3. General overview

INTERGATOR Patent Search supports the entire workflow of modern patent search: from the initial search query, through reviewing and evaluating results, to structured analysis, relevance assessment and documentable preparation of the results. The platform combines semantic search, classic Boolean search, AI assistants, re-ranking functions and report generation into one continuous workflow.

A typical workflow with INTERGATOR Patent Search consists of the following steps:

• Start search: A search begins with a natural-language input, an invention disclosure, a publication number, a reference document or a specialized search assistant.
• Review result list: The documents found are reviewed, filtered, sorted, marked, hidden or saved to lists in the result list.
• Use reference document: Particularly relevant results can be used as reference documents to find similar documents, create semantic filters or further refine the search.
• Run relevance analysis: Individual results or selected documents can be analyzed with regard to technical relevance, feature overlaps, key text references and technical proximity to the initial question.
• Support validity analysis: For opposition or validity-related questions, reference documents, priority dates, technical features and relevant prior art can be reviewed and assessed in a structured way.
• Use re-ranking: Additional filters, selected relevant documents, semantic similarity and technical evaluation criteria can be used to further refine the order of the results.
• Document results: Searches, assessments and analysis results can be used as the basis for traceable reports, for example for internal coordination, technical evaluations or further processing by specialist departments.

The indexing of patent data is a central component of INTERGATOR Patent Search. It is based on an extensive, regularly updated data set derived from Questel publications. This includes worldwide patent data – especially with English translations – and ensures that current and relevant information is available for patent search.

Each patent application is represented by its latest publication. Members of a patent family with the same priority are grouped into one result.

All content is processed exclusively on servers in an ISO 27001-certified data center in Germany during use – without connection to external AI cloud services.

INTERGATOR Patent Search supports three basic types of search:

• semantic search
• Boolean search
• hybrid search

These search methods differ in how search terms are processed and how results are identified and ranked.

The search methods form the foundation of the research process. Building on this, AI assistants and analysis functions support the further evaluation of the documents found. While the search identifies matching results, relevance analysis, validity analysis, re-ranking and report functions help to classify, prioritize and document the results in a technically traceable way.

Semantic search uses AI language models to evaluate terms, phrases and texts in their contextual meaning. This makes it possible to find documents that are relevant in terms of content, even if they do not contain the exact search terms.

A semantic search is typically created through:

• natural-language inputs,
• free text,
• text passages,
• or semantic concept filters.

Semantic search is particularly suitable as an entry point into an analysis workflow. Based on an invention disclosure, a technical question or a reference document, relevant results can first be identified. These results can then be further evaluated, compared and prepared for reports using AI-supported analysis functions.

Advantages:

• Relevance based on semantic proximity instead of exact word matches
• Recognition of related terms and technical relationships
• Particularly suitable for invention disclosures, questions and free-text inputs
• Good basis for downstream relevance analysis, feature comparison and re-ranking

Disadvantages:

• Broader result spread with unspecific texts or unclear terms
• Relevance assessment may require explanation and, if necessary, refinement through filters
• Technical assessment and legal classification must still be reviewed by a specialist

Boolean search is based on lexical search logic. It works with concrete terms, field references and Boolean operators such as AND, OR and NOT. In INTERGATOR Patent Search, a purely Boolean search is performed using the BOOL perspective.

The following applies:

• The input is not interpreted semantically.
• The search is performed on the basis of the entered search syntax.
• The Boolean query has a ranking effect and therefore also influences the order of the results.
• Semantic or similarity functions are not available in this perspective.

Boolean search is particularly helpful when known terms, specific patent numbers, applicants, classifications or fixed search strategies are to be used in a controlled way. It can also be used as a supplement to narrow down semantically generated result lists or to prepare specific aspects for later analysis.

