An overview of the Climate Finance Tracker, an initiative launched with OneEarth. Each circle or node on the map represents one private company in the U.S. Users can sort by themes, sectors, investors, and more.

The emerging threat of climate change to our national security

From as early as 2008, the U.S. Department of Defense flagged the increased risk to national security if climate change wasn’t urgently addressed by the federal government.

“We judge global climate change will have wide-ranging implications for U.S. national security interests over the next 20 years,” wrote the National Intelligence Committee during President George W. Bush’s tenure.  The political will to take action to address climate change as a national security issue is slowly growing. Recently the U.S. Army released its first ever climate strategy, outlining its goals to cut emissions in half by 2030, produce electric combat vehicles, and advance officer training for a warmer climate. 

At Primer, we build AI technology that helps subject matter experts and decision-makers identify, understand, and act on emerging threats. We believe that the intersection between climate change and national security is imperative to proactively address. 

“Climate change will affect the Department of Defense’s ability to defend the nation and poses immediate risks to U.S. national security.”

-U.S. Department of Defense, 2014 Climate Change Adaption Roadmap

“Increasing migration and urbanization of populations are also further straining the capacities of governments around the world and are likely to result in further fracturing of societies, potentially creating breeding grounds for radicalization.”

-U.S. National Intelligence Strategy, 2019

Our partnership with OneEarth.org

On the cusp of Climate Week in New York starting Sept. 19, Primer is proud to announce our partnership with OneEarth.org, and the public release of the Climate Finance Tracker that we worked on together this year. Leveraging Primer’s machine learning technology on publicly available data, the tracker “follows the money,” surfacing in one unified view, where U.S. private funding for combating climate change is going. In short, following the money helps everyone collaborate more effectively to tackle the climate crisis. 

For the first time, the Climate Finance Tracker is live and open-source, and uses AI to advance intelligence and decision cycles, to ultimately improve outcomes on solving a major global challenge.  

The Climate Finance Tracker allows anyone to search the funding landscape of all private companies and nonprofits in the U.S. that are addressing climate-related topics. By surfacing topics from organization and project descriptions provided by Crunchbase and Candid, the tracker allows us to better understand the shape of the current response, and with that, its potential gaps. This understanding means funders can better direct their resources, ensuring the highest possible impact of their investment.

The value of following the funding

Eric Berlow, a leading social impact data scientist and researcher, led the project with his nonprofit Vibrant Data Labs and OneEarth.org. Berlow explained how using Primer’s AI technology allows researchers to find the answer to questions they may not even know to ask. 

“We searched for companies working on known climate-related topics, and then the machine pulled out of their language that some also mention national security-related topics,” Berlow explained. “That’s novel. I wouldn’t have even known to look… And the machine found it.”

Text-based machine learning doesn’t just uncover answers – it also helps surface questions researchers may not even know to ask.

A snapshot of 32 organizations that were tagged with ’national security’, ‘unrest, conflicts and war’, or ‘peace and security.’ Explore the full Climate Finance Tracker here.

“We know climate intersects with national security, and now we can easily see about 25 private investors funding companies in that space,” Berlow explained. “It’s only 6 companies out of 6,000. That’s an important gap. Let’s talk to them and see what’s going on.”

Climate week begins

With this week’s launch of the Climate Finance Tracker, along with Climate Week in New York, we expect to see increased conversation around adoption of machine learning solutions to help solve global problems such as the climate crisis, making us more resilient toward emerging threats.

At Primer, we believe that when we empower the ecosystem with better information, intelligence, and decision making, we can truly solve really complex problems together. 

See the power of text-based AI and NLP. 

Request a demo of Primer’s NLP technology. 

AI is rapidly changing how wars are fought. Winners of the AI arms race will become the world’s dominant military powers. Here’s what’s at stake.

Primer CEO Sean Gourley testified at the U.S. Chamber of Commerce AI Commission on July 21, 2022, about the global AI arms race — and what the U.S. and our allies need to do next. Below are his testimony notes. 

