A method and system for automatically extracting process data sets from life cycle assessment literature

Abstract

This application provides a method and system for automatically extracting process datasets from life cycle assessment (Life Cycle Assessment) literature, relating to the field of Life Cycle Assessment technology. The method includes: preprocessing Life Cycle Assessment literature fragments; identifying at least one candidate process based on the obtained cleaned text; generating a process dataset conforming to the TIDAS data pattern for each candidate process; validating the process datasets according to preset validation requirements; in response to successful validation of the process datasets, determining standard flow datasets matching the input and output items from an external standard Life Cycle Assessment database based on structured flow retrieval prompts; exporting process datasets carrying correlations or unmatched tags, along with the corresponding literature source datasets, into dataset files conforming to the TIDAS data pattern; performing TIDAS data pattern validation on the exported dataset files; and outputting the pattern validation results. Thus, process datasets meeting the pattern requirements are automatically extracted.

Description

Technical Field

[0001] This application relates to the field of life cycle assessment technology, and in particular to a method and system for automatically extracting process datasets for life cycle assessment literature. Background Technology

[0002] Life Cycle Assessment (LCA) typically requires the construction of a life cycle inventory (LCI), which includes physical inputs and outputs such as raw materials, energy, products, by-products, waste, and emissions in a process or unit operation. Currently, a large amount of LCI data is scattered across academic papers, technical reports, and appendices, presented in various formats including text descriptions, tables, image illustrations, and supplementary materials. To utilize this data for database construction or downstream evaluation, the unstructured or semi-structured content in the literature needs to be transformed into a standardized process dataset that conforms to the data schema requirements of the Tiangong Life Cycle Assessment Data System (TIDAS), can establish citation relationships with standard life cycle assessment databases, and can undergo schema validation. However, existing data extraction methods are often not tailored to the specific structure of life cycle assessment process datasets, making it difficult to directly obtain standardized process datasets that meet the above requirements based on the literature content by directly reusing existing technologies. Summary of the Invention

[0003] In view of this, this application provides a method and system for automatically extracting process datasets from life cycle assessment literature, which aims to automatically convert physical life cycle inventory data in the literature into process datasets that are verifiable, traceable and meet the requirements of the TIDAS data model.

[0004] According to a first aspect of the embodiments of this application, a method for automatically extracting process datasets for life cycle assessment literature is provided, comprising: Preprocessing of life cycle assessment literature fragments yields cleaned texts that include life cycle inventory content.

[0005] Based on the cleaning text, at least one candidate process is determined, and each candidate process corresponds to a set of physical lifecycle inventory data including input-side data and output-side data.

[0006] A process dataset conforming to the TIDAS data model is generated for each candidate process. The process dataset includes a set of physical lifecycle inventory input and output items and structured stream retrieval prompts for each input and output item. The set of physical lifecycle inventory input and output items includes at least one input item and at least one output item. The structured stream retrieval prompts are used to express the domain semantics of the corresponding input and output items for retrieval.

[0007] The process dataset is validated according to preset validation requirements.

[0008] In response to the successful verification of the process dataset, a standard flow dataset matching the input and output items is determined from an external standard lifecycle assessment database based on the structured flow retrieval prompt.

[0009] For successfully matched input / output items, the association between the standard stream dataset and the input / output items is constructed; for unmatched input / output items, a non-matching tag is generated.

[0010] The process dataset carrying the aforementioned association or the aforementioned unmatched tag, along with the corresponding literature source dataset, are exported as dataset files conforming to the TIDAS data pattern. The exported dataset files are then subjected to TIDAS data pattern validation, and the pattern validation results are output.

[0011] In some possible implementations, the preprocessing of life cycle assessment documents includes at least one of the following operations: filtering document fragments that do not carry life cycle inventory content based on paragraph characteristics and / or paragraph position; performing numerical artifact correction on non-table and non-image description content based on the format type of the document content; and converting characters into expressions in a preset standard format.

[0012] In some possible implementations, determining at least one candidate process based on the cleaning text includes: determining the candidate process based on a set of physical lifecycle inventory data lines belonging to the same process boundary in the cleaning text.

[0013] The physical lifecycle inventory data rows include data rows used to quantify the resource consumption, product output, and / or environmental emissions of the candidate process, and the physical lifecycle inventory data rows have identifiable physical units.

[0014] In some possible implementations, the candidate process includes at least one of the following: evidence information, used to record the original content information of the candidate process in the literature; and judgment basis information, used to determine the data availability of the candidate process in life cycle assessment modeling.

[0015] In some possible implementations, the process dataset may also include the name and unit information for each input / output item.

