Abstract Social networks are a great source of data for municipal decision making, as they allow citizens to freely express their opinions. But due to their unstructured nature and large volume, these data are difficult to process. Therefore, we propose a social network data analysis framework for municipal decision making consisting of four main steps: (1) topic modeling, (2) sentiment analysis, (3) combining the results of the previous two steps into a triangular fuzzy number, which captures sentiment diversity, and (4) calculating the levels of positive and negative sentiment. For step (3) we propose a method of constructing a triangular fuzzy number by calculating weighted mean sentiment and its weighted “standard semi-deviation”, and for step (4) we propose using degree of similarity with prototypical fuzzy sets representing the concepts of positive and negative sentiment. The most important practical innovation of the framework is its ability do capture not only some sort of an average sentiment towards a topic, but also its diversity. The framework was evaluated on about 30,000 tweets published from Ostrava, Czechia, over a period of 5 months. We demonstrate how analyzing sentiment diversity can be useful for municipal decision makers by extracting several practical recommendations from the results provided by the framework. We also show that the framework provides more information than a naive approach deployed by many commercial tools which fails to properly distinguish between mean sentiment towards a topic and its diversity.

Many cities around the world are aspiring to become smart.However, smart initiatives often give little weight to the opinions of average citizens.Social media are one of the most important sources of citizen opinions.This paper presents a prototype of a framework for processing social media posts with municipal decision-making in mind.The framework consists of a sequence of three steps: (1) determining the sentiment polarity of each social media post (2) identifying prevalent topics and mapping these topics to individual posts, and (3) aggregating these two pieces of information into a fuzzy number representing the overall sentiment expressed towards each topic.Optionally, the fuzzy number can be reduced into a tuple of two real numbers indicating the "amount" of positive and negative opinion expressed towards each topic.The framework is demonstrated on tweets published from Ostrava, Czechia over a period of about two months.This application illustrates how fuzzy numbers represent sentiment in a richer way and capture the diversity of opinions expressed on social media.

Abstract Valuable information for decision-making can be obtained by collecting and analyzing opinions from diverse stakeholder or respondent groups, which usually have different backgrounds and are variously affected by the topics under survey. For this to succeed, it is necessary to manage the uncertainty of respondents’ opinions, different number of filled questionnaires among groups, different number of questions for each stakeholder group, and relevance of subsets of respondent groups. This work proposes handling the hesitance of respondents’ opinions for the rating scale questions. To evaluate the collected opinions, a three-level aggregation model is developed. In the first level, the overall opinion of each respondent is computed as a mean of fuzzy numbers covering uncertain answers and their respective hesitance. In the second level, stakeholder groups are considered as a whole. Aggregation by a relative quantifier is applied to calculate the validity of a proposition the majority of respondents have a positive or negative opinion . At the third level, the consensus among diverse subsets of stakeholder groups is calculated considering the relevance of each group independently as well as their so-called coalitions by Choquet integral. Finally, the proposed model is illustrated by a real-life case study.

Choosing a proper representation of textual data is an important part of natural language processing.One option is using Word2Vec embeddings, i.e., dense vectors whose properties can to a degree capture the "meaning" of each word.One of the main disadvantages of Word2Vec is its inability to distinguish between antonyms.Motivated by this deficiency, this paper presents a Word2Vec extension for incorporating domain-specific labels.The goal is to improve the ability to differentiate between embeddings of words associated with different document labels or classes.This improvement is demonstrated on word embeddings derived from tweets related to a publicly traded company.Each tweet is given a label depending on whether its publication coincides with a stock price increase or decrease.The extended Word2Vec model then takes this label into account.The user can also set the weight of this label in the embedding creation process.Experiment results show that increasing this weight leads to a gradual decrease in cosine similarity between embeddings of words associated with different labels.This decrease in similarity can be interpreted as an improvement of the ability to distinguish between these words.