NGI ENGINEROOM

News media

• The figure shows trending terms that frequently appear in articles with most relevant social issues
• Artificial intelligence and machine learning are key elements of discussion
• Elections and referendums are linked to Cambridge Analytica, Facebook and fake news
• Privacy issues are debated in the context of GDPR and location data

Reddit

• To dive deeper, co-occurrence is examined in Reddit comments
• Reddit is an open, pseudonymous community, where emerging technology and related challenges are popular topics
• Different vocabulary, more detailed discussion
• The figure shows the terms that relatively more frequently appeared in Reddit comments than in news media articles
• Blue outline: not trending in online news
• Insight on emerging open-source programming tools (Scikitlearn-Python), hardware (Odroid) and privacy enablers (Duckduckgo)
• Polarizing discussions on social issues: e.g. gig economy and serfdom, future of cryptocurrencies (Beanie Babies)

Filtering process

• Most relevant, trending uni/bigrams related to social issues are selected for analysis
• The co-occurrence is calculated for word pairs consisting of these terms and the most trending 15000 bigrams
• For each social issue, a separate filtering has been carried out
• The terms are sorted by the index described in the methodology section, and 10 most relevant are picked from the first 30 words
• All raw results are available. The columns represent the social issues, while rows the trending words. Columns ending with count_freq_weighted contain the index used for sorting
• Raw data: social issue unigrams
• Raw data: social issue bigrams

Issue mapping

Articles have been assigned to the European group if they have been written for a European source (The Guardian, The Register, The Conversation, Techforge, Euractiv or Politico.eu) or contain a noun or adjective representing European Union member states or their capitals. Otherwise, they are marked as American. It is a simplification, “Rest of the World” may have been more precise in some cases (see: ACCC – Australian Competition and Consumer Commission), but a vast majority of issues is related to American companies, products and politics.

Social words list has been derived through Latent Dirichlet Allocation analysis of recent articles in sources related mostly to social issues. Some topics remained technological, but others were clearly social. Out of social topics, 75 words have been chosen and their mean inverse frequency by month checked. Words which are social, but common have a lower inverse frequency (like tax: 4; on average, there have been 4 times more articles than occurrences of the word “tax”), and less frequent words’ inverse frequency is higher (e.g. discrimination: 35.5). If the sum of inverse frequencies of social words occurring in an article exceeds 25, it is a social article; otherwise it is categorized as technological. It does not matter how many times a word occurs in an article, so longer articles do not have a significantly increased probability of being classified as social.

After the assignments, top trending terms have been ranked according to their frequency in European and social articles. Because of bias in the choice of sources – American sources included in the analysis are specifically technological, while some European ones (like the Guardian) are rather mixed or even clearly social (Politico.eu), the result was an apparent linear dependence – the more frequently a word occurs in social articles, the more European it is. To find the terms which are more European than expected, we performed OLS regression on the ranks, to modify the EU/US axis to residuals from the regression. Even if a word is more often used in the American sources and is technological – like kubernetes or Python – it is shown on the European half, because it is more European than one could expect given the choice of sources.

Out of most social topics, anti-vaccine and Flat Earth movements and net neutrality repeal are clearly American. Common European social topics include democracy, GDPR, Cambridge Analytica and ethical issues. Apparently, the strengths of Europe are 5G, innovative cloud solutions, cryptocurrencies and possibly quantum computing, but the hardware-related discussion is limited to the US.

Sentiment analysis

The sentiment analysis has been prepared using VADER*, an open-source lexicon and rule-based sentiment analysis tool. VADER is specifically designed for social media analysis, but can be also applied for other text sources. The sentiment lexicon was compiled using various sources (other sentiment data sets, Twitter etc.) and was validated by human input. The advantage of VADER is that the rule-based engine includes word-order sensitive relations and degree modifiers.

As VADER is more robust in the case of shorter social media texts, the analysed articles have been divided into paragraphs. The analysis have been carried out for the social issues presented in the co-occurrence exercise.

The process included the following main steps:

• The 100 most frequently co-occurring terms are identified for every social issue (using the co-occurrence methodology)
• The articles containing the given social issue and co-occurring term are identified
• The identified articles are divided into paragraphs
• Social issue and co-occurring words are removed from the paragraph
• The VADER sentiment analysis is carried out for every identified and modified paragraph
• The average for the given word pair is calculated for the final result

Therefore, the procedure has been repeated for 100 words for all identified social issues.

The sentiment analysis resulted in a compound score for every paragraph. The score is calculated from the sum of the valence scores of each word in the paragraph, and normalised between the values -1 (most extreme negative) and +1 (most extreme positive). Finally, the average is calculated from the paragraph results. Removal of terms is meant to exclude sentiment of the co-occurring word itself, because the word may be misleading, e.g. when some technologies or companies attempt to solve a negative issue. The neighbourhood's scores would be positive, but the negative term would bring the paragraph's score down.