[0016] The verification of the process dataset according to the preset verification requirements includes at least one of the following verification contents: verifying whether the process dataset includes both input items and output items, and verifying whether the field contents of the input / output item names and the structured flow retrieval prompt information are complete; verifying whether the input / output item names, input / output directions and / or the unit information represent life cycle impact evaluation indicators.

[0017] In some possible implementations, the process dataset may also include process classification information.

[0018] The method further includes: classifying the process dataset corresponding to the candidate process step by step according to a preset process classification hierarchy.

[0019] The step of verifying the process dataset according to preset verification requirements includes: performing process classification verification on the process dataset.

[0020] In some possible implementations, the process dataset may also include geographic information and a reference year.

[0021] The step of generating a process dataset that conforms to the TIDAS data model requirements for each candidate process includes: determining a standardized description of the geographic information according to a preset geocoding method; and determining the reference year based on the data coverage year of the literature in response to the fact that the reference year is not explicitly given in the cleaned text.

[0022] The step of verifying the process dataset according to preset verification requirements includes: performing geographic information standardization verification on the process dataset.

[0023] In some possible implementations, determining the standard flow dataset matching the input and output items from an external standard lifecycle assessment database based on the structured flow retrieval prompts includes: Based on the stream type to which the input / output items belong, a stream retrieval query request is constructed using the input / output item names and / or the structured stream retrieval prompt information.

[0024] Based on the stream retrieval query request, a preset number of candidate stream datasets are determined from the external standard lifecycle assessment database.

[0025] In response to a successful match between the description information of the target candidate stream dataset and the input / output item, the target candidate stream dataset is determined to be the standard stream dataset.

[0026] In some possible implementations, the step of constructing the association between the standard stream dataset and the input / output item for a successfully matched input / output item includes: in response to a successful match, writing the identification information of the standard stream dataset into the stream reference data field of the input / output item.

[0027] The step of generating a non-matching flag for an input / output item that does not match includes: generating a non-matching record for the input / output item and / or associating the input / output item with a placeholder stream reference in response to the matching failure.

[0028] In some possible implementations, the method further includes: in response to the failure of the process dataset verification, regenerating the process dataset for the candidate process; in response to the number of times the process dataset is regenerated reaching a preset value, outputting the process dataset verification failure information for the candidate process, and associating the candidate process with a manual review marker.

[0029] According to a second aspect of the embodiments of this application, an automatic extraction system for process datasets of life cycle assessment literature is also provided, comprising: The document preprocessing module preprocesses the life cycle assessment document fragments to obtain cleaned text that includes the life cycle inventory content.

[0030] The process enumeration module is used to determine at least one candidate process based on the cleaning text, and each candidate process corresponds to a set of physical lifecycle inventory data including input-side data and output-side data.

[0031] The process dataset generation module is used to generate a process dataset that conforms to the TIDAS data pattern requirements for each candidate process. The process dataset includes a set of physical lifecycle inventory input and output items and structured stream retrieval prompts for each input and output item. The set of physical lifecycle inventory input and output items includes at least one input item and at least one output item. The structured stream retrieval prompts are used to express the domain semantics of the corresponding input and output items for retrieval.

[0032] The input / output item verification module is used to verify the process dataset according to preset verification requirements.

[0033] The flow matching module is used to respond to the successful verification of the process dataset, determine the standard flow dataset that matches the input and output items from the external standard lifecycle assessment database according to the structured flow retrieval prompts; and for successfully matched input and output items, construct the association relationship between the standard flow dataset and the input and output items, and for unmatched input and output items, generate a non-matching mark.

[0034] The dataset export and verification module is used to export the process dataset carrying the aforementioned correlation or the aforementioned unmatched marker, as well as the corresponding literature source dataset, into a dataset file that conforms to the TIDAS data pattern, and to perform TIDAS data pattern verification on the exported dataset file and output the pattern verification result.

[0035] According to a third aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described in the above embodiments.

[0036] According to a fourth aspect of the embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the above embodiments.

[0037] According to a fifth aspect of the embodiments of this application, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the method described in the above embodiments.

[0038] The technical solutions provided by the embodiments of this application may include the following beneficial effects: Preprocessing document fragments reduces the impact of non-lifecycle inventory content and identification noise on subsequent data extraction. Enumerating candidate processes and generating process datasets one by one allows multiple processes from the same document or document fragment to be processed separately, reducing the risk of process granularity confusion. Actively excluding lifecycle impact assessment indicators during process dataset generation reduces the risk of mistakenly writing lifecycle impact assessment indicators as physical lifecycle inventory data. Generating structured flow search prompts for each input and output item allows information such as the name, purpose, source, attributes, or synonyms of input and output items to participate in standard flow matching, improving matching accuracy. Timely verification, retry correction, and pattern verification after standard flow matching after process dataset generation result in better structural integrity, traceability, and verifiability of the final output. This enables the automatic extraction of verifiable, traceable process datasets that meet TIDAS data pattern requirements from lifecycle assessment literature.