“The stakes are high, and the winner of this AI arms race will become the dominant military power in the world. It is a race that we cannot afford to lose.”

-Primer CEO Sean Gourley

Testimony by Primer CEO Sean Gourley 

Thank you Congressmen Ferguson and Delaney, members of the Commission. 

While there is a huge amount of discussion about the impact of AI on our society, the biggest impact that Artificial Intelligence (AI) will have in the next decade will be in warfare, where advanced AI will fundamentally change the way wars are fought. The impact of AI on warfare will be akin to that of nuclear weapons, where AI is a technology so powerful that the country that wields it will quickly defeat any opponent who does not.  

Artificial intelligence represents what is known as the “third offset,” a set of technological capabilities so advanced that it gives whoever wields them an advantage so large that the opponent without the technology is effectively defeated before a conflict even starts. 

“AI is a technology so powerful that the country that wields it will quickly defeat any opponent who does not.”

-Primer CEO Sean Gourley

The first offset was nuclear weapons, which ended the Second World War. The second offset was stealth weaponry and precision munitions, which resulted in the U.S. defeating the Iraq army — which at the time was the world’s 6th most powerful army — in less than 72 hours during the First Gulf War. The third offset is artificial intelligence. And it will have as profound an impact on warfare as any of the previous two offsets combined. 

This is not a hypothetical discussion about something that might happen in the future. Today, we are already seeing the impact of AI on the battlefield in Ukraine. From computer vision being used with commercial drones to identify camouflaged Russian vehicles, to AI that listens to radio communication and triangulates these with videos from social media to track Russian troop movement and intention in real time, through to AI being deployed in the information war to attempt to manipulate the narrative and win over the enemy population. 

These changes are happening rapidly. But Russia and Ukraine are not widely regarded as AI superpowers – it is China that we need to turn our attention to. 

We need to acknowledge that we are currently in an AI arms race with China. The stakes are high, and the winner of this AI arms race will become the dominant military power in the world. It is a race that we cannot afford to lose. 

The United States and its allies come into this arms race with a considerable set of advantages, but speed is going to determine the winner here — and China is moving fast.

To read more, download the full testimony here.

“As everyone is aware, misinformation and disinformation are being sown by many of our competitors, and the problem is only growing. We have to be able to see that in real time. But we also have to be able to counter with all elements of statecraft.” – General Richard D. Clarke, Commander of U.S. Special Operations Command, April 2022 

Primer has been deploying core Natural Language Processing (NLP) infrastructure and Engines to help people detect, understand, and respond to disinformation campaigns. This has been a critical aspect of Primer’s work with the U.S. Air Force and U.S. Special Operations Command (SOCOM) and drives us to bring the best Artificial Intelligence (AI) tools to operators who make mission-critical decisions. We are committed to accelerating our work in information operations and today we are excited to announce our acquisition of Yonder, a pioneering company in disinformation analysis. 

Disinformation is changing the course of the world

From the war in Ukraine to market manipulation of cryptocurrency, disinformation changes the course of conflict, impacts elections, degrades brands, and distorts discussions. These narratives develop across the Internet at a pace and scale that humans simply can’t detect or keep up with. A single narrative, even if it’s false or misleading, may impact an organization’s competitive or strategic position in the world – seemingly overnight. Early intelligence about emerging narratives is critical, so organizations can be proactive and avoid an expensive and exhausting game of whack-a-mole. 

But detecting and mitigating the impact of influence operations is hard. Organizations have a small window of time to address narratives before they go viral. Getting in front of these messages is essential, because it can have lasting effects on perceptions, beliefs, and behavior. We saw this play out in the war in Ukraine when there were false messages circulating about the U.S. operating a chemical weapons facility in Ukraine. By detecting and proactively refuting these claims, the U.S. government got in front of the narrative before opinions were formed and solidified. 

