Hierarchical identification system and method for unstructured document content

Abstract

The invention discloses a hierarchical identification system and method for unstructured document content. The hierarchical identification system comprises a field hierarchical configuration unit, a mark coding and uniqueness maintenance unit, an association storage unit, a link packaging and analyzing unit and a multi-strategy positioning and backspacing retrieval engine. Configuring a field set and a hierarchical relationship according to a scene, setting a field sequence and a separator or a structured coding rule, generating a content identification mark at least containing a FileID and a content unit identification field, performing escape / coding on a field value, and updating the field and / or supplementing Version and a verification field during drifting or conflicting; through mapping indexes and / or metadata, maintaining correspondence between the marks and the content units, embedding the marks into private / universal links, and analyzing and extracting the marks; during positioning, automatic rollback is carried out among step-by-step analysis, index query and link analysis, after page-level convergence is carried out, positioning is completed in combination with offset and feature information approximate retrieval, and DeviceID / User ID cross-environment mapping is supported.

Description

Technical Field

[0001] This invention relates to the field of information identification and retrieval technology, specifically to a hierarchical identification system and method for unstructured document content (including but not limited to PDF, Word, web pages, etc.), and particularly to the generation, link encapsulation and transmission, and multi-strategy positioning and fallback retrieval of hierarchical, composable, and extensible unique identifiers for arbitrary granularity content units within a document. Background Technology

[0002] In large-scale document management, knowledge processing, and content referencing scenarios, the precise referencing and positioning of content in unstructured documents (such as PDFs, Word documents, and web pages) has long been a challenge. Taking PDFs as an example, relevant standards have defined fragment identifiers for PDFs, allowing the expression of positioning parameters such as page and name / ddest in the fragment portion of the URI, and specifying rules for parameter processing and encoding. However, in practical applications, different browsers / readers have varying levels of support for fragment identifiers, and many scenarios still rely on page-level navigation, making it difficult to provide a unified, flexible, and scalable universal identifier for fine-grained content such as "line / paragraph / sentence / character offset / region." On the other hand, fragment identifier standards for text / plain documents allow referencing text fragments by character position / range or line position / range, and can carry integrity verification information to enhance robustness. However, similar capabilities have not formed a consistent and transferable application-layer universal mechanism in complex document formats such as PDFs and Word documents and their cross-platform rendering environments.

[0003] Meanwhile, with the widespread application of large-scale models in document intelligence and retrieval augmentation generation (RAG) technologies, systems often need to provide more granular and verifiable evidence location and citations when generating answers. Existing research indicates that simply citing coarse-grained document identifiers increases user verification costs, and more granular grounded citations are needed to support verifiability and reduce the risk of illusion. Therefore, in scenarios such as multi-source, multi-format document fusion, question answering, summarizing, comparing, and auditing, there is an urgent need for a hierarchical identifier system that can uniformly express information across document types, has an extensible field set, and possesses a location fallback mechanism to support credible citations, rapid retrieval, evidence traceability, and cross-system collaboration.

[0004] In summary, the existing technology has at least the following shortcomings: 1) Insufficient uniqueness of content identifiers: There is a lack of a universal mechanism to establish a resolvable and unique identifier for content units of any granularity in multi-source documents, making it difficult to ensure uniqueness after cross-platform and cross-user transmission; 2) Insufficient robustness in positioning: When relying solely on positional elements such as page numbers / line numbers / offsets, document editing, reflow, addition, deletion, or pagination changes can easily cause positioning drift or failure; 3) Insufficient scalability and compatibility: Fields, levels, and granularity are difficult to flexibly tailor and expand according to different business scenarios, and there is a lack of multi-strategy positioning and failure rollback mechanisms; 4) Insufficient multi-user / cross-application / cross-device referencing: There is a lack of a unified solution for embedding identifiers into transitive links and automatically resolving and positioning them in the target application.

[0005] In view of the above problems, it is necessary to provide a new technical solution that can establish a hierarchical, composable, and scalable unique identification system for multiple types of document content, so as to support fine-grained content accurate referencing, fast retrieval, and multi-system, multi-user, and cross-application software collaboration in complex scenarios driven by large models. Summary of the Invention

[0006] This invention aims to provide a hierarchical identification system and method for unstructured document content. It generates content identification tags through a configurable set of fields and field hierarchical relationships, and combines associated storage, link encapsulation and parsing, as well as multi-strategy positioning and fallback retrieval to achieve fast, verifiable, and scalable referencing and positioning of target document content units (lines / paragraphs / sentences / character offsets / regions, etc.), thereby solving the shortcomings of existing technologies in terms of uniqueness, robustness, scalability, and cross-environment positioning.