[0039] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating an automatic extraction method for process datasets provided in an embodiment of this application; Figure 2This is an overall flowchart of an automatic extraction method provided in an embodiment of this application; Figure 3 This is a flowchart illustrating a process for generating and verifying a closed-loop data set, as provided in an embodiment of this application. Figure 4 This is a flowchart illustrating a standard stream matching, dataset export, and verification process provided in an embodiment of this application. Figure 5 This is a schematic diagram of the structure of an automatic process dataset extraction system provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0041] Currently, the process of extracting lifecycle inventory data and constructing standardized process datasets from lifecycle assessment literature typically relies on those skilled in the art reading each document, identifying processes, copying table data, standardizing units, matching standard flows in the database, and performing format validation. While this method offers high controllability, it is extremely inefficient and easily affected by document formatting issues, table spanning, character recognition errors, and inconsistencies in terminology.

[0042] Existing general document structure extraction methods can extract topics, entities, fields, or question-and-answer results from documents, but they are usually not geared to the specific structure of lifecycle assessment process datasets and are difficult to directly reuse in the extraction process of lifecycle inventory data.

[0043] For example, generic extraction methods struggle to distinguish between physical life cycle inventory inputs and outputs and Life Cycle Impact Assessment (LCIA) indicators. They are prone to misusing LCIA indicators such as global warming potential and acidification potential as physical life cycle inventory inputs and outputs, leading to unsatisfactory extraction results and requiring extensive manual review. When a single document contains multiple processes or system boundaries, generic extraction methods often merge inventory data from multiple processes, resulting in incorrect process granularity. During matching with standard flows, it is difficult to correctly align flows with similar names but different meanings to the standard life cycle assessment flow database. Furthermore, data output from generic extraction methods is prone to issues such as missing data fields, incomplete citations, and non-standard geographic information, failing to meet TIDAS data model constraints and rendering the extracted results unusable in subsequent modeling. Additionally, when extracting information from documents, numerical values, physical units, chemical formulas, years, or serial numbers are highly susceptible to numerical artifacts. These are crucial information in process datasets, and process datasets determined using generic extraction methods have a high probability of errors, resulting in low reliability.

[0044] In the field of life cycle assessment technology, current life cycle process inventory calculation or automatic life cycle assessment modeling schemes can convert documents, reports or image data into string fragments, lightweight markup language (Markdown) or vector retrieval objects, and use large language models to generate life cycle process inventories, inventory frameworks, HTML tables, life cycle assessment model frameworks or populate inventory data, but they still cannot solve the above problems.

[0045] In view of this, this application provides an automatic extraction method for process datasets from life cycle assessment literature, which aims to automatically convert physical life cycle inventory data in the literature into process datasets that are verifiable, traceable and meet the requirements of the TIDAS data model.

[0046] like Figure 1 As shown, Figure 1 This is a flowchart illustrating a method for automatically extracting process datasets according to an embodiment of this application. The method includes the following steps S101 to S107: Step S101: Preprocess the life cycle assessment literature fragments to obtain cleaned text including life cycle inventory content.

[0047] For example, life cycle assessment literature excerpts can come from full-text paper analysis results, supplementary material analysis results, PDF file conversion results, or manually compiled JSON or Markdown files, such as JSON files converted from manually compiled tables. It is understood that the data structure of these literature excerpts is machine-readable, including but not limited to JSON or Markdown formats.

[0048] A bibliographic fragment may include at least one of the following: text fragment, table fragment, and image caption fragment. For example, a bibliographic fragment may include multiple entries, each with a type field and a content field. The type field is used to indicate text, tables, image captions, titles, references, or other paragraphs, while the content field stores the corresponding text.

[0049] In some possible implementations, the preprocessing of life cycle assessment documents includes at least one of the following operations: filtering document fragments that do not carry life cycle inventory content based on paragraph characteristics and / or paragraph position; performing numerical artifact correction on non-table and non-image description content based on the format type of the document content; and converting characters into expressions in a preset standard format.

