Predictive Analysis on U.S. Midterm Elections on Twitter with RNN

We implemented an analysis (meaning both a method and a system) that aim to gauge local support for the two major US political parties in the 68 most competitive House of Representative districts during the 2018 U.S. mid-term elections.

The analysis attempts to mirror the “Generic Ballot” poll, i.e., a survey of voters of a particular district which aims to measure local popularity of national parties by querying participants on the likelihood they would vote for a “generic” Democrat or Republican candidate. We collect the tweets containing national parties and politicians in the 68 most competitive districts. By most competitive we mean that they are rated as: toss up50%-50%, or lean by the Cook Political Report.

This means we are addressing an extremely challenging analysis and prediction problem, while disregarding the simpler cases (everyone is good at predicting the obvious!).

Our solution employs the Twitter Search API to query for tweets mentioning a national leader or party, posted form a limited geographic region (i.e., each specific congressional district). For example, the following query extracts tweets on Republicans:

TRUMP OR REPS OR Republicans OR Republican OR MCCCONNELL OR ‘MIKE PENCE’ OR ‘PAUL RYAN’ OR #Republicans OR #REPS OR @realDonaldTrumpOR @SpeakerRyan OR @senatemajldr OR @VP OR GOP OR @POTUS

To limit the search to each congressional district, we use the geocode field in the search query of the API, which queries a circular area based on the coordinates of the center and the radius. Because of the irregular shape of the congressional districts, multiple queries are needed for each of them, therefore we built a custom set of bubbles that approximate the district shape.

For the analysis of the tweets, we adopted a Recurrent Neural Network, namely a RNN-LSTM binary classifier trained on tweets.

To build training and testing data we collected tweets of users with clear political affiliation, including candidates, political activists, and also lesser know users, well versed in the political vernacular.
The accounts selected yielded around 280,000 tweets in 6 months before election day, labeled based on the author’s political affiliation.

Notice that the method is a general political-purpose language-independent analysis framework, that can be applied to any national or local context.

Further details and the results can be found on this Medium post.

This work has been published as a short scientific paper presented at IEEE Big Data Conference in Seattle, WA on December 2018 and on a previous Medium post by Antonio Lopardo.

You can also download a poster format reporting the work:

poster-midterm

In case you want to cite the work, you can do it in this way:

A. Lopardo and M. Brambilla, “Analyzing and Predicting the US Midterm Elections on Twitter with Recurrent Neural Networks,” 2018 IEEE International Conference on Big Data (Big Data), Seattle, WA, USA, 2018, pp. 5389–5391.
doi: 10.1109/BigData.2018.8622441.
URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8622441&isnumber=8621858

The online running prototype, the full description of the project, its results, and source code are available at http://www.twitterpoliticalsentiment.com/USA/.

Understanding Polarized Political Events through Social Media Analysis

Predicting the outcome of elections is a topic that has been extensively studied in political polls, which have generally provided reliable predictions by means of statistical models. In recent years, online social media platforms have become a potential alternative to traditional polls, since they provide large amounts of post and user data, also referring to socio-political aspects.

In this context, we designed a research that aimed at defining a user modeling pipeline to analyze dis cussions and opinions shared on social media regarding polarized political events (such as a public poll or referendum).

The pipeline follows a four-step methodology.

 

  • First, social media posts and users metadata are crawled.
  • Second, a filtering mechanism is applied to filter out spammers and bot users.
  • Third, demographics information is extracted out of the valid users, namely gender, age, ethnicity and location information.
  • Fourth, the political polarity of the users with respect to the analyzed event is predicted.

In the scope of this work, our proposed pipeline is applied to two referendum scenarios:

  • independence of Catalonia in Spain
  • autonomy of Lombardy in Italy

We used these real-world examples to assess the performance of the approach with respect to the capability of collecting correct insights on the demographics of social media users and of predicting the poll results based on the opinions shared by the users.

Cursor_and_KDWEB_2018_paper_1_pdf

Experiments show that the method was effective in predicting the political trends for the Catalonia case, but not for the Lombardy case. Among the various motivations for this, we noticed that in general Twitter was more representative of the users opposing the referendum than the ones in favor.