Yet even if an organization detects the emerging narrative in time, understanding the risk is often less straightforward. Questions need to be answered. Which groups started or are promoting the narrative, what is their agenda, who are they connected with, how influential are they, what will they do next … is this a momentary blip or something more organized and persistent?

These are exactly the types of questions that Yonder helps customers answer – the who, what, where, why, and how of a disinformation campaign. Fortune 1000 companies, including the world’s largest retailer, a fast food brand, and telco company, use Yonder to proactively monitor and take action to avoid situations that could harm their brands and customers. Yonder provides contextual intelligence about the narrative and factions promulgating or amplifying it. Users can configure their settings in an easy-to-use User Interface (UI) and receive alerts via email if something in their world is changing. Yonder is able to understand evolving narratives and, importantly, does so across multiple languages.

A robust and scalable NLP information operations suite 

Detecting information manipulation operations is mission-critical for the US Department of Defense, Intelligence Community, and strategic international partner organizations. Identifying influence campaigns is increasingly challenging because of advances in synthetic text generation. These technological advancements are making it largely impossible for most tools to detect synthetic text because they lack the predictable “tells.” But this is Primer’s strength. Our Primer NLP Engines automatically detect content that has been algorithmically amplified by bots or even automatically generated by advanced AI programs. 

Recognizing our customers’ diverse needs for understanding fast-breaking events in real time, we launched Primer Command. With Command, organizations can easily monitor and cut through noisy and high-volume social and news media, so they can understand risk and take action. Our advanced NLP algorithms work around the clock, at human-level accuracy and machine scale and speed, to automatically detect and display the origin and source of disputed information – all accessible within a single and intuitive interface. 

By pairing Yonder’s contextual narrative intelligence capabilities with Primer Command and our state-of-the-art NLP/ML models, we can provide our customers with advanced early warning about emerging narratives and agenda-based online networks associated with propaganda. Our combined capabilities also allow us to draw out the factions pushing disinformation on social media for our customers. 

The need for tracking early signals has become increasingly critical as influence operations have become a key element of modern warfare. It was a tactic Russia used before they invaded Ukraine, portraying Ukraine as an aggressor preparing to attack Russia-backed forces on the eastern border. They also began promoting a narrative that the US is abandoning Ukraine by drawing comparisons to the US’s withdrawal from Afghanistan. Spreading disinformation has been a part of the Russian military arsenal ever since. 

What the war in Ukraine has shown is that it is no longer enough for the U.S. military to track intelligence about adversaries’ actions. The same caliber of signals is needed today about competitor influence operations. Having these signals at scale is exactly what USSOCOM Commander General Clarke mentioned as part of his SOFIC keynote in May 2022. He suggested that the US military needs to have sentiment analysis capabilities similar to how major companies track influence operations against their brand. That there is a requirement for the U.S. military to generate alerts warning about information manipulation campaigns so they can respond to it. 

Creating a “monitor” by combining brand terms, sources, exclusionary terms, and geofencing in Primer Command.

In Primer Command, users can get a quick snapshot of the sentiment around their brands.

What’s next

We believe that government and civil society organizations need to have the tools to be successful in their missions. Analysts and operators need early warning capabilities and contextual information from a multitude of sources to help them inform countermeasures and make high-stakes decisions. Every Fortune 1000 company should have the capabilities in place such that when subjected to an information attack, they will be alerted and have the ability to respond quickly to reduce the impact. 

Effective today, Yonder customers can expect the same great insights and services they are used to, boosted by Primer’s resources, infrastructure, and expertise. They will also have access to Primer’s world-class pre-trained NLP models, including more advanced bot detection algorithms, synthetic text detection capabilities, and claim detection and extraction algorithms, to further enhance their ability to identify and understand emerging risks. 