[0050] For example, document fragments can be read sequentially, and the main text, tables, and image descriptions can be merged into candidate texts. These candidate texts correspond to the sections of the document that contain lifecycle inventory content. For bibliographic metadata, copyright notices, references, acknowledgments, and sections containing only a list of citations, filtering can be performed based on paragraph position, headings, or regular expression features. In other words, document fragments that do not contain lifecycle inventory content are removed to reduce the workload of subsequent data processing.

[0051] During preprocessing, characters can also be standardized by converting them into a preset standard format to achieve standardized representation. For example, full-width characters can be converted to half-width characters, and compatible characters can be standardized into a unified form.

[0052] To address numerical artifacts caused by Optical Character Recognition (OCR), corrections can be performed in non-table and non-image caption contexts. For example, incorrectly space-separated decimals, exponents, or numbers near units can be corrected to their normal representation. For table fragments and image caption fragments, their original structural and contextual markers can be preserved, avoiding damage to column alignment or graphic semantics due to over-correction, and preventing disruption of table or image caption structure.

[0053] The above preprocessing can reduce the impact of non-list content and identification noise on data extraction, enabling subsequent extraction steps to obtain inputs that are closer to the actual technical content, which is beneficial to improving the accuracy and efficiency of data extraction.

[0054] Step S102: Determine at least one candidate process based on the cleaning text, and each candidate process corresponds to a set of physical lifecycle inventory data including input-side data and output-side data.

[0055] It is understandable that a single document or document fragment may contain multiple candidate process steps. This application employs an enumeration approach, separating at least one candidate process step and its corresponding physical lifecycle inventory data area from other candidate process steps. In subsequent processing, this allows multiple process steps from the same document or document fragment to be processed separately, thereby reducing the risk of process granularity confusion.

[0056] In some possible implementations, this step includes: determining the candidate process based on a set of physical lifecycle inventory data rows belonging to the same process boundary in the cleaning text.

[0057] The Life Cycle Inventory (LCI) data rows refer to data rows used to quantify the resource consumption, product output, and / or environmental emissions of the candidate process. Each LCI data row has an identifiable physical unit. A set of LCI data rows can collectively characterize both the input and output sides of the candidate process. For example, the set of LCI data rows includes at least one input data row representing resource consumption or material input, and at least one output data row representing product output, waste stream output, or environmental emissions.

[0058] For example, the candidate process is a process or unit operation that has a physical lifecycle inventory (LCI) data line in the cleaned text. In the cleaned text, content related to process name, process boundary, functional unit, input / output table, emission table, or energy consumption table can be filtered and identified. If a process or unit operation has at least one physical LCI data line in the literature, such as physical input, physical output, emission, energy consumption, or product flow, and has an identifiable physical unit, it can be considered a candidate process. In some possible implementations, when determining a candidate process, a process identifier can also be assigned to the candidate process, and the source paragraph or table of the candidate process can be saved as evidence based on the corresponding content in the cleaned text for subsequent verification.

[0059] Regarding life cycle assessment literature, it is important to clearly distinguish between physical life cycle inventory data and life cycle impact assessment indicators to avoid mistakenly extracting life cycle impact assessment indicators as process flow data. For example, in the step of determining candidate processes, this application also adopts an exclusion method. If a paragraph or table only contains content related to impact assessment indicators such as global warming potential, acidification potential, eutrophication potential, ozone depletion potential, ecotoxicity, or human toxicity, it will not be identified as a candidate process.

[0060] Step S103: Generate a process dataset that conforms to the TIDAS data model requirements for each candidate process. The process dataset includes a set of physical lifecycle inventory input and output items and structured stream retrieval hints for each input and output item. The set of physical lifecycle inventory input and output items includes at least one input item and at least one output item. The structured stream retrieval hints are used to express the domain semantics for retrieval of the corresponding input and output items.

[0061] That is, the input and output items include input items and output items. The process dataset is a set of physical lifecycle list input and output items, including at least one input item and at least one output item. Each input item or each output item corresponds to a structured flow retrieval prompt.

[0062] The TIDAS data schema requirements in this application are used to constrain the fields, hierarchy, referential relationships, and verifiability of the process dataset.

[0063] For example, for each identified candidate process, structured data processing is performed to generate a process dataset with a preset data structure based on the relevant content of the corresponding physical LCI data row.

[0064] For example, the TIDAS data structure of a process dataset includes process identification information, process classification information, geographical information, reference year, reference flow, and a set of physical lifecycle inventory input and output items. This set of physical lifecycle inventory input and output items can be represented as a list of input and output items, and includes at least input items and output items. Input items may include raw materials, energy, auxiliary materials, or waste streams as input; output items may include product streams, by-product streams, waste streams as output, basic environmental emission streams, or reference streams used to characterize service / treatment objects. Each input / output item includes the input / output item name, input / output direction, value, unit, and structured flow retrieval prompts. The above data structure is for illustrative purposes only, and the specific details can be determined according to the TIDAS data model requirements.

