EKG Language Processing is a tool for extracting facts from text. Here are examples of tasks that can be solved using it:
- automatic processing of messages to the technical support service: classification, determination of mass problems and failures, determination of the subject of the request, and so on;
- analysis of contractual documents: identification of the parties to the contract, subject matter of the contract, scope of obligations, additional terms and conditions;
- analysis of organizational and administrative documents, as well as standards and other regulatory documentation: search for definitions of terms, requirements, responsibilities - including for duplications or contradictions;
- extracting facts from various text reports, policy briefs, publications, etc.
- normalization of product catalogues, automating MDM data content filling, generation of database content based on text lists and non-normalized tables
- creation of dialog assistants: chat bots, corporate knowledge base with dialog interface using RAG for structured data and texts.
Addressing such challenges will allow businesses to:
- save the time employees spend reading and analyzing documents,
- Increase accessibility of information (you may not be able to find the document or fact you need manually),
- enrich the contents of corporate databases, collect more information for analysis,
- speed up the processing of customer requests, improve the quality of organizational and administrative documents,
- automate the control of compliance with the requirements of these documents.
Our product has commercial use cases and brings real value to organizations!
See our description of the AI assistant for the technical support service, offering search in the knowledge base and Jira incidents. See also detailed description of using our product to classify technical support requests.We use the GraphRAG approach, which is aimed at extracting precise facts from text, transforming them into a knowledge graph, and using the graph as structured data and a source for forming context for LLM. This allows us to solve problems that are not accessible to a brute-force solution using large language models (LLM) or regular RAG. Our approach minimizes hallucinations and increases the accuracy of answers.
The EKG LP algorithm is as follows:
- extract pure text from the analyzed document (PDF, office formats)
- perform grammatical analysis of each phrase of the document
- build a conceptual scheme of the subject area described in the document
- extract specific facts and present them in accordance with the conceptual scheme
- integrate information from different documents by combining their conceptual schemes and factual information
- perform useful work: write the extracted structured information to the database, generate answers to questions, and so on.