Sentiment analysis for Twitter

The aim of this component is to infer the opinion of consumers about various activities and actions taken by the LABs. To achieve this, this component gathers information related to the relevant actors from social network Twitter and performs sentiment analysis of the associated information (text of tweets and associated images), see Fig. 1. The sentiment analysis is performed with the help of latest technologies available in NLP (Natural Language processing) and Image Processing domains, namely the Artificial Neural Network (ANN) -based algorithms. After that, the results are aggregated and summarized. Finally, they are presented in a form of plots and tables for further analysis by a user. This information can later be used to correct action plans and perform accordingly.

Fig. 1: A schematic representation of the component

The provided visualization of different granularity can include historical (Fig. 2) and geographical (Fig. 3) distribution of tweets of different sentiment (Positive, Neutral and Negative).

Fig. 2a: Historical distribution (per day) of tweets neutral (NEU), negative (NEG) and positive (POS) tweets
Fig. 2b: Historical distribution (per hour) of tweets neutral (NEU), negative (NEG) and positive (POS) tweets
Fig. 3: Geographical distribution of tweets

Additionally, the component can perform the analysis of the images associated with the tweets to better target specific events/actions/discussions. A relevant example is shown in Fig. 4.

Fig. 4: Results of image analysis