The work has been presented at the KDWEB workshop at the ICWE 2018 conference.

A preprint of the paper can be downloaded from ArXiv and cited as reported here:

Roberto Napoli, Ali Mert Ertugrul, Alessandro Bozzon, Marco Brambilla. A User Modeling Pipeline for Studying Polarized Political Events in Social Media. KDWeb Workshop 2018, co-located with ICWE 2018, Caceres, Spain, June 2018. arXiv:1807.09459

Data Cleaning for Knowledge Extraction and Understanding on Social Media

 

Social media platforms let users share their opinions through textual or multimedia content. In many settings, this becomes a valuable source of knowledge that can be exploited for specific business objectives. Brands and companies often ask to monitor social media as sources for understanding the stance, opinion, and sentiment of their customers, audience and potential audience. This is crucial for them because it let them understand the trends and future commercial and marketing opportunities.

However, all this relies on a solid and reliable data collection phase, that grants that all the analyses, extractions and predictions are applied on clean, solid and focused data. Indeed, the typical topic-based collection of social media content performed through keyword-based search typically entails very noisy results.

We recently implemented a simple study aiming at cleaning the data collected from social content, within specific domains or related to given topics of interest.  We propose a basic method for data cleaning and removal of off-topic content based on supervised machine learning techniques, i.e. classification, over data collected from social media platforms based on keywords regarding a specific topic. We define a general method for this and then we validate it through an experiment of data extraction from Twitter, with respect to a set of famous cultural institutions in Italy, including theaters, museums, and other venues.

For this case, we collaborated with domain experts to label the dataset, and then we evaluated and compared the performance of classifiers that are trained with different feature extraction strategies.

The work has been presented at the KDWEB workshop at the ICWE 2018 conference.

A preprint of the paper can be downloaded and cited as reported here:

Emre Calisir, Marco Brambilla. The Problem of Data Cleaning for Knowledge Extraction from Social Media. KDWeb Workshop 2018, co-located with ICWE 2018, Caceres, Spain, June 2018.

The slides used in the workshop are available online here:

 

Iterative knowledge extraction from social networks

Yesterday, we presented a new work at The Web Conference in Lyon along the research line on knowledge extraction from human generated content started with our paper “Extracting Emerging Knowledge from Social Media” presented at the WWW 2017 Conference (see also this past post).

Our motivation starts from the fact that knowledge in the world continuously evolves, and thus ontologies and knowledge bases are largely incomplete, especially regarding data belonging to the so-called long tail. Therefore, we proposed a method for discovering emerging knowledge by extracting it from social content. Once initialized by domain experts, the method is capable of finding relevant entities by means of a mixed syntactic-semantic method. The method uses seeds, i.e. prototypes of emerging entities provided by experts, for generating candidates; then, it associates candidates to feature vectors built by using terms occurring in their social content and ranks the candidates by using their distance from the centroid of seeds, returning the top candidates.

Based on this foundational idea, we explored the possibility of running our method iteratively, using the results as new seeds. In this paper we address the following research questions:

  1. How does the reconstructed domain knowledge evolve if the candidates of one extraction are recursively used as seeds?
  2. How does the reconstructed domain knowledge spread geographically?
  3. Can the method be used to inspect the past, present, and future of knowledge?
  4. Can the method be used to find emerging knowledge?

This is the presentation given at the conference:

This work was presented at The Web Conference 2018, in the Modeling Social Media (MSM) workshop.

The paper is in the official proceedings of the conference through the ACM Digital Library.

You can also find here a PDF preprint version of “Iterative Knowledge Extraction from Social Networks” by Brambilla et al.

 

How Fashionable is Digital Data-Driven Fashion?

Within the context of our data science research track, we have been involved a lot in fashion industry problems recently.

We already showcased some studies in fashion, for instance related to the analysis of the Milano Fashion Week events and their social media impact.

Starting this year, we are also involved in a research and innovation project called FaST – Fashion Sensing Technology. FaST is a project meant to design, experiment with, and implement an ICT tool that could monitor and analyze the activity of Italian emerging Fashion brands on social media. FaST aims at providing SMEs in the Fashion industry with the ability to better understand and measure the behaviours and opinions of consumers on social media, through the study of the interactions between brands and their communities, as well as support a brand’s strategic business decisions.

