# US 2026/0073253 A1 · Hybrid Content Item Chunking For Retrieval Augmented Generation

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

- Number: US 2026/0073253 A1
- Document ID: US20260073253A1
- Status: Published application
- Kind: A1
- Application number: 18/916,272
- Filed: 2024-10-15
- Published: 2026-03-12
- Earliest claimed: 2024-09-06
- Assignees: Oracle International Corporation
- Applicants: Oracle International Corporation
- Inventors: Xin Zhang; Zheng Wang; Mengqing Guo; Yazhe Hu; Zhonghai Deng; Yimo Liu; Yuying Wang; Genyi Huang; Rongguang Wang; Tao Sheng
- Web page: https://patentgrep.com/patent/US20260073253A1

## CPC

- G06N 5/04: Inference or reasoning models
- G06F 40/211: Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
- G06F 40/30: Semantic analysis
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Hybrid content item chunking techniques for retrieval augmented generation (RAG) systems are disclosed. The techniques employ a dual approach, combining size-based and semantic chunking with layout-based chunking. The techniques analyze content for layout indicators, creating two sets of chunks that are then merged into a hybrid set. This hybrid set is loaded into a database for subsequent searches. The techniques offer several technical advantages, including improved handling of diverse document types, potential for parallel processing, and enhanced capture of both semantic meaning and structural layout. By maintaining size constraints and adapting to various formats, the techniques provide a more comprehensive representation of document content. The techniques overcome limitations of single-method approaches, potentially leading to more accurate information retrieval, improved context preservation, and enhanced RAG system performance across varied document types.

## Claims

### Claim 1

One or more non-transitory computer-readable media comprising instructions which, when executed by one or more hardware processors, cause performance of operations comprising:
performing a combination of size-based chunking and semantic chunking on at least a portion of a content item to obtain a first set of content item chunks;
analyzing at least a portion of a content item for layout chunking indictors to determine a set of one or more layout chunking markers;
based on the set of one or more layout chunking markers, performing a combination of size-based chunking and layout chunking on at least a portion of a content item to obtain a second set of content item chunks;
combining the first set of content item chunks and the second set of content item chunks to obtain a hybrid set of content item chunks, wherein the hybrid set of content item chunks comprises one or more content item chunks of the first set of content item chunks and one or more content item chunks of the second set of content item chunks;
loading the hybrid set of content item chunks into a database system; and
invoking the database system to perform a set of one or more searches of the hybrid set of content item chunks.

### Claim 2 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, the operations further comprising:
parsing at least a portion of a content item into a plurality of grammatical units;
forming a set of nodes wherein each node of the set of nodes comprises a respective grammatical unit of the plurality of grammatical units and associated metadata; and
wherein the performing the combination of size-based chunking and semantic chunking is based on the set of nodes.

### Claim 3 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, wherein the combining the first set of content item chunks and the second set of content item chunks to obtain the hybrid set of content item chunks comprises:
identifying a content item chunk in the second set of content item chunks that overrides a corresponding content item chunk in the first set of content item chunks; and
selecting the content item chunk in the second set of content item chunks for inclusion in the hybrid set of content item chunks and omitting the corresponding content item chunk in the first set of content item chunks from inclusion in the hybrid set of content item chunks.

### Claim 4 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, the operations further comprising:
determining a retrieval score reflecting a number of content item chunks retrieved from the database system by a retrieval augmented generation agent that are in a set of ground truth content item chunks associated with a query submitted to the retrieval augmented generation agent; and
modifying a hybrid content item chunking process based on the retrieval score.

### Claim 5 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, the operations further comprising:
determining a generation score reflecting a number of intermediate responses generated by a retrieval augmented generation agent that are sufficiently semantically similar to a set of one or more content item chunks retrieved from the database system by the retrieval augmented generation agent; and
modifying a hybrid content item chunking process based on the generation score.

### Claim 6 (depends on 1)

The one or more non-transitory computer-readable media of claim 1 the operations further comprising:
determining a generation score reflecting a number of a set of intermediate responses generated by a retrieval augmented generation agent that are sufficiently semantically similar to a query submitted to the retrieval augmented generation agent; wherein the set of intermediate responses are generated by the retrieval augmented generation agent based on a set of one or more content item chunks retrieved by the retrieval augmented generation agent from the database system; and
modifying a hybrid content item chunking process based on the generation score.

### Claim 7 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, the operations further comprising:
determining a retrieval augmented generation agent score reflecting how semantically similar an overall query response generated by a retrieval augmented generation agent is to a ground truth overall query response; wherein the overall query response is generated by the retrieval augmented generation agent based on a set of one or more content item chunks retrieved by the retrieval augmented generation agent from the database system; and
modifying a hybrid content item chunking process based on the retrieval augmented generation agent score.

### Claim 8

A method comprising:
performing a combination of size-based chunking and semantic chunking on at least a portion of a content item to obtain a first set of content item chunks;
analyzing at least a portion of a content item for layout chunking indictors to determine a set of one or more layout chunking markers;
based on the set of one or more layout chunking markers, performing a combination of size-based chunking and layout chunking on at least a portion of a content item to obtain a second set of content item chunks;
combining the first set of content item chunks and the second set of content item chunks to obtain a hybrid set of content item chunks, wherein the hybrid set of content item chunks comprises one or more content item chunks of the first set of content item chunks and one or more content item chunks of the second set of content item chunks;
loading the hybrid set of content item chunks into a database system;
invoking the database system to perform a set of one or more searches of the hybrid set of content item chunks; and
wherein the method is performed by at least one device including a hardware processor.

