressed within the network through the interactions established between users, related to a product, a service, an individual, an event, etc., in a given space and in a defined time frame. This is closely related to the theme of networks and specifically the reconstruction of networks generated by the use of social media. For this purpose, the footprints that each person leaves behind are especially relevant, since from them it is possible to reconstruct the bonds and identify the strategic nodes. In fact, with the diffusion of one own opinions on the web, a volume of «big data» has been generated that is useful from various points of view (institutions, political figures, brands, etc.) because of the impact it can have. In this regard, SA's utility and applications are multiple: reputation; perception of brands, personalities and products; measuring social media marketing activities; degree of satisfaction; improving services and products; online prevention and crisis management; identification of competitors; identification of influencers. SA is on one hand a popular tool, just think of the many tools for its implementation made available by the same social media (Facebook Insights, TweetStats, Google Analytics, etc.), as well as the availability of paid platforms that offer this type of services (Talkwalker, Digimind, etc.). On the other hand, the methodological robustness of this type of analysis is often questioned, especially when compared to traditional (textual and not only) analysis tools and techniques, such as surveys (by telephone) and focus groups. The issue has recently been discussed by the relevant literature, that in addition to having dealt with some methodological questions related to the use of digital footprints and the data provided by them (accessibility in view of commercial and/or legal limits, ethics in their free use, representativeness), has highlighted how the two perspectives (traditional and modern), instead of being opposed, can be used in an integrated way. In this context, an additional element of interest at the heart of this contribution is the comparison between the scenario offered by the analysis of digital footprints using the above mentioned network analysis and sentiment analysis and the well-known «two-step flow communication theory». First exhibited in 1944 by Paul Lazarsfeld, later reworked with Elihu Katz in 1955, it deals with the social influence that the media indirectly exercise on the population, passing through the «personal influence» of the «opinion leaders», who play a strategic role in a certain social context, having first access to a certain type of information, then spread it to the others (large public) conveying it through their opinion, which will then have a certain influence on the opinion among the so-called «opinion followers». From this comparison, what one wonders about is whether what is happening in the context of the analysis of big data provided by the widespread diffusion of digital footprints is able to add some innovation element compared to what has already been highlighted by the« two-step flow of communication theory», or if it merely represents its explanation. A first point of reflection is that, while the concept of «personal influence» requires the existence of networks and influence that pre-empt the flow of communication, on the other hand the concept of «influencer» seems to allude to the fact that nodes are created because of such use of communication. Is something happening that Katz and Lazarsfeld had predicted? Or do network and sentiment analyses even contrast or contradict the two authors?