5.4 Language & Patent Model

Patent Search includes two important toggles that affect how searches are executed and which data and models are used:

• the Patent Model Toggle (data corpus selection)
• the Language Toggle (input language and semantic model)

Both controls are located in the header of the search input field and influence how the query is processed.

To the left of the input field is the selector for choosing the patent corpus to be searched:

• ALL (EN) – searches all globally available patent documents in English (including translations).
• DACH (DE) – limits the search to German-language patent documents from Germany, Austria, and Switzerland.

Note: The selected corpus determines which documents are searched – regardless of the input language.

Directly to the right of the corpus selector is the language toggle, which determines the language in which the input is interpreted and which semantic language model is used:

• EN – uses the English-language AI model (e.g., trained on USPTO data)
• DE – uses the German-language AI model (e.g., trained on DPMA data)

Note: This toggle only affects semantic filters. Classical Boolean queries or facet filters are not affected.

• When entering a text or concept, the language toggle defines which semantic model is used to evaluate similarity.
• When starting from a publication, the system often automatically detects the language and selects the appropriate mode.
• Switching between the EN and DE corpus does not automatically switch the semantic model – both toggles must be set manually and deliberately.

Patent Search supports cross