Exploring the World’s Largest News Knowledge Graph — Newsplunker
Build, Explore & Analyze the World’s Largest News Knowledge Graph
The AskNews team is excited to announce the release of the Newsplunker dashboard, your custom window into the largest news knowledge graph on the planet.
Newsplunker is a state-of-the-art analyst dashboard designed to let you easily build, explore, and analyze your own custom news topics. This includes building your own knowledge graphs, reading enriched article summaries and tracked news events, generating your own forecasts — using our brand-new forecasting model, Voir — and more, all while benefiting from AskNews’ rich metadata and event-tracking research.
Whether you’re a dashboard user or an API customer on either our Analyst or Enterprise plans, you can start using Newsplunker right away. We’ve also created a new Spelunker plan for those who are doing deep research into multiple news topics and need more access than our Pro plan currently offers but who don’t need the additional add-ons that come with the Analyst plan. (Compare all plans here.)
All of the Information You Need to Explore, Track, and Analyze Any News Topic
The Newsplunker dashboard brings all of Emergent Methods’ state-of-the-art research into news transparency, diversity and narrative tracking together in one highly explorable place. Newsplunker monitors any topic imaginable, including companies, people, wars, investigations, technology trends, politics, finance, sports, entertainment and much more.
In addition, you can create multiple topics and switch between them with just a click. This functionality allows you to track not only different topics, but how a given topic is covered in the global media. For example, you can easily track how many different political parties cover the same news topics, or how the same topic is covered objectively vs. sensationally.
The Newsplunker dashboard breaks down information silos by automatically including news sources from different countries, cultures, and languages. Translation of non-English sources ensures diversity, equity, and competing perspectives as part of your analysis of a given topic or event. This approach — powered by Emergent Methods’ research and our open source models for equitable entity extraction — creates least-biased interactions for analysts who need a variety of perspectives.
Furthermore, human editors working alongside AI (a.k.a., human-in-the-loop or HITL) ensure that the AskNews data powering the Newsplunker dashboard follow journalistic guidelines and minimize LLM hallucinations. Humans review articles and tracked events on a daily basis, injecting journalistic integrity into foundational components of the news information storage and retrieval.
AskNews powers the largest news knowledge graph in the world. The AskNews team uploads 1 million news entities and analyzes 500,000 articles per day. We also curate and update 100+ tracked news events per day, creating a high-level overview report of what’s happened in relation to a particular topic. (If you’re interested in the tech behind AskNews and the Newsplunker dashboard, skip ahead to the behind-the-scenes section below.)
A Top-to-Bottom Tour of the Newsplunker Dashboard
Newsplunker has a lot to offer, so we’ll break it down bit by bit.
Creating & Customizing Your Own Topic
The powerful features of the Newsplunker dashboard give you the ability to explore a news topic through AskNews’ highly enriched data. You can explore example topics, get inspiration from the top news events from the past 7 days (streaming along the top of the dashboard), or you can create your own topic from scratch.
When you create your own topic, provide a name and an optional natural-language description of what you want to track. The more descriptive, the better. Write a whole paragraph if you want!
It’s important to note: If you use a description, the dashboard retrieval is based on a similarity search of what articles have content most similar to your description. This might result in some articles being only tangentially The Overview gives you a condensed rundown of the information related to your given topic, but can also yield a greater diversity of perspectives.
If you want or need absolute precision on what articles are brought into your Newsplunker dashboard, then skip the description and use only the filters (detailed below). This way, the dashboard retrieval is based purely on which articles have metadata matching your given filters.
From there, you set up basic filters for your dashboard, including:
- Number of articles
- Look-back hours (up to 60 days), or a specific Time window (e.g., May-August of 2024)
And filters for the content of the articles:
- Reporting voice: Do you want to look only at articles with objective reporting voice? Or maybe satirical, or emotional? Or a combination? For instance, compare this example Objective Gaza Monitor versus this Sensational Gaza Monitor.
