Summary
- Profile Type
- Technology offer
- POD Reference
- TODE20240425006
- Term of Validity
- 29 April 2024 - 29 April 2025
- Company's Country
- Germany
- Type of partnership
- Commercial agreement with technical assistance
- Targeted Countries
- All countries
Contact the EEN partner nearest to you for more information.
Find my local partner
General information
- Short Summary
- A German SME seeks a technological/commercial partner for high-quality Large Language Model (LLM) solutions for question answering (QA) to business customers. They possess a technology for largely automatic knowledge graph construction from text, databases, news feeds etc. Integrating domain knowledge from the graph will enhance answer accuracy, aid inferencing, and improve explainability. This will allow the partner, to develop a replicable QA offering, and ensure high customer satisfaction.
- Full Description
-
The German SME supplies resources for deep semantic processing with LLMs, such as knowledge graphs (KGs). A KG is a domain specific, symbolic network of concepts organized in simple statements, and comes with a domain's specific ontology. KGs are kind of complementary to LLMs, in that the latter have excellent question understanding and answer generation capabilities, while KGs provide factual and exact knowledge with inferencing and explanation capabilities.
The SME has a unique information extraction technology which allows them to generate KG elements with high speed from natural language text, from tabular data or other input formats. In many customer cases, this is an automatic process where no human intervention is needed once the extraction engine is set up. A product offering for automatic or semi-automatic knowledge graph creation is in preparation.
Once the KG for a customer's domain has been created, the SME supplies a library supporting the analysis of incoming queries, and the retrieval of KG statements (so called "triples") under various inferencing options. Based on these triples, and together with other information, a powerful prompt in RAG-style (Retrieval Augmented Generation) can be constructed, or an encoder model can be trained secondarily with the KG's triple information. In both ways, powerful KG question answering solutions knowledgeable about a customer's specific semantic domain can be built.
The SME is not interested in the business of supplying LLMs or foundation models to end customers. Instead, they are looking for partners to work with them on provisioning semantically enabled chatbots for such end customers.
Through knowledge graph-based QA technology, the SME will help the partner to differentiate from competition, and to develop a replicable QA offering with high end customer satisfaction achievable. The partnership should be technological as well as commercial. - Advantages and Innovations
-
• The SME's technology enables high-speed automatic or semi-automatic creation of KGs from various input formats
• KGs offer factual knowledge and precise knowledge with the possibility of drawing conclusions and providing explanations. They are thus complementing LLMs.
• The KGs created through the SME's technology support the analysis of incoming queries and prompt-building for LLM's, thus facilitating efficient question-answering and decision-making processes.
• Tailor-made KG question answering solutions knowledgeable about a customer's specific semantic domain can be built.
• The knowledge graph-based QA technology provides a competitive advantage and high end customer satisfaction. - Stage of Development
- Available for demonstration
- Sustainable Development Goals
- Not relevant
- IPR description
- The partner should be knowledgeable in the areas of LLM training and/or prompting and/or NLP-oriented web application development. For customer scenarios with local or hybrid NLP (Natural Laguage Programming) solution requirements, the question of sufficient computing resource needs to be settled.
Partner Sought
- Expected Role of a Partner
-
- Specific area of activity of the partner: Software developer/ NLP solution provider
- Tasks to be performed by the partner sought: Lead training, verification and testing activities with data delivered by the SME. - Type and Size of Partner
- R&D InstitutionSME 11-49UniversityOtherSME <=10SME 50 - 249Big company
- Type of partnership
- Commercial agreement with technical assistance
Call details
- Coordinator required
-
Yes
Dissemination
- Technology keywords
- 01003003 - Artificial Intelligence (AI)01005005 - Information Filtering, Semantics, Statistics01003015 - Knowledge Management, Process Management01003018 - User Interfaces, Usability
- Market keywords
- 02007016 - Artificial intelligence related software02007018 - Natural language02007023 - Web semantics
- Sector Groups Involved
- Digital
- Targeted countries
- All countries