[0007] To achieve the above objectives, the present invention provides a hierarchical identification system for unstructured document content, which includes at least: a field hierarchy configuration unit, a tag encoding and uniqueness maintenance unit, an association storage unit, a link encapsulation and parsing unit, and a multi-strategy positioning and fallback retrieval engine; the units interact with each other through direct connection or indirect connection via bus, middleware, or interface service.

[0008] 1. Field-level configuration unit This is used to configure the set of fields and the hierarchical relationship of the content identifier tags in different application scenarios, and to configure the preset field order and preset delimiter rules or structured encoding rules.

[0009] The set of fields includes at least the following: File ID field; Content unit identifier field (at least one of line identifier field, paragraph identifier field, sentence identifier field, character offset field, or region identifier field).

[0010] The field hierarchy includes at least: FileID → Content Unit Identifier field. Optionally, the field set may further include a Page Identifier field (PageID) and / or a Device Identifier field (DeviceID) and / or a User Identifier field (UserID), and allows for scenario-based trimming to omit DeviceID and / or UserID; it can also be expanded to include collaboration or management entity fields, which are at least one of a Group Identifier field (GroupID), a Project Identifier field, a Permission Identifier field, or a Session Identifier field. In one embodiment, the field hierarchy includes at least one of the following dependencies: FileID → Content Unit Identifier field; or FileID → PageID → Content Unit Identifier field; when the field set includes GroupID, it also includes a GroupID → FileID dependency.

[0011] 2. Tag Encoding and Uniqueness Maintenance Unit Connected to the field hierarchy configuration unit, it is used to generate content identifier tags based on the field set, field hierarchy relationship, and delimiter rules or structured encoding rules. The generation includes: combining the field values ​​in a preset order to form a tag, and escaping or encoding the field values ​​to avoid ambiguity between the field values ​​and delimiters.

[0012] The structured encoding rules may include at least one of URL encoding, Base64 encoding, JSON encoding, or CBOR encoding, and may escape field values ​​to avoid parsing ambiguity.

[0013] The uniqueness maintenance includes updating at least one field value and / or adding a version field and / or a verification field to the tag when document content editing, reflow, addition, deletion, or pagination changes cause positioning drift and / or tag conflicts are detected, to generate a new resolvable tag and maintain uniqueness. The verification field may be at least one of checksum, digest hash, digital signature, random code, or timestamp.

[0014] 3. Associated storage unit Connected to the tag encoding and uniqueness maintenance unit, it is used to establish and maintain a one-to-one correspondence between content identification tags and unstructured document content units; wherein the correspondence is achieved through one or a combination of the following methods: (1) Maintain a mapping table or index structure for “tag-location information”, wherein the location information includes at least one of the following: file identifier, page identifier, content unit identifier, character offset range, or rendering coordinate area; (2) Store at least some field values ​​in the metadata of the database object, file object, page object and content unit object respectively, so as to support the level-by-level matching and convergence to the target content unit according to the field hierarchy.

[0015] When using a mapping table or index structure, the location information may include at least a file path or file object reference, a page identifier, and a character offset range and / or a rendering coordinate region; the index structure may be at least one of a hash index, a B-tree index, or an inverted index to support one-hop query location based on content identifier tags.

[0016] 4. Link Encapsulation and Resolution Unit Connected to the tag encoding and uniqueness maintenance unit, it is used to embed content identifier tags in an encoded manner into the path segment and / or query parameters of the link to generate a reference link, and to parse and extract the content identifier tags from the reference link; the reference link is an application-private link or a general link.

[0017] The reference link may include a version field and / or a verification field to indicate parsing rules and integrity checks. Upon receiving the reference link, the target application triggers the opening of the target unstructured document and triggers at least one of the following: navigation, scrolling, or highlighting of the target content unit.