- Sentiment: This allows you to compare the positive and negative reporting on a topic.
You can also set the disambiguation parameters for building the knowledge graph:
- Disambiguation probability: The probability of grouping entities.
- Use geographic information: Whether or not geographic information is relevant to the disambiguation. For example, if you have a topic on international politics, being able to distinguish between the President of France and the President of the U.S. is important and requires geographic information.
A quick note on disambiguation: An entity can be identified in different ways in different articles. For example, articles talking about the British singer-songwriter Freddie Mercury might use “Freddie Mercury,” “Freddie,” “Mercury,” “Mr. Mercury,” or any other disambiguation. Entity disambiguation ensures these entities would not be duplicate nodes in your graph.
In terms of filterability, the settings above are only scratching the surface. The advanced parameters give you even more custom control of which articles and event summaries do (or don’t) show up on your dashboard.
This section lets you filter on just about any factor imaginable. We’ve split it up into filters specifically for the article content, including:
- Major news categories: For example, politics, finance, health, sports, technology, entertainment, etc.
- Provocation level: Filter articles based on the use of provocative language and emotional vocabulary.
- Relevant continents: If you only want your dashboard to include article content that cover Africa and South America, use this filter. Note that the media sources behind the articles might originate in other regions (like Europe or Asia), but the dashboard will only include articles about developments in Africa and South America.
- Guaranteed and excluded strings: Do you need the articles to include a specific name, title, slogan, phrase, or other string? Or do you want to make sure they don’t mention it at all? Use these sections to narrow your search.
- Entities: Major entities and entity types to include in your local news graph, including:
And more filters specifically for the media sources behind the articles:
- Specific news sources: Want your dashboard only tuned in to specific sources like the BBC, Nikkei Asia, Fox News, etc.? Use this section to specify one or multiple sources to obtain articles from.
- Source language and country: These filters only apply to the media sources feeding into your dashboard, not the topic itself. For example, this is where you can restrict your media sources to be only from Germany or to only include media sources written in Arabic, etc.
- Similarity threshold: If you have provided a topic description, this parameter determines how similar the retrieved articles are to that description.
You won’t find a more filterable or customizable news dashboard than Newsplunker anywhere else on the planet. And remember: if you’re an API customer — or if you work with developers with API access — you can easily retrieve your Newsplunker topic through one of our SDKs by clicking the “Code block </>” button.
Once you’re done creating your topic, it’s time to explore the generated graph and dashboard.
Exploring & Using Your Custom News Topic Dashboard
Topic Overview & Summary
The first thing you’ll find on the Newsplunker dashboard is a topic Overview with multiple media source citations. So whether you’re new to a given topic or you’re an in-depth expert, you have a snapshot of the most important aspects and developments. Each statement is cited so that you can follow up information in the original news source.
Scroll down for an overview of the sources cited in your dashboard summary, and click on any one of them to read the full original article.
Explore the Relationship Graph
One of the flagship features of the Newsplunker dashboard is the Relationship Graph.
The Relationship Graph shows the relationships between entities (nodes in the graph) identified in your requested articles. The size of a node represents the number of articles that mention that entity. The color of the node represents the category of the entity (like Countries, Organizations, People, Locations, etc.)
The graph is interactive, so you can zoom in and out, pan across the graph, and hover over nodes to see more information. Clicking on a single node will highlight all the Articles (left of the graph) and Mentions (right of the graph) related to the node, and it will show a popup detailing the relationships between the selected node and other nodes of the graph.
Along the right-hand side of the Relationship Graph, you’ll first see Mentions to show which entities are included in your graph, sorted by entity type, and how many times each entity was mentioned in source articles (increasing or decreasing the size of the node in the graph visualization).
You can click on an entity to focus the graph on the corresponding node and its relationship. You can also filter your local graph by selecting entity types to display — this creates a sub-graph featuring only your selected entity type.
When you select an entity to highlight in the graph (or a node in the graph), related entities (i.e., entities mentioned in the same articles) are indicated with green dots in the Mentions section.
