Entity Extraction is the process of extracting named entities – like people, places, organizations and more – that appear in structured or unstructured text documents. We use a combination of proprietary and licensed text analytics models to correctly isolate these entities in text, and to categorize them according to their type.

Entity Disambiguation refers to the ability to resolve an entity’s identity to a knowledgebase. It is not merely entity-type classifying – as in determining that a reference to “George Washington” is a reference to the person and not the bridge in New York City, for example. Instead, entity disambiguation involves correctly distinguishing between two identically named entities of the same type – as in, John Roberts the Chief Justice of the U.S. Supreme Court, John Roberts the Fox News correspondent, or any one of the hundreds of individuals in the world named John Roberts.

Sentiment Assignment involves identifying a piece of content as positive, negative or neutral and returning a numeric score that indicates how positive, negative or neutral the item is. Finch for Text performs sentiment assignment at the document, sentence and entity level. We also use topic classification to drive accuracy. This approach allowed us to out-perform the sentiment tool in place at a major, international news content aggregator.

Relationship Extraction refers to the ability to discern how two entities are connected to one another. For companies, this can mean understanding supplier-customer, parent-subsidiary, or acquirer-acquired relationships. Understanding these connections can help organizations examine many types of risk or opportunity. Doing so in real-time and on huge volumes of text accelerates their ability to identify and assess these risks and opportunities, and then to make critical business decisions as a result.

Text Summarization involves creating short, fluent summaries of documents without losing key information or the document’s overall meaning and intent. Doing this accurately is challenging – doing it accurately, quickly and at scale proves even more so. We use abstractive and extractive summarization to produce better, more accurate summaries for our customers. We leverage cutting-edge summarization models that combine both approaches and use reinforcement learning to ensure the two approaches work in harmony.

Entity enrichment is a complement to the text analytics functions of extraction and disambiguation that involves adding additional data about an entity to a knowledge base entry of an entity so that it becomes richer and more valuable for further analysis. For example, understanding that business is also affiliated with a certain industry, has a certain URL or street address, uses a certain ticker symbol or social media handle, has a certain CEO and board members all make that entity more meaningful. From these enrichments, you can make connections in the data and link two seemingly disparate things to one another based on a characteristic they share – even if it’s not expressly mentioned in the text.