[0065] The names, directions, values, and units of the input and output items can be determined based on the corresponding physical LCI data rows, and supplemented with text content for description or error correction. During the generation of this information, structured stream retrieval suggestions can be generated.

[0066] The structured stream retrieval hints described here are used to express the domain semantics of the input / output item for retrieval. For example, structured stream retrieval hints can be represented as dictionaries, JSON objects, or key-value pair text, and serialized and written into the annotation fields of the corresponding input / output items to provide richer semantic information than the input / output item names. The annotation fields contain a fixed prefix for stream retrieval and multiple key-value pairs. These key-value pairs can include one or more of the following: base name, processing method, hybrid or geographic location, source path, English synonyms, state purity, stream attributes, and usage context, to enrich the semantic information from different dimensions. This key-value pair text is only used to express the field structure and does not represent specific experimental data or specific literature conclusions.

[0067] For example, structured stream search suggestions may include one or more of the following fields: The basename field is used to express the name of the main substance, energy source, or product after removing modifiers.

[0068] The treatment field is used to express semantics related to processes such as incineration, landfill, recycling, market mixing, production, and transportation.

[0069] The mix_location field, which represents a mixed or geographic location, is used to express a regional market, country, or region.

[0070] The flow_properties field is used to express attributes such as quality, energy, volume, and quantity.

[0071] The `synonyms_en` field is used to record English synonyms.

[0072] The `synonyms_zh` field is used to record Chinese synonyms.

[0073] The state_purity field records state and purity information.

[0074] The source_or_pathway field is used to record source or path information.

[0075] The usage_context field is used to record the usage context in the current process.

[0076] Based on the above field content, this structured stream retrieval suggestion can provide richer semantic information than the input and output item names, which is beneficial to improving the retrieval accuracy in subsequent searches of standard stream databases.

[0077] Step S104: Verify the process dataset according to the preset verification requirements.

[0078] In other words, after generating a process dataset for each candidate process, the validity of the process dataset is also verified to ensure that the aforementioned steps effectively generate the process dataset and avoid wasting subsequent processing resources.

[0079] For example, as described above, a valid process dataset should include a set of physical lifecycle inventory inputs and outputs that characterize the input and output sides of the candidate process. Each set of inputs and outputs includes at least one input and at least one output. A valid process dataset should also include the name, input / output direction, value, and unit of each input / output. This step validates the inputs and outputs of the process dataset for each candidate process to ensure that the process dataset meets the TIDAS data schema requirements. Similarly, other information can be validated in this step, ensuring that the necessary fields required by the TIDAS data schema are valid before proceeding to subsequent processing steps.

[0080] Step S105: In response to the successful verification of the process dataset, a standard flow dataset matching the input and output items is determined from an external standard lifecycle assessment database based on the structured flow retrieval prompt.

[0081] As mentioned above, structured flow retrieval hints are used to express the domain semantics of the input / output items for retrieval, providing richer semantic information than just the names of the input / output items. If the validation passes, meaning the process dataset is valid and meets the TIDAS data schema requirements, a query is performed from the external standard lifecycle assessment database based on the aforementioned structured flow retrieval hints.

[0082] For example, the input / output item list in the process dataset is traversed. For each input / output item, an external standard lifecycle assessment database is retrieved based on the structured flow retrieval prompts to query candidate flow datasets. Then, a comprehensive judgment is made based on information such as the name, category, short description, geographical information, and flow attributes of the candidate flow datasets to determine the standard flow dataset that matches the input / output item from the candidate flow datasets.

[0083] In some possible implementations, the method further includes: in response to the failure of the process dataset verification, regenerating the process dataset for the candidate process; in response to the number of times the process dataset is regenerated reaching a preset value, outputting the process dataset verification failure information for the candidate process, and associating the candidate process with a manual review marker.

[0084] In other words, if the validity check of the process dataset fails, the error result can be used as feedback information to return to step S103 to regenerate the process dataset and complete the retry correction until the validation passes or the preset number of retries is reached. The preset number of retries can be set to at least one, such as two or three, and is not limited here. If the validation still fails after the preset number of retries, the subsequent processing flow is stopped, and the process dataset validation failure information is output for manual review or rejection.

[0085] Step S106: For successfully matched input / output items, construct the association between the standard stream dataset and the input / output items; for unmatched input / output items, generate unmatched tags.