Dig Deep into the News with Articles & Events
Next to the Relationship Graph you’ll find the Articles section serving as your launch point into the global media coverage of your topic.
Each article is designated with the logo of the news source, a sentiment analysis (positive, negative, etc.), and the national flag of the source origin so you can easily scan to determine who is covering your topic and from what angle.
Click on an article or expand the Articles section and select an article to receive an AI-generated summary of the full content, with color-coded highlights to mark mentioned entities. The summary also includes analysis on the sentiment, reporting voice, provocation rating, and source origin of the article so you can evaluate the source quickly before clicking through to read the full article via the link in the footer — or moving onto the next summary.
The geographical overview gives you an unprecedented look at how your topic is being covered throughout the globe:
Scrolling further down, you’ll find the Tracked Events section which features custom editorial from AskNews based on narrative tracking. Each AI-generated event summary is curated by the AskNews team and summarizes a recent incident covered by multiple news sources (never just a single article!).
Similar to individual article summaries, each Tracked Event summary includes media citations and highlighted entities in addition to analysis on sentiment, reporting voice, provocation rating, source origin percentage, overall sentiment movement, and coverage volume over time.
Take your meta-media analysis one level deeper by clicking on the Key Takeaways & Contradictions button. This pop-out summarizes the top 3–5 most important takeaways of a given event (backed by multiple source citations), in addition to summarized analysis of when and how various sources disagreed on important points.
Most Tracked Events include a RedditPerspective where discussions in related Reddit threads have been summarized and give additional context.
We write about 100 new event summaries (or “stories” as we also call them) or updates to events we are tracking per day. This covers global news narratives with a large volume of reporting, but not smaller ones. Depending on the nature of your custom topic, this means that you may see more or fewer Tracked Events in this section of your dashboard and they may be less relevant to your topic.
Evaluate News Sources with Article Analytics
The media articles that feed the Newsplunker dashboard are translated from 16 different languages and represent news outlets across the world. (Explore the diversity of the sources underpinning the AskNews content here.) So which ones are informing your dashboard’s content?
In multiple locations on the dashboard, you can assess the mix of media articles feeding into your custom topic. On the right-hand side under Mentions, you’ll find the Source Origin section. This section displays the percentage of media coverage in your dashboard that’s sourced from different countries, in addition to a map giving you an geographical overview of the quantity of each nation’s coverage.
Just below the Source Origin section, you’ll find the Article Analytics for your dashboard. This section offers four charts that are completely customizable to help you weigh and analyze the media sources feeding into your Newsplunker dashboard.
You can customize each chart according to nine data types and five chart types, including the following:
The result is a wide array of options for quickly analyzing the media coverage that’s included in your Newsplunker dashboard for that particular topic. Want to include the Article Analytics data in an external report you’re working on? Click on the icon in the lower right corner to export the data as a CSV file!
Chat with Your Topic
Just as how you can chat with the news at AskNews.app/chat, you can chat with the news data in your Newsplunker dashboard. If you missed it earlier, hit the “Chat” button in the far top right of your dashboard to interact with your topic directly.
Choose your favorite LLM and then ask your natural language questions directly to your custom graph/topic. Answers always include media source citations, with links to the full articles at the bottom of every answer.
Want to broaden your sources for potential answers? Swap to the Global Chat option and talk directly to the full AskNews database of global news.
Chatting directly with your news lets you easily uncover important information and insights. Find something particularly insightful? Pin it to your Analyst’s Notes!
Explore Forecasts by Voir
Now it’s time for one of the most exciting features of the Newsplunker dashboard: Forecasts by Voir.
Voir is the AskNews forecasting model that is also available via API. Using a combination of the AskNews data, live web search (like wiki sources, statistics, and other news), LLMs, and state-of-the-art forecasting methodologies, Voir gives you a headstart on researching all the news media sources backing (or opposing) a particular outcome.
