Exploring SEC Filings with SightXR & Senzing

Context on the SEC filings

The Securities and Exchange Commission (SEC) filings are comprehensive documents that public companies and certain insiders must submit to the SEC, providing detailed information about their financial performance and operations. These filings are essential for maintaining transparency and protecting investors by ensuring they have access to critical information. Essential filings include Form 10-K, which provides an annual overview of a company’s financial health; Form 10-Q, offering quarterly updates; and Form 8-K, used for reporting significant events. Additionally, insider trading reports and proxy statements are part of these mandatory disclosures, enabling investors to make informed decisions based on a company’s financial status and corporate governance.

The Form 8-K filing includes a specific section dedicated to cybersecurity incidents, which companies must complete when they experience a significant cybersecurity event. This section requires companies to disclose any breaches or cyber attacks that could have a material impact on their business, financial condition, or operations. The information typically includes the nature and scope of the incident, the potential impact on the company, measures taken to address the breach, and any ongoing or anticipated consequences. By reporting these incidents promptly, companies ensure transparency and help maintain investor confidence while highlighting the importance of robust cybersecurity practices.

Loading SEC Filings into GraphXR with Streamlit App

Looking to retrieve these forms, we developed an app in Streamlit that connects with EDGAR text search (Electronic Data Gathering, Analysis, and Retrieval), enables the search for keywords in SEC Filings, downloads the text of these filings, and uploads them into GraphXR.

Figure 1 shows an example focusing on Material Cybersecurity Incidents, especially section 1.05 of 8-K forms, where we can get a list of the latest incidents in many public companies and a description of how these incidents happened. So, we get the complete filing and load them through this Streamlit app into the GraphXR environment for further analysis.

We see fields such as file date, companies display names, and locations, among others. Then, we can mark a checkbox to select the filings of interest to see the original text on the links and load them.

Figure 1 — Material Cybersecurity Incidents filings upload

Figure 2 — SightXR with 4 Sources from SEC Filings loaded

Knowledge Mapping from unstructured documents with SightXR

With the SEC filings loaded into GraphXR, we use SightXR, which leverages LLMs to map the entities and their relationships inside the content of the text of the filings. Figure 3 shows the resulting entities after running the knowledge map task.

Figure 3 — Entities detected in 4 SEC Filings as Sources

Figure 4 — Entities detected in one chunk of the original text in blue

Figure 4 shows a chunk of one of the SEC Filings text content with the found entities in blue (and separately in the right side panel).

To show the relationships between the entities we use GraphXR for the presentation, as seen in Figure 5.

Figure 5 — GraphXR presenting the entities and relationships from SEC Filings

Entities Resolution with Senzing

Since the entities are associated with chunks that depend on the size of the context of the LLM that was used, we can have more than one occurrence of the same entity in different chunks with some spelling differences. We can merge these entities, but in most cases, we don’t want to because we want to know where each piece of information came from. To resolve these entities, we use the Senzing Framework. In this case, we disambiguate the organizations while creating relationships in red to preserve provenance. We could merge the entities that are relatable to each other with a Senzing matching level indicating that they are possibly the same entity, but we opted to keep all the instances of the entities and show those relationships in red.

Figure 6 — Complete graph after creating Senzing relationships in red for match level, indicating they are possibly the same entity

Bringing Senzing results as relationships into Neo4j

The final step involves bringing Senzing results as relationships into Neo4j for a comprehensive and connected data view. This integration facilitates advanced querying and deeper insights into the interconnected entities and relationships derived from SEC filings.

Figure 7 — Entities related by a red relationship indicating they are possibly the same entity

Conclusion

By leveraging SEC filings and integrating advanced tools like Streamlit, GraphXR, SightXR, and Senzing, we can significantly enhance our understanding of cybersecurity incidents and their impacts on public companies. This approach ensures transparency and investor confidence and highlights the importance of robust cybersecurity practices in today’s digital landscape.

These tools and methodologies provide a powerful framework for analyzing and visualizing complex data from SEC filings, ultimately aiding in better decision-making and risk management for investors and stakeholders.