​The Ministry of Foreign Affairs wants to acquire an IT system based on artificial intelligence (AI) to monitor online content in real time from social media platforms, news sites, blogs, as well as the accounts of the most influential people – journalists, celebrities and dignitaries. The goal is to identify, analyze and combat disinformation and propaganda.

The Ministry of Foreign Affairs (MAE), which until recently was headed by the liberal Bohdan Aurescu, announced on Saturday, June 10, an open tender for the procurement of an IT system based on advanced artificial intelligence algorithms to analyze and verify online content to determine its origin, with the aim of identifying, analysis, identification and fight against disinformation from the online environment in the areas of responsibility of the IAEA.

The contract with an estimated value of more than 3.2 million lei (excluding VAT) is financed by EU funds and will be implemented within 4 months, according to data from the electronic system of public procurement (SEAP).

The contract also provides for the purchase of training services related to the training of the company’s own personnel in the use of the developed IT tool.

500 GB of data per month will be collected: what online content will be monitored in real time

The tender specification states that the process of collecting data from sources such as Facebook, Twitter etc. Data collection modules must provide a minimum of 500 GB of data collected per month at no additional cost to the MAE.

• SEE THE ASSIGNMENT HERE

The IT system will include, among other things, a COTS or open source online data collection module for real-time monitoring of online content (social media platforms, news sites, blogs, etc.) with the following capabilities:

• geographic segmentation – the ability to separate data/references by country of origin;
• quick discovery – articles, mentions or posts should be discovered within a maximum of 30 minutes;
• interaction monitoring – to the extent that the source systems allow it, the system should be able to collect statistical data about the interaction of users with the monitored content: the number of views, likes, shares, etc. (the system should include the ability to schedule data updates at intervals to be able to cover the entire life of the article or publication);
• receive publicly available data from social platforms, at least: Facebook, Twitter, Instagram, a minimum of posts and comments for each social network, with or without the need for authentication;
• allow extraction of metadata related to the web page: date, author, keywords, description, title; (…)”.

The IT tool will be a web portal based solution where designated MAE representatives will have access to the necessary data and information.

The artificial intelligence system will track influencers and news sites, as well as the emotional state created by the articles

Here are some of the services that a future IT system based on artificial intelligence should provide:

Customizable dashboard (without graphical administration interface):

• Configurable dashboard in the form of widgets (drag and drop) to be able to compare different topics, influencers, keywords and more.
• Feed – the user can view and manage all mentions, share them with other users of the system, add notes or tags. It can also archive/classify information in personal files. (…)

View data:

• Automatic determination of semantic similarity between analyzed texts, selection of hyperlinks and links from online articles.

• These links will be used to build a meaningful graph of semantic connections that reveals how information moves across different online environments, tracing it back to its point of origin (original source).

Data filtering

• The system should allow filtering of stored data by topic/category, keywords, media channel, usernames, time periods (today, yesterday, last week/month, etc.), sentiment, language, country, tags, etc.

Analysis of emotional states caused by the article

• The system will automatically assign the type of emotional state caused by the article (positive, negative, neutral) in at least Romanian and English languages, and allow their analysis according to media channel, trends, topics, etc. (..)
• At the same time, the participant will have to provide clear evidence of the existence at the time of bidding of the module for the analysis of the emotional state caused by the article, and for a minimum of Spanish, Italian, Russian, French, German, Portuguese languages. , Hungarian, which MAE will be able to purchase after the completion and production of the IT tool.

Analyze influencers and profiles

Discover the most active/influential authors, most popular sources and online communities.

Create an influencer profile:

• Name and Username
• Number of posts on the topic
• Number of articles ordered by the state caused by the article (positive, negative, or neutral) associated with the posts
• Number of subscribers

Influencer accounts from different platforms will be merged into a single influencer profile (e.g. profiles of journalists, celebrities, dignitaries). Unification will be based on common attributes or the degree of similarity of profile images.

List of top 10 influencers for various topics/categories.

The most popular authors

• The purpose of this section is to enable users to identify the most popular profiles that mention controlled keywords and phrases.
• An influential author with a significant social reach clearly has a huge audience with massive social media access and simply reaches more people who may be interesting and worthy of attention.
• The section will include a list of the top 50 authors ranked by influence and social reach (with the option to view the entire list) (…)

Summary of the most important indicators

The purpose of this section is to allow users to see a summary of the most important values ​​(volume of posts over time, type of states induced, users, etc.) related to the project/keywords being tracked.

The summary will include the following indicators:

• Number of mentions – the total number of mentions/topics detected in the selected time interval:
• Social Media Mentions – Total number of mentions/discussions detected on social media platforms for the selected time period
• Non-Social Media Mentions – The total number of mentions/discussions found outside of social media networks (eg forums, blogs, etc.).
• Social media reach is the estimated number of people who may have come into contact with relevant social media mentions containing the monitored phrase(s). Estimated social media reach is based on the number of authors who talk about controlled phrases on social networks, the number of their followers/followers/friends
• Non-Social Media Reach – Estimated number of impressions with relevant mentions (containing controlled keywords)
• You should take into account the number of domains in which the controlled phrases are mentioned and the visibility ratio of these mentions (percentage of the controlled keywords appearing in a certain domain: URL, title, content, comments, etc.).
• Social Media Interactions – The number of interactions generated containing tracked words/phrases on social media platforms.
• User Generated Content – ​​The number of mentions generated by users of social media platforms, forums, blogs, etc.
• Likes on social networks – the number of likes for posts that contain controlled keywords.
• Number of videos – the number of video mentions found on social networks.
• Positive mentions – the number of mentions of tracked keywords with a potentially positive sentiment (numerical value and percentage of total
• Negative mentions – number of mentions of tracked keywords with a potentially negative sentiment
• Neutral mentions

Top public profiles

• Profile Name | platform | coverage percentage | influence
• Reach percentage – what percentage of the entire discussion was created by the selected author (user, profile, page, etc.)
• Impact – the estimated number of views generated by the selected author containing the monitored phrases. Influence is based on number of mentions, number of followers/followers/friends
• The section should also link to this profile.
• Ability to remove the author from the monitoring list (if it does not meet the purpose of the respective project) to ensure a more accurate analysis.

The most active profiles

• Profile Name | platform | audience | CLAIMS
• Audience – The number of users to whom a post containing monitored phrases has been shared.
• Mentions – The number of profile mentions that contain tracking phrases
• The section should also link to this profile.
• Ability to remove the author from the monitoring list (if it does not meet the purpose of the respective project) to ensure a more accurate analysis.

The most active sites

A list of the most active sites containing monitoring phrases

– Name of the site

– Number of mentions

You can display the first 10 sites, with the option to open the full list.

The most influential sites

A list of the most influential domains containing monitored phrases

– Domain name

– Impact assessment

– Estimated number of visitors

metrics

The system should contain the following indicators:

• – Feelings count
• – Evaluation of the influencer
• – Volume of posts, articles, distributions for the last hour, day, month, etc.
• – Average number of messages per minute/hour for specific topics/keywords
• – Most shared hashtags and links
• – Users who are mentioned most often in posts
• – Post/share ratio
• – Growth of evolution in a certain period
• – Viralization – based on links, shares and comments

The deadline for receiving bids or applications for participation in this auction is July 17, 2023. A bid opening meeting will also be held on the same day.

Photo source: Melpomenem / Dreamstime.com