These rich forecast data help you analyze different scenarios and the scope of the world’s news covering what experts believe could happen, saving you hours of initial research and letting you dive into deep analysis from the get go.
To get started, just enter the natural-language query you want to forecast and then let Voir do the rest.
The resulting forecast gives you a high-level probability of your query and then follows up with analysis that breaks down the forecast in a number of ways, including sources, summary, and reasoning.
The prediction is meant to give you a high-level overview of the forecast but the real value is in the assembled data. And this is where Voir truly shines — in less than a minute, you’re provided with data that would take hours to gather with conventional methods.
So, after you’ve perused the high-level summary and likelihood of a given event, you’re ready to dive into the sources and analysis that informed the forecast. The Articles section gives links to all of the full-content news articles that fed into Voir’s forecast. Scroll down and you’ll also see Live Web Search results that informed the forecast, allowing you to explore the topic outside of just news media coverage.
You’re also provided with:
- The Resolution Criteria, i.e., the key to understanding the predicted outcome of your forecast. This section summarizes the standards and limitations used to define the resolution of your question, providing transparency and clarification on how Voir came to its particular forecast.
- The Reasoning and logic behind the forecast
- The Unique Information that clarifies key outliers and not-to-be-overlooked factors that informed Voir’s forecast. This section identifies information that might not apply to similar situations.
- The Timeline of key events as applicable to your query. Some events might be historical, while others might be the latest development in today’s headlines (and thus their true impact may vary as a situation develops), and the final date is likely to be tied to the Resolution Criteria determined and explained by Voir higher up on the dashboard.
- The Key People, i.e, the most important figures involved in the situation, each presented with a short summary of their role and possible impact on the forecast.
- The Key Facets build on the context of when and where. These 5–6 points each summarize a major factor related to the query and how it affected the probability of the Resolution Criteria (whether increasing, decreasing, or complicating the resolution). These Key Facets provide another launching point for further analysis.
- The Reconciled Information highlights differing aspects and conflicting pieces of information, and explains how they have been reconciled in the final prediction for the forecast.
- The Candidate Models helps you understand the reasoning models and frameworks that were weighed by and influenced Voir’s forecast. Again, these models provide a jumping-off point for your further, in-depth analysis into the given issue or query.
- The Expert Information wraps up the forecast and acts as a portal into the multiverse, allowing you to look into multiple different possibilities and timelines depending on a number of influencing factors. In particular, Voir provides six different outcomes that could differ from the main forecast provided above.
Behind the Scenes: How We Built the Newsplunker Dashboard
The AskNews team is always working to re-imagine the way people and LLMs consume and understand news. Building the Newsplunker dashboard was a key part of that mission. Here’s a glimpse at the tools, libraries, and tech we used to build it.
First, Newsplunker is made possible by Flowdapt — our in-house, open source orchestration software for large-scale adaptive modeling challenges. In turn, Flowdapt uses Ray or Dask as back-end executors.
If you didn’t already know, AskNews is actually a plugin for Flowdapt that leverages Qdrant and vLLM for RAG (Retrieval Augmented Generation) applications. For more information on how AskNews serves as a low-latency, natural-language link for any LLM chain, see our blog post on infusing any LLM with news using just one line of code.
Another big part of the Newsplunker dashboard is the news knowledge graph that underpins it. Read more about how we created the AskNews knowledge graph using a fine-tuned Phi-3 Mini Graph.
The quality of AskNews data and the news knowledge graph is fundamentally shaped by the diversified fine-tuning of our open-source entity extraction models, GLiNER-news.
Conclusion
From all of us here at AskNews, we hope you enjoy the Newsplunker dashboard. Please let us know what you think of it in the AskNews Discord.
We wanted to create a tool for analysts and experts to build and explore custom topic windows into the largest news knowledge graph on the planet — helping you sift through the noise of global media coverage and find the information you need in a contextualized way.
But you don’t have to take our word for it: Current AskNews Analysts can try out Newsplunker here or new customers can find an AskNews plan that works for you.
