US 2026/0099559 A1 · Systems and Methods for Link Validation with Machine Learning

Abstract

A link validation system can iterate through a set of source documents to identify link errors associated with the documents. For a source document, the system can extract a source link that includes source text and a resource identifier. The system can retrieve one or more resources identified by the resource identifier and generate one or more images of the resource(s). The system can generate one or more machine learning prompts that include the image(s) and the source text. The prompt(s) can be provided to a multimodal sequence processing model that is configured to respond to the prompt by determining whether there is a link error such as a semantic dissimilarity and/or a target resource error. If a semantic dissimilarity and/or a target resource error is detected, the system can generate an indication of a link error associated with the source link.

Claims20

Dependent claims

A computer-implemented method for validating documentation links, the method comprising, by a computing system including one or more computing devices:

extracting a source link from a source document, the source link including source text and a resource identifier;
generating an image of a resource specified by the resource identifier;
providing one or more prompts including the image of the resource and the source link from the source document to a machine-learned multimodal sequence processing model;
determining, with the machine-learned multimodal sequence processing model and based on the image of the resource and the source link from the source document, at least one of a semantic dissimilarity between the source link and the resource or an error associated with the resource;
and
generating, in response to the semantic dissimilarity or the error associated with the resource, an indication of a link error associated with the source link.

3 more elements

A computing system, comprising:

one or more processors;
and
one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, comprising:
extracting a source link from a source document, the source link including source text and a resource identifier;
generating an image of a resource specified by the resource identifier;
providing one or more prompts including the image of the resource and the source link from the source document to a machine-learned multimodal sequence processing mode;
determining, with the machine-learned multimodal sequence processing model and based on the image of the resource and the source link from the source document, at least one of a semantic dissimilarity between the source link and the resource or an error associated with the resource;
and
generating, in response to the semantic dissimilarity or the error associated with the resource, an indication of a link error associated with the source link.

6 more elements

One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more processors, cause the one or more processors to perform operations, comprising:

extracting a source link from a source document, the source link including source text and a resource identifier;
generating an image of a resource specified by the resource identifier;
providing one or more prompts including the image of the resource and the source link from the source document to a machine-learned multimodal sequence processing mode;
determining, with the machine-learned multimodal sequence processing model and based on the image of the resource and the source link from the source document, at least one of a semantic dissimilarity between the source link and the resource or an error associated with the resource;
and
generating, in response to the semantic dissimilarity or the error associated with the resource, an indication of a link error associated with the source link.

3 more elements

Dependencies are inferred from the claim text.

Systems and Methods for Link Validation with Machine Learning

Life of this patent

Systems and Methods for Link Validation with Machine Learning

Life of this patent

Abstract

Claims20