[0018] 5. Multi-strategy location and fallback search engine Connected to the associated storage unit and the link encapsulation and parsing unit, it is used to locate unstructured document content units based on content identifier tags, wherein the location includes at least one of the following strategies or a combination thereof: (a) Step-by-step parsing and positioning: Parse the tags according to the delimiter and field order or structured encoding rules, and locate them step by step according to the file → page → content unit; (b) Index lookup location: Use the tag as the key to query the mapping table or index structure to obtain location information or object references to achieve location; (c) Link parsing and location: Parse the reference link to extract the tag, and select the corresponding parsing rule when the version field is included, and perform integrity verification before location when the verification field is included; Furthermore, the multi-strategy positioning and fallback retrieval engine supports automatic fallback to another positioning strategy when any positioning strategy fails, and after positioning to the page level range, it performs approximate retrieval by combining character offset and / or feature information to complete the positioning.

[0019] The feature information may include at least one of the following: text fragment summary hash of the target content unit, keyword set, context window text, or page area features, used to determine the target content unit after locating to the page level.

[0020] 6. Cross-device / cross-user transfer and mapping (optional implementation) When content identifiers are transmitted across devices and / or users and parsed and located in the target environment, the tag encoding and uniqueness maintenance unit can replace, rewrite, or map the DeviceID and / or UserID in the tag based on the device identifier and / or user identifier in the target environment to generate a content identifier tag suitable for the target environment and maintain parsability. In one embodiment, a mapping table or index structure between source field values ​​and target field values ​​is maintained, recording at least the source device identifier and / or source user identifier, the target device identifier and / or target user identifier, and the corresponding FileID, so that the tag containing the source field value can be converted into the target field value to continue locating the target content unit.

[0021] The present invention also provides a hierarchical identification method for unstructured document content, applied to the above-mentioned system, comprising the following steps: S1. The field hierarchy configuration unit configures the set of fields for content identification tags and the field hierarchy relationship for different application scenarios, and configures the preset field order and preset delimiter rules or structured encoding rules. S2. The tag encoding and uniqueness maintenance unit generates content identification tags based on the field set, field hierarchy and delimiter rules or structured encoding rules, and escapes or encodes field values ​​to avoid ambiguity; when document content changes and / or tag conflicts are detected, at least one field value is updated, and / or a version field and / or a validation field are added to generate a new unique content identification tag. S3. The associated storage unit establishes and maintains a one-to-one correspondence between content identifiers and target content units. The one-to-one correspondence is achieved by maintaining a mapping table or index structure of “tag-location information” and / or writing at least some field values ​​into the metadata of database objects / file objects / page objects / content unit objects to support hierarchical matching. S4. The link encapsulation and parsing unit embeds the content identifier into the path segment and / or query parameters of the link in an encoded manner to generate a reference link, and parses and extracts the content identifier when receiving the reference link; S5. The multi-strategy positioning and fallback retrieval engine locates the target content unit based on the content identifier tag using a step-by-step parsing positioning strategy and / or an index query positioning strategy and / or a link parsing positioning strategy. If any positioning strategy fails, it automatically falls back to another positioning strategy. After locating the page-level range, it performs an approximate retrieval by combining character offset and / or feature information to complete the positioning.

[0022] In one implementation, S5 may first employ an index query positioning strategy; when the index query fails or the index expires, a fallback positioning strategy is adopted to converge the positioning range to the file and page; and an approximate retrieval is performed within the page level based on character offset and / or feature information to determine the target content unit.

[0023] Compared with the prior art, the present invention has at least the following beneficial effects: 1) A unique identifier for content units is achieved through multi-field hierarchical combination, reducing reference conflicts in cross-document, cross-user, and cross-device scenarios; 2) By combining multiple strategies such as "mapping index location + hierarchical parsing location + link parsing location" and automatic rollback, the success rate and robustness of location are improved in cases of missing indexes, expired indexes, or changes in document structure. 3) Through field pruning and expansion mechanisms, as well as the addition of version fields / validation fields, the field set and parsing rules can be made more evolving and compatible; 4) By embedding tags into application-private or universal links, it enables cross-application, cross-device, and cross-user transitive referencing and automatic location recall capabilities; 5) By approximating the retrieval of character offsets and feature information after page-level convergence, the failure rate caused by positioning drift during document editing is reduced, thereby enhancing usability. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the hierarchical identification system for unstructured document content according to the present invention; Figure 2 This is a schematic diagram of the process of encapsulating and parsing reference links in this invention. Detailed Implementation

[0025] The system and method of the present invention will be further described below with reference to the accompanying drawings. The following embodiments are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0026] Implementation Method 1: System Overall Architecture Please see Figure 1 , Figure 1 This is a schematic diagram of a hierarchical identification system for unstructured document content in one embodiment of the present invention.