Primer customers who are using Command, our NLP infrastructure, and our suite of pre-trained models to track and understand fast-moving crisis situations will have the ability to integrate Yonder into their workflows. They can get a high-resolution picture of the disinformation space they are operating within, easily understand what information they should trust, and identify what is likely to be an information operations campaign.

Primer offers a warm welcome to the Yonder team. These are the people that have defined and shaped the disinformation detection space and we are thrilled to work with them to accelerate our work on combating this growing and critical challenge.

For more information about Primer and how to request a Yonder demo, contact us here.

BabyBear cuts GPU time and money
on large transformer models

When you’re processing millions of documents with dozens of deep learning models, things add up fast. There’s the environmental cost of electricity to run those hungry models. There’s the latency cost as your customers wait for results. And of course there’s the bottom line: the immense computational cost of the GPU machines on premises or rented in the cloud. 

We figured out a trick here at Primer that cuts those costs way down. We’re sharing it now (paper, code) for others to use. It is an algorithmic framework for natural language processing (NLP) that we call BabyBear. For most deep learning NLP tasks, it reduces GPU costs by a third to a half. And for some tasks, the savings are over 90%. 

The basic idea is simple: The best way to make a deep learning model cheaper is to not use it at all, when you don’t need it. The trick is figuring out when you don’t need it. And that’s what BabyBear does. 

In our algorithm, the expensive deep learning model is called mamabear. As the documents stream in to be processed by mamabear, we place another model upstream: babybear. This model is smaller, faster, and cheaper than mamabear, but also less accurate. For example, babybear might be a classical machine learning model like XGBoost or Random Forest. 

The babybear model can be whatever you want—as long as it produces confidence scores along with its predictions. We use the confidence of babybear to determine whether an incoming document requires mamabear. If it’s an easy one, babybear handles it and let’s mamabear sleep. If it’s a difficult one, babybear passes it to mamabear.

How does babybear learn this task? During a “warm up” period, we send the stream of documents to mamabear, just as we would in production without the BabyBear framework. The babybear model directly observes mamabear as an oracle, making its own predictions to learn the skill. Once it has sufficient skill, it gets to work. 

For example, we took this open source sentiment analysis model as a mamabear. For the babybear warm up we trained an XGBoost model using 10,000 inferences from mamabear as gold data. With hyperparameter optimization it took 113 minutes. But the babybear had already learned the task sufficiently to get to work after about 2000 examples—less than half an hour of training.

Every NLP practitioner has created keyword filters upstream of expensive models to save unnecessary processing. BabyBear just replaces that human data science work with machine learning.

All that you need to do is tweak one parameter: performance threshold. That determines how much loss in overall f1 score you’re willing to pay in return for computational savings. Whatever you set it to—10%, 5%, or 0%—BabyBear will save as much compute as possible by adjusting its own confidence threshold to meet that target.

What I’ve described so far is the simplest version of BabyBear. It applies to document classification, for example, where babybear is a single model upstream of mamabear and is learning to perform the same task. In our paper, we describe more complicated versions of BabyBear. For NLP tasks such as entity recognition, we achieve greater savings using a circuit of multiple babybear models. And some of those babybear models can be distilled (cheaper, faster) versions of mamabear. But the same algorithm applies. 

BabyBear is the framework that we use here at Primer to automate some of the data science we once laboriously did ourselves. For the vast majority of NLP tasks—basically everything not requiring text generation—BabyBear can help prevent needless draining of your wallet and the electrical grid.

Try it out on your data.

Alliance Trucking is a regional trucking company that’s been serving the Dallas-Fort Worth area for over 30 years. Alliance acts as a broker between truck owner-operators and construction companies who need loads of construction material delivered. 

To request pricing for materials, Alliance’s customers email requests for quotes (RFQs) describing the location, the materials that need to be transported, and other details pertinent to the job. Alliance’s team then analyzes the emails to estimate trucking routes, availability, and costs and return a quote to the customer. 