[0086] In conjunction with step S105 above, if a standard stream dataset matching the input / output item can be determined from the candidate stream dataset, the input / output item is associated with the matching standard stream dataset for subsequent tracking, thereby ensuring that the final extracted process dataset meets the requirements of being traceable and verifiable.

[0087] For example, the data structure of the input / output item includes a stream reference data field for recording the stream reference status, supporting the traceability of the process dataset. The step of constructing the association between the standard stream dataset and the input / output item includes: in response to a successful match, writing the identification information of the standard stream dataset into the stream reference data field of the input / output item.

[0088] For example, one or more of the information such as the identifier, version, Uniform Resource Identifier, or short description of the matched standard stream dataset can be written into the stream reference data field. This allows natural language input and output items extracted from the literature to be linked to the normative stream object in the external standard lifecycle assessment database, which is beneficial for subsequent data import, comparison, tracing, and pattern reference verification operations.

[0089] In some possible implementations, the correspondence between input / output items and the original literature sources is also preserved, thereby preserving the chained correspondence between the original literature, the process dataset, and the standard flow dataset.

[0090] If none of the candidate stream datasets returned from the external standard lifecycle assessment database can match the input / output item, a non-matching tag is generated for that input / output item for subsequent manual review or new stream creation.

[0091] For example, generating a non-matching flag for an input / output item that did not match successfully includes: generating a non-matching record for the input / output item and / or associating the input / output item with a placeholder stream reference in response to the matching failure.

[0092] For example, an additional unmatched record can be generated to record input and output items that did not match the standard stream dataset. This unmatched record can exist as a standalone record, a list, or a file, and can retain the name, direction, value, unit, structured stream retrieval hints, the stream dataset to which it belongs, and source information (such as document fragment identifiers or specific paragraphs or figures in the document).

[0093] Alternatively, an unmatched state can be recorded in the stream reference data field of the input / output item. This serves as a temporary reference to mark a state awaiting review or a state awaiting stream creation. For example, a placeholder stream object or a local stream object can be generated in this stream reference data field to establish an association between the input / output item and the placeholder stream reference. To avoid creating unresolved dangling references, placeholder stream references can be expressed using annotation fields, extended fields, or local placeholder stream objects allowed by the TIDAS data schema.

[0094] Step S107: Export the process dataset carrying the aforementioned association or the unmatched tag, along with the corresponding literature source dataset, into a dataset file conforming to the TIDAS data pattern. Perform TIDAS data pattern verification on the exported dataset file and output the pattern verification result.

[0095] For example, the exported dataset files include process dataset files and source dataset files. For instance, process datasets can be exported to a process dataset directory, and literature source datasets can be exported to a source dataset directory. The exported datasets can be linked using identifiers, versions, and Uniform Resource Identifiers. The exported directory is validated using the TIDAS schema validator; the validation result can serve as the delivery status for this round of automatic extraction or indicate subsequent manual revisions or regeneration operations.

[0096] In conjunction with steps S105-S106 described above, in the event of input / output item mismatch, the exported dataset file in step S107 also includes the stream dataset file corresponding to the unmatched marker. For example, if a separate unmatched record is generated for an unmatched input / output item, then the list or file of unmatched records is exported. Alternatively, if a placeholder stream object or local stream object is generated in the stream reference data field, then its identifier, version, or reference path, etc., are exported along with it.

[0097] Based on the exported dataset file, the final output pattern validation result can include various information such as pattern errors, warnings, a list of unmatched input / output items, and the export path.

[0098] For example, if the TIDAS schema validator returns errors such as missing fields, invalid enumeration values, incorrect reference paths, or schema inconsistencies, these errors can be written into the schema validation results and output to indicate subsequent manual revisions. If the validator does not return any errors, meaning the validation result is successful, it indicates that the process dataset meets the TIDAS data schema requirements, and the process dataset can be delivered to complete this round of automatic extraction tasks.

[0099] The pattern validator here can be a local command-line program, a service interface, or an embedded validation library. As long as it can perform TIDAS pattern validation on the process dataset, source dataset, and stream dataset in the exported directory when local placeholder stream objects exist, and return errors, warnings, and path information, it can be used in this embodiment.

[0100] Figure 2 This is an overall flowchart of an automatic extraction method provided in an embodiment of this application.

[0101] As shown in steps S201-S211, firstly, LCA literature fragments are obtained, and then preprocessed. Next, all candidate processes are enumerated, and process datasets are generated one by one. Then, input / output items and pattern constraints are validated to ensure that the generated process datasets are valid and meet the pattern constraint requirements. If the validation fails, the process returns to step S203 and repeats the above process. If the validation passes, a flow retrieval query request is constructed to match the standard flow dataset, thereby writing a flow reference relationship (i.e., the constructed association relationship with the standard flow dataset) or writing a non-matching tag to the process dataset. Afterward, the process dataset, source dataset, and non-matching tags (if any) are exported, pattern validation is performed, and the final pattern validation result is output.