[0027] The hierarchical identification system for unstructured document content provided by this invention includes: a field hierarchy configuration unit 10, a tag encoding and uniqueness maintenance unit 20, an association storage unit 30, a link encapsulation and parsing unit 40, and a multi-strategy positioning and fallback retrieval engine 50. The above units can interact with each other through direct connections or indirect connections via buses, middleware, or interface services. The above units can be implemented by software, hardware, or a combination of both, and the function of one unit can be shared by multiple units or combined from a single unit.

[0028] The field hierarchy configuration unit 10 is used to configure the field set and field hierarchy relationship of content identifiers in different application scenarios, and to configure the preset field order and preset delimiter rules or structured encoding rules.

[0029] The tag encoding and uniqueness maintenance unit 20 is connected to the field hierarchy configuration unit 10. It is used to generate content identification tags based on the field set, field hierarchy relationship and delimiter rules or structured encoding rules; and when document content editing, reflow, addition, deletion or pagination changes cause positioning drift and / or tag conflicts are detected, it updates at least one field value and / or adds a version field (Version) and / or a validation field to the tag to generate a new parsable tag and maintain uniqueness.

[0030] The associated storage unit 30, connected to the tag encoding and uniqueness maintenance unit 20, is used to establish and maintain a one-to-one correspondence between content identification tags and unstructured document content units. The correspondence can be achieved by maintaining a mapping table or index structure of "tag-location information" and / or by distributing at least some field values ​​in the metadata of database objects / file objects / page objects / content unit objects to support hierarchical matching.

[0031] The link encapsulation and parsing unit 40, connected to the tag encoding and uniqueness maintenance unit 20, is used to embed content identifier tags into the path segment and / or query parameters of the link in an encoded manner to generate a reference link, and to parse and extract content identifier tags from the reference link; the reference link is an application-private link or a general link, and may contain a version field and / or a verification field to indicate parsing rules and integrity verification.

[0032] The multi-strategy positioning and fallback retrieval engine 50, connected to the associated storage unit 30 and the link encapsulation and parsing unit 40, is used to locate unstructured document content units based on content identifiers. It supports hierarchical parsing positioning, index query positioning, and / or link parsing positioning, and supports automatic fallback to another strategy when any strategy fails. After locating to the page level range, it can combine character offset and / or feature information to perform approximate retrieval to complete the positioning.

[0033] The above system enables unified identification, transferable referencing, and robust positioning of content units within unstructured documents (such as lines, paragraphs, sentences, character offset ranges, or page layout areas).

[0034] Implementation Method 2: Design of Tag Field Set and Field Hierarchy In one implementation, the content identifier consists of multiple fields, each corresponding to a different entity and having an inclusion or subordinate relationship, thereby forming a hierarchical structure.

[0035] 1) The minimum required fields for a field set The field set includes at least: a file identifier field (FileID) and a content unit identifier field; wherein the content unit identifier field is at least one of a line identifier field, a paragraph identifier field, a sentence identifier field, a character offset field, or a region identifier field. The field hierarchy relationship includes at least: FileID → content unit identifier field.

[0036] 2) Optional fields and scenario-based cropping In different application scenarios, the field set may optionally include: a page identifier field PageID and / or a device identifier field DeviceID and / or a user identifier field UserID; and can be trimmed according to the scenario to omit DeviceID and / or UserID.

[0037] When the field set includes PageID, the field hierarchy can include: FileID → PageID → Content Unit Identifier field.

[0038] 3) Extended fields for collaborative or management entities The field set can also be expanded to include collaboration or management entity fields, which are at least one of the following: GroupID, project ID, permission ID, or session ID. When the field set includes GroupID, the field hierarchy can further include: GroupID → FileID.

[0039] 4) Field value examples and tag format examples In one example implementation, the content identifier can be represented as a "sequence of fields grouped in a preset field order", for example: Separator format example: <fileid> # <pageid> @ <contentunitid> or <groupid> # <fileid> # <pageid> # <contentunitid> ContentUnitID can be a line number, paragraph number, sentence number, character offset range (start and end positions), or region identifier (such as the encoded representation of rectangular coordinates / polygon point sets).