The presented tables include the most extreme co-occurring terms for the analysed social issue. The examples are chosen from the list of words with 30 most positive and 30 most negative sentiment. The presented graphs show the evolution of sentiments for social issues. The analysed paragraphs are selected the following way:

• The articles containing the given social issue are identified
• The paragraphs containing the social issue are selected for sentiment analysis

Charts

This section compares the trends of terms across different sources and provides interesting insights. Tech journalists and scientists have different focus in their analyses, which is well reflected by such words as AI or machine learning. These more general words have a robust trend in popular news sources, with much less significant increase in academic sources. On the other hand, scientists do research on smaller sub-fields in the field of artificial intelligence, such as deep neural networks.

The charts also show the difference in social science (SSRN) and computer science (arXiv) research fields. As an example, the term "fake news" has been on a stronger trend in abstracts of working papers published in SSRN than in Arxiv.

Trend robustness

• 100 most trending uni-grams over 38 months (January 2016-February 2019)
• Additional analysis of last 3, 6 and 12 months
• The table shows that trends are not linear
• AI achieved the second largest increase over 38 months, but has not been increasing in the last 12, 6 or 3 months
• Verification of methodology: news articles are covering future events and predictions (term 2017 is not rising in the last 12 months, while 2019 has been increasing in last periods)

Topic models – Latent Dirichlet Allocation

Topic models are of great help in automating the analysis process of exploring the structure of a large set of documents by clustering documents based on the words that occur in them. It is assumed that documents on similar subjects tend to use a similar vocabulary. Topic modeling is one of the most powerful techniques in text mining used for discovering the latent topics that occur in collection of text documents and examining the relationships among them (Jelodar et al. 2017).

Visualization explanation: the bubbles represent topics identified by the LDA. The number of topics is arbitrarily set to 20. The positions of bubbles are determined by computing the distance between topics, and then the inter-topic distances are projected onto two dimensions (see: Chuang et al., 2012). The topic’s prevalence is shown by the size of the bubble. In the right panel, bars represent the individual terms that help interpreting the currently selected topic on the left. A pair of overlying bars represent both the corpus-wide frequency of a given term as well as the topic-specific frequency of the term (see: Chuang et al., 2012).

Countries map

The exploration of research strengths in the field of computer science and related social sciences is presented in this section. The main source of data is ACM that manages a large portfolio of scientific conferences, journals, books and magazines. A unique data set was prepared from the publications in the ACM Digital Library, containing the metadata of publications by institutions that are part of the 37 ACM Special Interest Groups. As ACM contains publications across most areas of computer science, the papers were divided into 3 groups based on their keywords for a more detailed analysis. The 3 groups are: AI/ML, social issues and technology.

Top institutions

The top 2 institutions are located in the EU: Karlsruhe Institute of Technology and University of Toulouse. From the top 10, the sole US based institution is Google Inc, while the remaining 7 are all located in China.
Considering the number of papers in the AI/ML area, Chinese institutions publish the largest quantities. The only EU institution in the top 10 is the University of Athens. The US is again represented by an institution that belongs to a major technology company: Yahoo Research Labs.
European institutions have a rich output in papers related to social challenges, as 7 out of 10 institutions are EU based.
In the papers discussing technologies, Asian and EU institutions form the top 10.

Clusters

The presented maps show the clusters of institutions whose affiliated researchers were likely to publish together. 5 maps were prepared: one based on SSRN data, and 4 based on ACM, also taking into consideration the different research areas. Each cluster has its own unique color on the map. Additionally, the tooltip reveals the number of the corresponding cluster for every institution on the interactive visualisations.
The analysis reveals that cooperation between institutions in computer science and social science is mostly limited to areas with geographical and cultural proximity. This is especially visible on the default map with all publications from the ACM data set. Most clusters are constrained to the same continent, while the exceptions (e.g. Iberian Peninsula - South-America) may be explained by reasons related to history or distance (China-Australia).
The separate network analyses on papers in the field of AI/Technology/Social issues show a similar geographical and cultural fragmentation. Therefore, research is typically carried out in teams that are in close proximity of each other, or with strong cultural connections.

Meetup

The data set contains metadata on Meetup groups in Europe and North-America. Meetups are particularly popular among tech-savvy persons, forming groups around various technologies and programming languages. Meetups are being held in all EU member states, making the platform an attractive source of information to track the activities of tech communities. The examined meetup categories are: Tech, Career and Business. Furthermore, meetups have been categorised into 3 groups, based on keywords in the name and description:

• Web (focused on JacaScript, HTML, web)
• Agile (agile task planning, popular in tech projects)
• AI (artificial intelligence, machine learning)