### Claim 9 (depends on 8)

The method of claim 8, further comprising:
parsing at least a portion of a content item into a plurality of grammatical units;
forming a set of nodes wherein each node of the set of nodes comprises a respective grammatical unit of the plurality of grammatical units and associated metadata; and
wherein the performing the combination of size-based chunking and semantic chunking is based on the set of nodes.

### Claim 10 (depends on 8)

The method of claim 8, wherein the combining the first set of content item chunks and the second set of content item chunks to obtain the hybrid set of content item chunks comprises:
identifying a content item chunk in the second set of content item chunks that overrides a corresponding content item chunk in the first set of content item chunks; and
selecting the content item chunk in the second set of content item chunks for inclusion in the hybrid set of content item chunks and omitting the corresponding content item chunk in the first set of content item chunks from inclusion in the hybrid set of content item chunks.

### Claim 11 (depends on 8)

The method of claim 8, further comprising:
determining a retrieval score reflecting a number of content item chunks retrieved from the database system by a retrieval augmented generation agent that are in a set of ground truth content item chunks associated with a query submitted to the retrieval augmented generation agent; and
modifying a hybrid content item chunking process based on the retrieval score.

### Claim 12 (depends on 8)

The method of claim 8, further comprising:
determining a generation score reflecting a number of intermediate responses generated by a retrieval augmented generation agent that are sufficiently semantically similar to a set of one or more content item chunks retrieved from the database system by the retrieval augmented generation agent; and
modifying a hybrid content item chunking process based on the generation score.

### Claim 13 (depends on 8)

The method of claim 8, further comprising:
determining a generation score reflecting a number of a set of intermediate responses generated by a retrieval augmented generation agent that are sufficiently semantically similar to a query submitted to the retrieval augmented generation agent; wherein the set of intermediate responses are generated by the retrieval augmented generation agent based on a set of one or more content item chunks retrieved by the retrieval augmented generation agent from the database system; and
modifying a hybrid content item chunking process based on the generation score.

### Claim 14 (depends on 8)

The method of claim 8, further comprising:
determining a retrieval augmented generation agent score reflecting how semantically similar an overall query response generated by a retrieval augmented generation agent is to a ground truth overall query response; wherein the overall query response is generated by the retrieval augmented generation agent based on a set of one or more content item chunks retrieved by the retrieval augmented generation agent from the database system; and
modifying a hybrid content item chunking process based on the retrieval augmented generation agent score.

### Claim 15

A system comprising:
at least one device including a hardware processor;
the system being configured to perform operations comprising:
performing a combination of size-based chunking and semantic chunking on at least a portion of a content item to obtain a first set of content item chunks;
analyzing at least a portion of a content item for layout chunking indictors to determine a set of one or more layout chunking markers;
based on the set of one or more layout chunking markers, performing a combination of size-based chunking and layout chunking on at least a portion of a content item to obtain a second set of content item chunks;
combining the first set of content item chunks and the second set of content item chunks to obtain a hybrid set of content item chunks, wherein the hybrid set of content item chunks comprises one or more content item chunks of the first set of content item chunks and one or more content item chunks of the second set of content item chunks;
loading the hybrid set of content item chunks into a database system; and
invoking the database system to perform a set of one or more searches of the hybrid set of content item chunks.

### Claim 16 (depends on 15)

The system of claim 15, the operations further comprising:
parsing at least a portion of a content item into a plurality of grammatical units;
forming a set of nodes wherein each node of the set of nodes comprises a respective grammatical unit of the plurality of grammatical units and associated metadata; and
wherein the performing the combination of size-based chunking and semantic chunking is based on the set of nodes.

### Claim 17 (depends on 15)

The system of claim 15, wherein the combining the first set of content item chunks and the second set of content item chunks to obtain the hybrid set of content item chunks comprises:
identifying a content item chunk in the second set of content item chunks that overrides a corresponding content item chunk in the first set of content item chunks; and
selecting the content item chunk in the second set of content item chunks for inclusion in the hybrid set of content item chunks and omitting the corresponding content item chunk in the first set of content item chunks from inclusion in the hybrid set of content item chunks.

### Claim 18 (depends on 15)

The system of claim 15, the operations further comprising:
determining a retrieval score reflecting a number of content item chunks retrieved from the database system by a retrieval augmented generation agent that are in a set of ground truth content item chunks associated with a query submitted to the retrieval augmented generation agent; and
modifying a hybrid content item chunking process based on the retrieval score.

### Claim 19 (depends on 15)

The system of claim 15, the operations further comprising:
determining a generation score reflecting a number of intermediate responses generated by a retrieval augmented generation agent that are sufficiently semantically similar to a set of one or more content item chunks retrieved from the database system by the retrieval augmented generation agent; and
modifying a hybrid content item chunking process based on the generation score.

### Claim 20 (depends on 15)

The system of claim 15, the operations further comprising:
determining a generation score reflecting a number of a set of intermediate responses generated by a retrieval augmented generation agent that are sufficiently semantically similar to a query submitted to the retrieval augmented generation agent; wherein the set of intermediate responses are generated by the retrieval augmented generation agent based on a set of one or more content item chunks retrieved by the retrieval augmented generation agent from the database system; and
modifying a hybrid content item chunking process based on the generation score.