Given the importance of Fashion as an economic and cultural resource for Lombardy Region and Italy as a whole, the project aims at leveraging on the opportunities given by the creation of an hybrid value chain fashion-digital, in order to design a tool that would allow the codification of new organizational models. Furthermore, the project wants to promote process innovation within the fashion industry but with a customer-centric approach, as well as the design of services that could update and innovate both creative processes and the retail channel which, as of today, represents the core to the sustainability and competitiveness of brands and companies on domestic and international markets.

Within the project, we study social presence and digital / communication strategies of brands, and we will look for space for optimization. We are already crunching a lot of data and running large scale analyses on the topic. We will share our exciting results as soon as available!

 

Acknowledgements

FaST – Fashion Sensing Technology is a project supported by Regione Lombardia through the European Regional Development Fund (grant: “Smart Fashion & Design”). The project is being developed by Politecnico di Milano – Design dept. and Electronics, Information and Bioengineering dept. – in collaboration with Wemanage Group, Studio 4SIGMA, and CGNAL.

logo_w_fondo_transparent 2

Myths and Challenges in Knowledge Extraction and Big Data Analysis

For centuries, science (in German “Wissenschaft”) has aimed to create (“schaften”) new knowledge (“Wissen”) from the observation of physical phenomena, their modelling, and empirical validation.

Recently, a new source of knowledge has emerged: not (only) the physical world any more, but the virtual world, namely the Web with its ever-growing stream of data materialized in the form of social network chattering, content produced on demand by crowds of people, messages exchanged among interlinked devices in the Internet of Things. The knowledge we may find there can be dispersed, informal, contradicting, unsubstantiated and ephemeral today, while already tomorrow it may be commonly accepted.

Picture2The challenge is once again to capture and create consolidated knowledge that is new, has not been formalized yet in existing knowledge bases, and is buried inside a big, moving target (the live stream of online data).

The myth is that existing tools (spanning fields like semantic web, machine learning, statistics, NLP, and so on) suffice to the objective. While this may still be far from true, some existing approaches are actually addressing the problem and provide preliminary insights into the possibilities that successful attempts may lead to.

I gave a few keynote speeches on this matter (at ICEIS, KDWEB,…), and I also use this argument as a motivating class in academic courses for letting students understand how crucial is to focus on the problems related to big data modeling and analysis. The talk, reported in the slides below, explores through real industrial use cases, the mixed realistic-utopian domain of data analysis and knowledge extraction and reports on some tools and cases where digital and physical world have brought together for better understanding our society.

The presentation is available on SlideShare and are reported here below:

Urban Data Science Bootcamp

We organize a crash-course on how the science of urban data can be applied to solve metropolitan issues.

crowdinsights_bootcamp_2017_en

The course is a 2 days face-to-face event with teaching sessions, workshops, case study discussions and hands-on activities for non-IT professionals in the field of city management. It is issued in two editions along the year:

  • in Milan, Italy, on  November 8th-9th, 2017
  • in Amsterdam, The Netherlands, on November 30th-December 1st, 2017.

You can download the flyer and program of the Urban datascience bootcamp 2017.

Ideal participants include: Civil servants, Professionals, Students, Urban planners, and managers of city utilities and services. No previous experience in data science or computer science is required. Attendees should have experience in areas such as economic affairs, urban development, management support, strategy & innovation, health & care, public order & safety.

Data is the catalyst needed to make the smart city vision a reality in a transparent and evidence-based (i.e. data-driven) manner. The skills required for data-driven urban analysis and design activities are diverse, and range from data collection (field work, crowdsensing, physical sensor processing, etc.); data processing by employing established big data technology frameworks; data exploration to find patterns and outliers in spatio-temporal data streams; and data visualization conveying the right information in the right manner.