Running into friction scaling their core operations 

Producing accurate quotes keeps Alliance competitive while ensuring a sound operating margin. But estimating the cost of a delivery is a complex human process that represents a lot of overhead. An estimator receives a bid in their help desk software, HelpScout. The bids come in as unstructured text with the details, including a schema of the location, material, and time frame. Estimators take that information, do the route research, and then manually input the job into Alliance’s purpose-built software. 

This manual process created variation in their pricing, inefficient routes, and constrained the number of RFQs Alliance could respond to. Alliance wanted to scale their operations by offloading the route estimation to machine learning. 

Using LightTag to build a dataset

Alliance needed to be able to feed emails directly into an NLP model that could understand RFQs and produce precise estimation criteria. The estimates would then be loaded into their software to be approved and executed. To build an NLP model for the trucking industry, they first needed to build a high quality dataset of job estimates from which their custom model could learn. 

To ensure data quality, the labeling job fell to their CFO, Eric Dance. To maximize the return on investment he invested in labeling data, Alliance set out to find a solution that would make labeling simple and efficient and selected LightTag.

Choosing the right data labeling solution

LightTag learned from Eric as he annotated and then provided pre-annotations that automated a large fraction of Eric’s work. An intuitive user interface made the software easy for Eric to use and build a dataset he was happy with, despite the fact that he is not a technical user.

“I’m not a developer and I don’t have those skills, but I understand my business and wanted to digitize it. I found LightTag easy to use to build a dataset of thousands of emails so we could get a quality dataset.” 

-Eric Dance, CFO, Alliance Trucking

Results

Eric was able to build a dataset with tens of thousands of annotations and train a precise RFQ estimation model. From there, Alliance automated the ingestion of bid emails and responses to RFQs in a faster, more consistent manner.  Today, RFQs can be created and approved by a truck dispatcher in seconds, rather than minutes. 

Alliance is a great example of how NLP can be applied not just to cutting-edge use cases, but to any business process. And it’s a shining example of how LightTag can dramatically improve the labeling process for subject matter experts to spend as little time labeling as possible, while still building a quality dataset to train a machine for human-quality results. 

To learn more about LightTag, Primer or request a demo, contact us here.

Since the development of the Transformer model, finding and re-labeling mislabeled examples is becoming a much more important factor in model performance. Because Transformer models mean ML engineers are finding that they need less and less data to create a model that performs at production quality for their task, the impact of mislabeled data has a greater and greater impact on model quality.

As an example of how this plays out: for a simple classification task, a Transformer-based model like XLNet may only require a couple hundred examples in order to perform well. When your dataset only has 200 examples, and 10% of them are mislabeled, those mislabeled examples can have a big impact, dragging performance down and making it harder to improve the model even when more examples are added. As a result, nowadays it’s more important for model creators to focus on data quality than data quantity

This is because Transformers can be “pre-trained” on natural language (such as random internet text) learning the rules of language simply by reading lots of documents. They can then be “fine-tuned” to perform a specific task, such as classification or named-entity recognition. Since we can start with a pre-trained model, which has already learned from millions of documents, fine-tuning for a more specific use case requires orders of magnitude less data. As an aside, speeding up model development is the rationale behind Primer’s pre-trained, domain specific models

Let’s say you’re doing a classification between two classes, Class A and B. When your model sees an example that clearly looks like Class B, and instead it’s labeled Class A, it now has to adjust its idea of what Class A looks like to accommodate this weird Class-B-like example, which is likely an outlier relative to the other Class A data points. If this happens too many times, it can distort the model’s overall understanding of the task, causing it to underperform on both Class A and Class B.

Data Map simplifies identifying mislabeled data

Data Map is a Primer feature that makes it easy to visualize the outliers so that data scientists and labelers can quickly and easily spot and correct mislabeled examples.

Most data science teams are leveraging some form of external help with annotating examples, as businesses generally want to minimize the time data scientists or business subject matter experts spend time on data labeling. As a result, the annotators’ understanding of the task often differs from that of the SMEs, and examples can get labeled incorrectly as a result. 