Advantages:

• High precision with known terms, field queries and fixed search patterns
• Controlled search with exact search syntax
• Particularly suitable for specific technical terms, patent numbers and structured queries

Disadvantages:

• Purely word- and syntax-based
• Semantically related or differently formulated content is not automatically captured

Hybrid search combines semantic search with additional lexical or bibliographic restrictions. It is particularly useful when relevant results should be found based on meaning while also being narrowed down in a targeted way.

Typical components of a hybrid search include:

• semantic search using free text or concepts,
• syntax filters,
• bibliographic filters,
• facets,
• or classifications such as IPC or ICM.

The following applies:

• The semantic search remains the foundation of the research process.
• Additional lexical or bibliographic elements have a filtering effect in this case, but generally do not affect ranking.

Hybrid search is often the practical standard case in INTERGATOR Patent Search. It combines broad, meaning-based result identification with controlled restrictions. This creates a reliable basis for further result evaluation, relevance analysis, validity analysis and report generation.

Advantages:

• Combines semantic breadth with precise restriction
• Particularly suitable for complex questions
• Flexibly adaptable during the ongoing search process
• Supports the gradual refinement of results before analysis and reporting

Disadvantages:

• Requires a certain understanding of the interaction between semantic and lexical filters
• May require more coordination and fine-tuning for very complex queries

With INTERGATOR Patent Search, entire texts, paragraphs or questions can be used directly as search queries. Natural-language input is usually an entry point into semantic search.

The system analyzes the input in terms of content, identifies relevant concepts and creates a search query from it, which can then be refined using additional filters.

Natural-language inputs are particularly suitable when no fully formulated search strategy exists yet. They can be used, for example, on the basis of an invention disclosure, a technical problem or a known patent document. The resulting hits can then serve as a starting point for reference documents, re-ranking, AI-supported relevance analyses or reports.

Advantages:

• Simple entry point without search syntax
• Ideal for research based on invention disclosures or technical problem descriptions
• Well suited as a starting point for semantic or hybrid search
• Supports the transition from an open question to a structured analysis

Disadvantages:

• Very general or vague formulations may require subsequent refinement
• For strictly controlled search strategies, Boolean search is often more suitable

The AI assistants in INTERGATOR Patent Search support typical search and analysis tasks through guided workflows. They help generate suitable search queries and filters from inputs such as texts, publication numbers or reference documents.

The assistants do not make the technical decision, but support the structured preparation, execution and evaluation of the search. They can take into account relevant terms, technical concepts, document similarities, priority information, family relationships and other criteria.

The analysis functions build on the search results. They help to evaluate results more precisely and classify their relevance for the respective question. These include in particular:

• the technical summary of individual patent documents,
• the identification of key features and feature groups,
• the analysis of feature overlaps,
• the highlighting of relevant text passages,
• the assessment of technical proximity to a reference document,
• support for validity and FTO-related questions,
• as well as the preparation of results for documentable reports.

After the initial search, the result list can be further refined. In addition to classic filters and facets, relevant results or reference documents can be used to sharpen the search direction and better align the order of results with the specific question.

In this context, re-ranking means that a result list is not viewed only once, but can be reprioritized using additional signals and technical criteria. These include, for example:

• selected relevant results,
• semantic document similarity,
• technical features,
• feature overlaps,
• bibliographic restrictions,
• priority or publication dates,
• as well as manually set filters.

Re-ranking helps to gradually condense large result sets down to the technically most important documents. It does not replace assessment by the searcher, but it facilitates prioritization and further analysis.

An essential goal of modern patent search is not only to find relevant documents, but also to document the search results in a traceable way. INTERGATOR Patent Search supports this step through structured analysis and report functions.

Reports can be used to prepare search results for internal coordination, technical evaluations, validity analyses, FTO assessments or monitoring evaluations. They can combine, among other things, the following content:

• initial question and search context,
• search strategy and filters used,
• relevant results and reference documents,
• technical summaries,
• feature overlaps,
• text references and reasoning,
• assessments of technical relevance,
• as well as notes on further analysis steps.

This makes the search traceable, reusable and easier to share with other stakeholders.