[0102] In other words, during the automatic extraction and verification of the process dataset provided in this application embodiment, preprocessing the document fragments can reduce the impact of non-lifecycle inventory content and identification noise on subsequent data extraction. By enumerating candidate processes and generating process datasets one by one, multiple processes in the same document or document fragment can be processed separately, thereby reducing the risk of process granularity confusion. Furthermore, lifecycle impact assessment indicators are actively excluded during the generation of process datasets, reducing the risk of mistakenly writing lifecycle impact assessment indicators as physical lifecycle inventory data. By generating structured flow search prompts for each input and output item, information such as the name, purpose, source, attributes, or synonyms of the input and output items can participate in standard flow matching, which helps to improve matching accuracy. Timely verification, retry correction, and pattern verification after standard flow matching after the process dataset is generated can make the final output result have better structural integrity, traceability, and verifiability, ensuring that the process dataset automatically extracted based on lifecycle assessment documents is verifiable, traceable, and meets TIDAS pattern requirements.

[0103] The following section, in conjunction with steps S102 to S105 above, further explains the closed-loop process of determining candidate processes, generating process datasets, and verifying them. Figure 3 This is a schematic diagram of a process for generating and verifying a closed-loop data set, as provided in an embodiment of this application.

[0104] In conjunction with the foregoing, for the cleaned text, as shown in steps S301 to S303, processes including physical LCI data are first identified, and LCIA index content is excluded to generate a list of candidate process processes corresponding to the physical lifecycle inventory data.

[0105] In some possible implementations, the candidate process includes at least one of the following: evidence information, used to record the original content information of the candidate process in the literature; and judgment basis information, used to determine the data availability of the candidate process in life cycle assessment modeling.

[0106] For example, in step S303 above, a unique process identifier, process name, evidence information, and judgment basis information are generated for each candidate process. The evidence information may record relevant paragraphs or tables as support for subsequent verification. The judgment basis information records whether the necessary information for the process exists. For example, the prerequisite for data availability in life cycle assessment modeling of a candidate process is the existence of input / output tables, physical units, functional units, and process boundary descriptions; therefore, the judgment basis information here may record whether such information exists.

[0107] As shown in steps S304 and S305, in this embodiment of the application, a process dataset is generated one by one, focusing on a single candidate process. When generating the process dataset, only paragraphs, tables, and image descriptions related to that candidate process are used; input and output items from other candidate processes are not merged into the current process dataset. Based on this, when describing multiple production routes, multiple processing methods, or multiple regional system boundaries in the same document or document fragment, the granularity of the process dataset can be maintained.

[0108] In conjunction with the foregoing, in some possible implementations, the process dataset may also include the name, input / output direction, and unit information of each input / output item. During the step of validating the process dataset according to preset validation requirements, as shown in step S306, validation will be performed on the input / output item names, units, and structured flow retrieval prompt information.

[0109] For example, in this step, an input / output item integrity check can be performed, that is, to check whether the process dataset includes both input and output items, and to check whether the field content of the input / output item names and the structured stream retrieval prompt information is complete.

[0110] For example, validation ensures that input / output item names are not empty and cannot be placeholder text such as "unknown," "placeholder," or "TBD." As mentioned earlier, when generating the input / output item information for the process dataset, structured flow retrieval hints are generated simultaneously. This validation step also checks for missing, null, placeholder, or obviously inconsistent fields in the structured flow retrieval hints, including base name, processing method, mixed or geographic location, source path, English synonyms, state purity, flow attributes, and usage context. Furthermore, it checks for conflicts between mixed location, source path, and geographic information. This step also verifies whether the process dataset meets the necessary fields required by the TIDAS schema.

[0111] Alternatively, in this step, one can start from the content level to perform life cycle impact assessment indicator exclusion verification to determine whether the data in the process dataset corresponds to the physical life cycle inventory data, but not to the life cycle impact assessment indicators.

[0112] For example, the system verifies whether the input / output item name, input / output direction, and / or the unit information characterizes a life cycle impact assessment indicator. When the input / output item name or unit corresponds to at least one of the impact assessment indicators such as global warming potential, acidification potential, eutrophication potential, ozone depletion potential, ecotoxicity, or human toxicity, the input / output item is determined not to belong to the physical life cycle inventory. Therefore, this input / output item can be reported as an error, and the system requests the regeneration of the input / output item list for the current candidate process. For instance, if the unit for the input / output item of the global warming potential indicator is kg CO2-eq, and the verification in this step determines that the process dataset includes this unit, then this input / output item is reported as an error, and the system requests the regeneration of the input / output item list for the current candidate process.