[0040] Example of structured encoding format: The key-value pairs of fields are encapsulated into structured objects (such as a collection of key-value pairs), and then URL-securely encoded or Base64 encoded to form a tag.

[0041] It should be understood that the delimiter can be any character, string, or delimiter rule; the field hierarchy does not necessarily have to exist, and can be selected, trimmed, or expanded according to specific business scenarios without affecting the basic concept of the present invention.

[0042] Implementation Method 3: Tag Generation, Escape / Encoding, and Uniqueness Maintenance In one implementation, the tag encoding and uniqueness maintenance unit 20 generates content identification tags according to the field set, field hierarchy relationship, preset field order, and delimiter rules or structured encoding rules configured by the field hierarchy configuration unit 10.

[0043] 1) Tag generation mechanism The tag encoding and uniqueness maintenance unit 20 combines the field values ​​in a preset order to form a tag; when using delimiter rules, the field values ​​are escaped or encoded to avoid ambiguity between the field values ​​and the delimiters; when using structured encoding rules, at least one of URL encoding, Base64 encoding, JSON encoding or CBOR encoding can be used to encode the field values ​​or field sets to ensure parsing consistency.

[0044] 2) Uniqueness maintenance and tag update mechanism When document content editing, reflow, addition, deletion, or pagination changes cause positioning drift and / or detect tag conflicts, the tag encoding and uniqueness maintenance unit 20 can update at least one field value and / or supplement the tag with a version field (Version) and / or a validation field to generate a new resolvable tag and maintain uniqueness. Preferably, the update / supplementation may include, but is not limited to: Update the content unit identifier fields (e.g., line identifier / paragraph identifier / sentence identifier / character offset range / region identifier) ​​and / or update the page identifier field PageID; The Version field has been added to indicate the set of fields and the parsing rules. An additional verification field is added to perform integrity verification on the combination of field values; the verification field can be at least one of checksum, digest hash, digital signature, random code or timestamp.

[0045] Therefore, new tags can be generated to distinguish old tags when tag conflicts or location drift occur, while maintaining uniqueness while preserving resolvability.

[0046] Implementation Method 4: Association Storage Method of Tag and Content Unit In one implementation, the associated storage unit 30 is used to establish and maintain a one-to-one correspondence between content identifiers and unstructured document content units, wherein the correspondence can be implemented in any or a combination of the following ways: 1) Mapping table / index structure method The system maintains a mapping table or index structure of "tag-location information". The content identifier tag is stored as the key and the location information is stored as the value to support one-hop query location based on the content identifier tag.

[0047] The location information includes at least one of the following: file identifier, page identifier, content unit identifier, character offset range, or rendering coordinate region; in one example, the location information may also include a file path or a file object reference.

[0048] The index structure can be at least one of a hash index, a B-tree index, or an inverted index.

[0049] When a tag is updated (e.g., rewritten due to drift / conflict, or supplemented with Version or validation fields), the associated storage unit 30 can synchronously update the mapping relationship to maintain consistency between the tag and the location information.

[0050] 2) Distributed metadata approach At least some field values ​​are stored in the metadata of the database object, file object, page object, and content unit object, respectively. For example, FileID is stored in the file object, PageID is stored in the page object, and the content unit identifier field is stored in the content unit object. During location, a step-by-step matching is performed based on the field hierarchy, so that the location range converges to the target document, target page, and target content unit in sequence.

[0051] 3) Combination method In practical systems, mapping table / index structures and distributed metadata methods can be used simultaneously: high-frequency location requests are prioritized for fast location through the index structure, and when the index is missing or expired, convergence location is achieved through step-by-step matching to improve robustness.

[0052] Implementation Method 5: Reference Link Encapsulation and Parsing with Multi-Strategy Location, Fallback, and Approximate Search Please see Figure 2 , Figure 2 In one embodiment of the present invention, the link encapsulation and parsing unit 40 is used to embed content identifiers into the path segment and / or query parameters of the link in an encoded manner to generate a reference link, and is used to parse and extract content identifiers from the reference link. The reference link is an application-private link or a general link. Preferably, the reference link may include a version field and / or a verification field to indicate parsing rules and integrity checks.

