# US 12,585,875 B2 · Device and method for processing temporal expressions from unstructured texts for filling a knowledge database

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,585,875 B2
- Document ID: US12585875B2
- Status: Granted
- Kind: B2
- Application number: 18/305,896
- Filed: 2023-04-24
- Published: 2023-11-02
- Granted: 2026-03-24
- Earliest claimed: 2022-05-02
- Assignees: Robert Bosch GmbH
- Applicants: Robert Bosch GmbH
- Inventors: Lukas Lange; Jannik Stroetgen; Heike Adel-Vu
- Web page: https://patentgrep.com/patent/US12585875B2

## CPC

- G06F 40/216: using statistical methods
- G06F 40/279: Recognition of textual entities
- G06F 40/30: Semantic analysis
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method and device for processing temporal expressions from unstructured texts for filling a knowledge database. A temporal expression in a text is determined. A type of the temporal expression is determined as a function of the text. The temporal expression and the type are mapped on a prediction of a value of the temporal expression in a context-free representation of the temporal expression.

## Claims

### Claim 1

A computer-implemented method for processing temporal expressions from unstructured texts for filling a knowledge database, the method comprising:
determining a temporal expression in a text;
determining a type of the temporal expression as a function of the text; and
mapping the temporal expression and the type on a prediction of a value of the temporal expression in a context-free representation of the temporal expression, wherein:
a model is configured to map the temporal expression and the type on the prediction, the temporal expression and the type being mapped on the prediction using the model, and
the temporal expression is provided, the context-free representation being provided, and the model is trained as a function of the temporal expression and of the context-free representation to map the temporal expression on a prediction for the context-free representation.

### Claim 2 (depends on 1)

The method as recited in claim 1, wherein the text that includes the temporal expression is provided, a reference date for the text being determined or predefined, a temporal characteristic being determined as a function of the context-free representation and as a function of the reference date, and the knowledge database being filled with the temporal characteristic.

### Claim 3 (depends on 1)

The method as recited in claim 1, wherein the context-free representation includes multiple slots, the temporal expression being represented by a content of the slots in a context-free manner, at least one of the slots being masked and the model is trained as a function of the temporal expression to map the temporal expression on a prediction for a masked content of the at least one masked slot.

### Claim 4 (depends on 3)

The method as recited in claim 3, wherein the model is trained in a first training phase, in which the at least one slot is unmasked and in which at least one slot is masked, to map the temporal expression and an unmasked content of the at least one slot, which is unmasked, on a prediction for a masked content of the at least one slot, which is masked.

### Claim 5 (depends on 4)

The method as recited in claim 4, wherein the model trained in the first training phase is trained in a second training phase, in which all of the slots are masked, to map the temporal expression on a prediction for a content of the slots.

### Claim 6 (depends on 3)

The method as recited in claim 3, wherein the model is trained in multiple iterations, fewer slots being masked in a first iteration than in a second iteration that is carried out after the first iteration.

### Claim 7 (depends on 1)

The method as recited in claim 1, wherein, using the model as the prediction, either a digit, or a number, or an arithmetic operator, or a letter, or a word element, or a word, or a placeholder is determined from a predefined vocabulary.

### Claim 8 (depends on 7)

The method as recited in claim 7, wherein the model is used to determine, for each masked content of a slot, a probability distribution across the vocabulary and, as a function of the probability distribution, the content of the slot.

### Claim 9 (depends on 1)

The method as recited in claim 1, wherein a text excerpt from the text is determined in a rule-based manner as a function of the text, which includes the temporal expression, and the prediction being determined using the model as a function of the text excerpt that includes the temporal expression.

### Claim 10 (depends on 1)

The method as recited in claim 1, wherein a text excerpt from the text is determined in a rule-based manner as a function of the text, which includes the temporal expression, and the model being trained as a function of the text excerpt that includes the temporal expression.

### Claim 11

A device configured to process temporal expressions from unstructured texts for filling a knowledge database, the device comprising:
a computing device configured to:
determine a temporal expression in a text,
determine a type of the temporal expression as a function of the text, and
map the temporal expression and the type on a prediction of a value of the temporal expression in a context-free representation of the temporal expression, wherein:
a model is configured to map the temporal expression and the type on the prediction, the temporal expression and the type being mapped on the prediction using the model, and
the temporal expression is provided, the context-free representation being provided, and the model is trained as a function of the temporal expression and of the context-free representation to map the temporal expression on a prediction for the context-free representation.

### Claim 12

A non-transitory computer-readable medium on which is stored a computer program including computer-readable instructions for processing temporal expressions from unstructured texts for filling a knowledge database, the instructions, when executed by a computer, causing the computer to perform the following steps:
determining a temporal expression in a text;
determining a type of the temporal expression as a function of the text; and
mapping the temporal expression and the type on a prediction of a value of the temporal expression in a context-free representation of the temporal expression, wherein:
a model is configured to map the temporal expression and the type on the prediction, the temporal expression and the type being mapped on the prediction using the model, and
the temporal expression is provided, the context-free representation being provided, and the model is trained as a function of the temporal expression and of the context-free representation to map the temporal expression on a prediction for the context-free representation.