The CrowdInsights professional school “Urban Data Science Bootcamp” provides a no-frills, hands-on introduction to the science of urban data; from data creation, to data analysis, data visualization and sense-making, the bootcamp will introduce more than 10 real-world application uses cases that exemplifies how urban data can be applied to solve metropolitan issues. Attendees will explore the challenges and opportunities that come from the adoption of novel types of urban data source, including social media, mobile phone data, IoT networks, etc.

Analysis of user behaviour and social media content for art and culture events

In our most recent study, we analysed the user behaviour and profile, as well as the textual and visual content posted on social media for art and culture events.

The corresponding paper has been presented at CD-MAKE 2017 in Reggio Calabria on August 31st, 2017.

Nowadays people share everything on online social networks, from daily life stories to the latest local and global news and events. In our paper, we address the specific problem of user behavioural profiling in the context of cultural and artistic events.

We propose a specific analysis pipeline that aims at examining the profile of online users, based on the textual content they published online. The pipeline covers the following aspects: data extraction and enrichment, topic modeling based on LDA, dimensionality reduction, user clustering, prediction of interest, content analysis including profiling of images and subjects.

Picture1We show our approach at work for the monitoring of participation to a large-scale artistic installation that collected more than 1.5 million visitors in just two weeks (namely The Floating Piers, by Christo and Jeanne-Claude). In the paper we report our findings and discuss the pros and cons of the work.

The full paper is published by Springer in the LNCS series in volume 10410, pages 219-236.

The slides used for the presentation are available on SlideShare:

 

Urbanscope: Digital Whispers from the Urban Landscape. TedX Talk Video

Together with the Urbanscope team, we gave a TedX talk on the topics and results of the project here at Politecnico di Milano. The talk was actually given by our junior researchers, as we wanted it to be a choral performance as opposed to the typical one-man show.

The message is that cities are not mere physical and organizational devices only: they are informational landscapes where places are shaped more by the streams of data and less by the traditional physical evidences. We devise tools and analysis for understanding these streams and the phenomena they represent, in order to understand better our cities.

Two layers coexist: a thick and dynamic layer of digital traces – the informational membrane – grows everyday on top of the material layer of the territory, the buildings and the infrastructures. The observation, the analysis and the representation of these two layers combined provides valuable insights on how the city is used and lived.

You can now find the video of the talk on the official TedX YouTube channel:

Urbanscope is a research laboratory where collection, organization, analysis, and visualization of cross domain geo-referenced data are experimented.
The research team is based at Politecnico di Milano and encompasses researchers with competencies in Computing Engineering, Communication and Information Design, Management Engineering, and Mathematics.

The aim of Urbanscope is to systematically produce compelling views on urban systems to foster understanding and decision making. Views are like new lenses of a macroscope: they are designed to support the recognition of specific patterns thus enabling new perspectives.

If you enjoyed the show, you can explore our beta application at:

http://www.urbanscope.polimi.it

and discover the other data science activities we are conducting at the Data Science Lab of Politecnico, DEIB.

 

Extracting Emerging Knowledge from Social Media

Today I presented our full paper titled “Extracting Emerging Knowledge from Social Media” at the WWW 2017 conference.

The work is based on a rather obvious assumption, i.e., that knowledge in the world continuously evolves, and ontologies are largely incomplete for what concerns low-frequency data, belonging to the so-called long tail.

Socially produced content is an excellent source for discovering emerging knowledge: it is huge, and immediately reflects the relevant changes which hide emerging entities.

In the paper we propose a method and a tool for discovering emerging entities by extracting them from social media.

Once instrumented by experts through very simple initialization, the method is capable of finding emerging entities; we propose a mixed syntactic + semantic method. The method uses seeds, i.e. prototypes of emerging entities provided by experts, for generating candidates; then, it associates candidates to feature vectors, built by using terms occurring in their social content, and then ranks the candidates by using their distance from the centroid of seeds, returning the top candidates as result.

The method can be continuously or periodically iterated, using the results as new seeds.

The PDF of the full paper presented at WWW 2017 is available online (open access with Creative Common license).

You can also check out the slides of my presentation on Slideshare.

A version of the tool is available online for free use, thanks also to our partners Dandelion API and Microsoft Azure. The most recent version of the tool is available on GitHub here.