[0053] After receiving the referenced link, the target application can trigger the opening of the target unstructured document and trigger at least one of the following: jump to, scroll positioning, or highlighting of the target content unit.

[0054] In one implementation, the multi-strategy location and fallback retrieval engine 50 is used to locate unstructured document content units based on content identifier tags, wherein the location strategy includes at least one of the following or a combination thereof: 1) Step-by-step parsing and positioning strategy: Parse the tags according to the delimiter and field order or structured encoding rules, and locate them step by step in the order of file → page → content unit; 2) Index query positioning strategy: Use the tag as the key to query the mapping table or index structure to obtain positioning information or object references to achieve positioning; 3) Link parsing and location strategy: Parse reference links to extract tags; select the corresponding parsing rule when the version field is included; perform integrity verification before location when the verification field is included.

[0055] Preferably, the multi-strategy positioning and fallback retrieval engine 50 supports automatic fallback to another positioning strategy when any positioning strategy fails. For example, the index query positioning strategy is used first to obtain positioning information; when the index query fails or the index expires, the fallback uses a step-by-step parsing positioning strategy to converge the positioning scope to files and pages.

[0056] Furthermore, after locating the page-level range, the system can perform approximate retrieval by combining character offsets and / or feature information to complete the location. The feature information may include at least one of the following: text fragment summary hash of the target content unit, keyword set, context window text, or page layout area features, used to determine the target content unit within the page-level range, thereby improving the location success rate in cases of document structure changes, missing indexes, or expired indexes.

[0057] Implementation Method Six: Field Replacement, Rewriting, or Mapping in Cross-Device and / or Cross-User Scenarios (Optional) In one implementation, when a content identifier is transmitted across devices and / or across users and is parsed and located in the target environment, the tag encoding and uniqueness maintenance unit 20 can replace, rewrite, or map the DeviceID and / or UserID in the content identifier based on the device identifier and / or user identifier in the target environment to generate a content identifier suitable for the target environment, and the generated content identifier maintains parsability.

[0058] In one example implementation, the tag encoding and uniqueness maintenance unit 20 maintains a mapping table or index structure that maps source field values ​​to target field values. This mapping table or index structure records at least: a source device identifier and / or a source user identifier, a target device identifier and / or a target user identifier, and the corresponding file identifier field FileID. Thus, when the system receives a content identifier tag containing source field values, it can convert it to the target field value and then continue locating the target content unit.

[0059] It should be understood that the above replacement, rewriting or mapping are merely exemplary implementation methods. The specific replacement rules and mapping granularity may vary depending on the scenario configuration such as collaboration domain, permission domain or session domain, and should not be construed as a limitation of the present invention.

[0060] The above-disclosed embodiments are merely a few preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art will understand that implementing all or part of the above embodiments and making equivalent changes in accordance with the claims of the present invention are still within the scope of the invention.< / contentunitid> < / pageid> < / fileid> < / groupid> < / contentunitid> < / pageid> < / fileid>