We built Data Map to empower data scientists to find these mislabeled examples in an automated way. On top of that, Data Map has the added benefit of giving us additional information about how the model perceived the examples in the dataset during training. (We’ll get into that during the technical explanation of Data Map.)

Our internal experiments using Data Map to find errors showed that, for small datasets, Data Map can identify up to 80% of the errors in the data by examining only 5% of the dataset. 

More directly, instead of re-reviewing the entire labeled dataset, data scientists can simply look at the 5% “most likely mislabeled” examples and re-label those, and rest assured that the remainder of the dataset is likely correct.

How does Data Map work?

Data Map is based on the paper Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics. The idea is to track three relevant metrics for each example as the model is being trained on those examples. 

The metrics are:

  • Correctness: How often the model got this example correct during training; or, more technically, the proportion of training epochs during which the model predicted the example correctly.
  • Confidence: The model’s average level of confidence when making predictions for this example during training.
  • Variability: How often the model changed its mind about the example’s predicted label during training. An example that oscillates between being classified as Class A and Class B repeatedly will have high variability; an example consistently classified as Class A will have low variability.

Each metric falls in the range [0, 1]. Usefully, these metrics also allow us to group data points into descriptive categories: easy-to-learn, ambiguous, and hard-to-learn. 

An easy-to-learn example will typically have high correctness, high confidence, and low variability.
An ambiguous example will typically have high variability, because the model was indecisive about where to put it. There won’t necessarily be a trend with respect to correctness or confidence.

A hard-to-learn example will most importantly have low correctness. There won’t necessarily be a trend with respect to variability or confidence.

These three metrics, taken together, give us a better picture of how challenging the model found those examples over the course of the training. 

The authors of the original paper found many interesting correlations with respect to the role that each group of data points plays in helping a model learn a task, which we would encourage you to check out if you’re interested. Here, we will focus on the hard-to-learn group.

Hard-to-learn examples are the examples we single out as being “possibly mislabeled.” The authors of the original paper found that hard-to-learn examples were frequently mislabeled, and our internal research has replicated this finding. Hard-to-learn examples which also have high confidence and low variability are particularly likely to be mislabeled (rather than just challenging), because it means the model was very confident about its answer, but still got the example consistently wrong.

Technical implementation

First, a research engineer on our Applied Research team had to implement the tracking of Data Map metrics into our internal deep learning library’s classifier training module. This required digging into the training loop for the model and inserting additional logic to store data about the relevant concepts after each epoch, then calculate them at the end of the training session.

Then, we had to spec out an ideal way to track and store the Data Map information our Postgres database. This required the creation of a dedicated “cartography” table, and updates to business logic throughout the application to introduce the concept of “possibly mislabeled” documents, as well as to keep track of whether or not users have “reviewed” their mislabeled documents. This was important because we didn’t want to constantly resurface the same documents to users if they had already addressed them. 

Part of the data modeling work was to implement additional functionality into the deep learning module to track training examples by ID from beginning to end, so we could link them back to their original document IDs – something we hadn’t needed to worry about up to this point.

Next, our ML platform’s training service needed to make the requisite updates to pass the Data Map information through and save it to S3 whenever Automate sends it a new training job. We also needed to update the number of epochs we were using to give us more granular values from the data map calculations (Particularly for correctness – it wouldn’t be ideal if an example could only ever have .33, .66, or 1.0 correctness.).

Finally, our design and frontend team created an informative, user-friendly interface to communicate this information, and provide multiple entry points from Data Map to our document labeling experience. Customers can see Data Map results for individual documents while labeling, filter their list of labeled documents by whether they’re mislabeled, and can preview and jump directly to a mislabeled document by clicking on its data point in the graph. 

