FLOWCHART 1_Custom Built Agents



Topic Agents are designed to search through a large quantity of text and assign a similarity score between 0 and 1 at either the paragraph or document level.  They use the power of artificial intelligence to do advanced searches and categorization of documents.

Bintel works with clients to create an initial list of themes or topics they want to track or discover in their data.  An important component of the agents is keywords, so the Bintel Team collaborates with the client to develop a list of keywords for each Topic Agent. In order to avoid false positives the keywords and phrases must be precise and only used for the topic the agent is being trained for.

Bintel then takes the keywords and example text from the clients data to train the Topic Agents against sample texts from the clients data.  It is important to use a client's bodies of text for training because the Topic Agents artificial intelligence capabilities are based on text patterns with boosting from keywords and phrases.  Each type of writing uses different language styles and words have different meaning based on context, for example what an aerospace company would write about "space" would be completely different from what an architect or interior designer would write about.

Our method saves clients time and money while being more comprehensive than traditional research methods.  And because the Topic Agents use text patterns they are able to find information hidden in the text that is exactly what the agent is looking for, but also find information that provides a different solution to the same problem.