Claims

1. A hierarchical identification system for unstructured document content, characterized by, Comprise: Field level configuration unit, mark coding and uniqueness maintenance unit, association storage unit, link packaging and parsing unit, and multi-strategy positioning and backoff retrieval engine; Wherein each unit interacts with data through direct connection or indirect connection via bus, middleware or interface service; The field level configuration unit is configured to configure the field set and field hierarchy relationship of the content identification mark in different application scenarios, and configure the preset field order and the preset delimiter rule or structured coding rule; wherein the field set at least includes a file identification field FileID and a content unit identification field; the content unit identification field is at least one of a line identification field, a paragraph identification field, a sentence identification field, a character offset field or a region identification field; the field hierarchy relationship at least includes FileID→content unit identification field; and the field set can optionally further include a page identification field PageID and / or a device identification field DeviceID and / or a user identification field UserID, and allows to be trimmed according to the scene to omit DeviceID and / or UserID; and can be extended to add cooperative or management entity fields, which are at least one of group identification field GroupID, project identification field, permission identification field or session identification field; The mark coding and uniqueness maintenance unit is connected with the field level configuration unit, and is configured to generate a content identification mark according to the field set, the field hierarchy relationship and the delimiter rule or structured coding rule; the generation includes: combining each field value in a preset order to form a mark, and escaping or encoding the field value to avoid ambiguity between the field value and the delimiter; the uniqueness maintenance includes: updating at least one field value when the document content is edited, rearranged, added, deleted or paginated, resulting in positioning drift and / or detecting mark conflict, and / or supplementing a version field Version and / or a check field in the mark to generate a new resolvable mark and maintain uniqueness; The association storage unit is connected with the mark coding and uniqueness maintenance unit, and is configured to establish and maintain a one-to-one correspondence between the content identification mark and the unstructured document content unit; wherein the correspondence is realized by one or a combination of the following ways: (1) maintaining a mapping table or index structure of "mark-positioning information", wherein the positioning information at least includes at least one of file identification, page identification, content unit identification, character offset range or rendering coordinate region; (2) storing at least part of the field values in the metadata of the database object, file object, page object and content unit object, to support matching and converging to the target content unit step by step according to the field hierarchy relationship; The link packaging and parsing unit is connected with the mark coding and uniqueness maintenance unit, and is configured to embed the content identification mark in the path segment and / or query parameter of the link in a coded manner to generate a reference link, and to parse and extract the content identification mark from the reference link; the reference link is an application private link or a general link; the reference link can include a version field and / or a check field, which are used to indicate the parsing rule and integrity check; wherein the reference link is used to trigger the opening of the target unstructured document after the target application receives the reference link, and trigger at least one of the jump, scroll positioning or highlight presentation of the target content unit; The multi-strategy positioning and fallback retrieval engine is connected with the association storage unit and the link packaging and parsing unit, and is configured to locate the unstructured document content unit according to the content identification mark, and the positioning at least includes one of the following strategies or a combination thereof: (a) hierarchical parsing positioning: parsing the mark according to the separator and field order or structured coding rule, and positioning at file→page→content unit level; (b) index query positioning: querying the mapping table or index structure with the mark as the key to obtain positioning information or object reference to realize positioning; (c) link parsing positioning: parsing the reference link to extract the mark, and selecting the corresponding parsing rule when the version field is included, and performing integrity check when the check field is included, and then positioning; And the multi-strategy positioning and fallback retrieval engine supports automatically falling back to another positioning strategy when a certain positioning strategy fails, and after positioning to the page level range, combining character offset and / or feature information to perform approximate retrieval to complete positioning.

2. The hierarchical identification system for unstructured document content according to claim 1, wherein The field hierarchy configuration unit is configured to include at least one of the following subordinate relationships in the field hierarchy relationship: FileID→content unit identification field; or FileID→PageID→content unit identification field; and when the field set includes a group identification field GroupID, it also includes the subordinate relationship of GroupID→FileID.

3. The hierarchical identification system for unstructured document content according to claim 1, wherein The structured coding rule includes at least one of URL coding, Base64 coding, JSON coding or CBOR coding; and the mark coding and uniqueness maintenance unit is configured to perform escape processing on the field value to avoid ambiguity caused by the inclusion of separators in the field value.

4. The hierarchical identification system for unstructured document content according to claim 1, wherein The mark code and the uniqueness maintenance unit update the content unit identification field and / or the page identification field when detecting positioning drift and / or mark conflict during the uniqueness maintenance; and / or supplement a version field in the content identification mark to indicate the field set and the parsing rule; and / or supplement a check field for integrity check of the field value combination result, the check field being at least one of checksum, digest hash, digital signature, random code or timestamp.

5. The hierarchical identification system for unstructured document content according to claim 1, wherein, The association storage unit, when using a mapping table or index structure of "mark-positioning information", the positioning information at least including file path or file object reference, page identification, character offset range and / or rendering coordinate area; and the index structure being at least one of hash index, B-tree index or inverted index; and the multi-strategy positioning and fallback retrieval engine, when performing the approximate retrieval, the feature information including at least one of text snippet digest hash, keyword set, context window text or layout area feature of the target content unit, for determining the target content unit after positioning to the page level range.