Results

Data Map is a key utility for quality assurance of datasets in Primer, saving customers time by allowing them to review a subset of examples for errors rather than the entire dataset. Techniques like Data Map are particularly powerful because they help customers to get more performance out of existing data, rather than label new data and are an obvious example of Primer’s focus on data efficiency. We’re determined to make time-to-value short so customers can hit the ground running on their projects with a performant model.

For more information about Primer and to access product demos, contact Primer here.

From a security perspective, the Beijing Olympics has all the ingredients of a perfect storm. Global tensions are ratcheted up as the US and Russia negotiations about Ukraine’s fate hit a wall. The US and its allies have declared a diplomatic boycott of the Beijing Olympics due to alleged human rights violations by the Chinese government against the Uyghur population. These same dissidents are looking at the games as a platform for voicing their grievances. 

Adding to the tensions are reports that movements of people in the city, including players and media, are limited under COVID-19 protections.  The games themselves are expected to be more opaque than years prior as Beijing has limited foreign correspondents to a “closed loop” bubble. The correspondents are provided with limited interactions with athletes taking part in the games and no movement in the city itself. The athletes are under extra security protocols as officials warn of Chinese surveillance operations that will target them while in the country. 

Security analysts from global security operations centers (GSOCs) around the world with assets or people in Beijing during the games are on high alert. They need to monitor incoming information around the clock to anticipate any threats and provide instructions if any security incidents occur.

Surface key insights

For security teams charged with the herculean task of monitoring threats emanating from and against the Beijing Olympics, and its commercial sponsors, Primer Command® is a game changer.  Command not only identifies people and places mentioned, but it also shows live feeds of news and social media posts side by side streaming in. This saves analysts from using multiple apps — it’s all in a single pane. Further, this allows them to leverage news reports to corroborate, in seconds, social media posts with alleged threats and emerging issues of importance. With these capabilities in hand, organizations can maximize the safety of their people, operations, and assets. 


Primer Command automatically structures and summarizes millions of news and social media posts across 100 languages to surface key people, locations, organizations, topics, suspected disinformation, sentiment, and media to better understand how the situation is evolving. 


The following puts a spotlight on the power of Command’s ability to zero in on the information that matters most through advanced filters and AI-enabled classifiers. To learn more about Command’s other capabilities, click here for a free trial of Command for Tactical Insights.

Humanitarian aid filters

Command’s advanced filtering capabilities allow security teams and first responders to unlock mission-critical information during a crisis. Primer’s humanitarian aid filters include drilling down on tweets of displaced people and requests for evacuations, caution and advice, and reports of injuries or deceased people. These filters will be particularly operative during the winter games to zero-in on any violence and safety concerns for personnel there.

Chinese language filters

Learning what is being conveyed to the local population will be more important during the Beijing games given the limited movements allowed for foreign media outlets. Additionally, it will give security teams early indicators of unrest. This filter can also illuminate posts by local nationals – automatically translated – expressing concern about physical security threats.

Additional filters

Command can filter on numerous other entities to drill down into the information security analysts care about most.  

  • Event Types: Security analysts can filter the information feeds by the event types such as diplomacy, military, or law enforcement topics within news reports. This will prove to be of particular importance if any security incidents break-out during the winter games. Analysts will be able to home in on reporting related to law enforcement to get the latest actions to contain the threat. Focusing on these posts also provides GSOCs with the latest official statements and guidance by security forces for the people in the area. 
  • Disinformation: Command can detect and flag disputed information in news reports. Analysts can filter by disputed information and use this as an indicator of disinformation campaigns occurring during the events. 
  • Social Data: Analysts can segment social media data feeds based on the number of likes or retweets, key threat indicators, or even by sentiment. Primer’s state-of-the art sentiment filters are hand tuned for the highest accuracy so analysts can quickly identify the social media posts that matter. By filtering for negative sentiment, analysts can uncover the threats hidden within the deluge of data — separating chatter from hazards.

Learn more

Contact sales to discuss your specific needs. You can also stay connected on Linkedin and Twitter.