6. A hierarchical identification method for unstructured document content, applied to the hierarchical identification system for unstructured document content as claimed in claim 1, characterized in that, Comprising the following steps: S1. Configuring, by the field hierarchy configuration unit, the field set and field hierarchy relationship of the content identification mark for different application scenarios, and configuring the preset field order and the preset separator rule or structured encoding rule; S2. Generating, by the mark code and uniqueness maintenance unit, the content identification mark according to the field set, field hierarchy relationship and separator rule or structured encoding rule, and escaping or encoding the field values to avoid ambiguity; updating at least one field value when detecting document content changes and / or mark conflict, and / or supplementing a version field and / or a check field to generate a new unique content identification mark; S3. Establishing and maintaining, by the association storage unit, a one-to-one correspondence between the content identification mark and the target content unit, the one-to-one correspondence being realized by at least one of maintaining a mapping table or index structure of "mark-positioning information", and / or writing at least part of the field values into database object / file object / page object / content unit object metadata to support hierarchical matching; S4. Embedding, by the link packaging and parsing unit, the content identification mark in the path segment and / or query parameter of the link to generate a reference link, and parsing and extracting the content identification mark when receiving the reference link; S5. Positioning, by the multi-strategy positioning and fallback retrieval engine, the target content unit according to the content identification mark using hierarchical parsing positioning strategy and / or index query positioning strategy and / or link parsing positioning strategy, and automatically falling back to another positioning strategy when a certain positioning strategy fails, and performing approximate retrieval combining character offset and / or feature information after positioning to the page level range to complete the positioning.

7. The hierarchical identification method for unstructured document content according to claim 6, wherein, The generating or updating of the content identification mark in S2 includes: combining field values in a preset field order based on a field hierarchy relationship when the content identification mark is generated; and when the content identification mark is updated, at least one of the following operations is performed: updating a page identification field or a content unit identification field, supplementing a version field, supplementing a check field, or performing integrity checking on a field combination result, and other fields can be optionally updated according to specific scene requirements, so as to distinguish old marks and maintain parsability.

8. The hierarchical identification method for unstructured document content according to claim 6, characterized in that, In S5, when locating the target content unit, the index query positioning strategy is used to obtain positioning information; when the index query fails or the index is expired, the hierarchical parsing positioning strategy is used to converge the positioning range to a file and a page; and approximate retrieval is performed in a page-level range based on at least one of character offset information and / or a text summary hash, a keyword set, a context window text, or a layout area feature of the target content unit to determine the target content unit.

9. The hierarchical identification system for unstructured document content according to claim 1, characterized in that, When the content identification mark is transmitted in a cross-device and / or cross-user scenario and is parsed and positioned in a target environment, the mark encoding and uniqueness maintenance unit is configured to: replace, rewrite, or map a device identification field DeviceID and / or a user identification field UserID in the content identification mark according to a device identification and / or a user identification in the target environment, to generate a content identification mark suitable for the target environment, and the generated content identification mark maintains parsability.

10. The hierarchical identification system for unstructured document content according to claim 9, characterized in that, When the mark encoding and uniqueness maintenance unit performs the replacement, rewriting, or mapping, a mapping table or index structure of a corresponding relationship between a source field value and a target field value is maintained, and the mapping table or index structure at least records a source device identification and / or a source user identification, a target device identification and / or a target user identification, and a corresponding file identification field, so that when a content identification mark containing a source field value is received, the target field value can be converted to continue positioning the target content unit.

11. The hierarchical identification system for unstructured document content according to claim 1, characterized in that, The association storage unit establishes the one-to-one correspondence by maintaining a mapping table or index structure of a content identification mark and positioning information, wherein the content identification mark is stored as a key value and the positioning information is stored as a value, to support one-hop query positioning based on the content identification mark.

12. The hierarchical identification system for unstructured document content according to claim 1, characterized in that, The association storage unit establishes the one-to-one correspondence by storing the field values in the metadata of the database object, file object, page object and content unit object respectively, and the multi-strategy positioning and fallback retrieval engine performs step-by-step matching according to the field hierarchy relationship to gradually converge the positioning range to the target document, target page and target content unit.

13. The hierarchical identification system for unstructured document content according to claim 1, wherein, The link packaging and parsing unit is configured to embed the content identification mark in the path segment and / or query parameter of the application private link or general link, and include a version field and / or a verification field in the reference link; after the target application receives the reference link, the content identification mark is extracted by the link packaging and parsing unit, and the target content unit is positioned and presented by the multi-strategy positioning and fallback retrieval engine based on at least the index query positioning strategy and / or step-by-step resolution positioning strategy.

Images