[0113] By eliminating candidates during the process identification phase and verifying the process dataset in a timely manner, the problem of mistakenly writing lifecycle impact assessment indicators as physical lifecycle inventory data can be effectively avoided.

[0114] When verification fails, the error is reported to the system program in the process dataset generation step to regenerate the dataset for the corresponding candidate process, until verification passes or the preset retry conditions are met. The number of retries can be limited by system configuration parameters. If the process still fails after meeting the preset retry conditions, an error report is output and the candidate process is marked as requiring manual review.

[0115] In conjunction with the foregoing, in some possible implementations, the process dataset may also include process classification information.

[0116] The method further includes: classifying the process dataset corresponding to the candidate process step by step according to a preset process classification hierarchy.

[0117] The step of verifying the process dataset according to preset verification requirements includes: performing process classification verification on the process dataset.

[0118] Similarly, if the process classification verification fails, the classification process can be repeated. This closed-loop process of hierarchical classification and process classification verification improves the discoverability, comparability, and trustworthiness of the process dataset, facilitating automatic process matching during subsequent modeling.

[0119] In some possible implementations, the process dataset may also include geographic information and a reference year.

[0120] The step of generating a process dataset that conforms to the TIDAS data model requirements for each candidate process includes: determining a standardized description of the geographic information according to a preset geocoding method; and determining the reference year based on the data coverage year of the literature in response to the fact that the reference year is not explicitly given in the cleaned text.

[0121] The step of verifying the process dataset according to preset verification requirements includes: performing geographic information standardization verification on the process dataset.

[0122] In other words, based on a preset geocoding method, different ways of expressing geographic information in document content can be unified and described in a standardized manner. If the geographic information standardization verification fails, the corresponding standardized description can be redefined. Through a closed-loop generation and verification process, the consistency of the final process dataset is ensured.

[0123] When determining the reference year, if the literature explicitly provides a reference year for the process dataset, then that year should be used. In some literature, the reference year for each candidate process may not be explicitly displayed. When the reference year is not explicitly provided in the literature, it can be extracted from the preset character window at the beginning of the literature as a candidate reference year. For example, search for four-digit years from 1900 to the year following the current year within the first 5000 characters of the cleaned text, or use the identified publication year or data overlay year as a candidate reference year.

[0124] Next, we will further explain the process of standard stream matching, dataset export, and pattern verification in steps S106 to S107 described above. Figure 4 This is a flowchart illustrating a standard stream matching, dataset export, and verification process provided in an embodiment of this application.

[0125] In some possible implementations, determining the standard stream dataset matching the input and output items from an external standard lifecycle assessment database based on the structured stream retrieval prompts includes the following steps: First, as shown in steps S401 to S403, a stream retrieval query request is constructed based on the stream type to which the input / output item belongs, and on the input / output item name and / or the structured stream retrieval prompt information.

[0126] Secondly, as shown in steps S404 to S405, a preset number of candidate stream datasets are determined from the external standard life cycle assessment database according to the stream retrieval query request.

[0127] Finally, as shown in step S406, it is determined whether the returned candidate stream dataset can successfully match the input / output item. In response to the description information of the target candidate stream dataset successfully matching the input / output item, the target candidate stream dataset is determined to be the standard stream dataset.

[0128] For example, the flow types to which the input / output items belong include product flows, basic flows, and waste flows. When the input / output item is a product flow or a waste flow, the flow retrieval query request constructed in step S403 may include one or more of the following: basic name, processing method, product use, input / output direction, and regional information. For example, for a certain input material, the query request includes not only the material name but also its context as a raw material, fuel, solvent, or auxiliary material; for a waste flow as input or output, the query request may also include processing method or destination information. Accordingly, after the standard LCA flow database returns the candidate flow dataset, the matching item can be determined from the candidates based on their short description, flow attributes, geographical information, and category, that is, the matching standard flow dataset can be determined.

[0129] When the input / output items are based on a flow, the flow retrieval query request constructed in step S403 can further utilize environmental compartment or chemical identification information to construct the retrieval query based on one or more of the input / output item name, chemical registry number, environmental medium, and environmental compartment. For example, for air emissions, water emissions, or soil emissions, the query request includes the emission medium and environmental compartment to avoid confusion between standard flows of the same substance in different environmental compartments.

[0130] As shown in steps S407 and S408, if a match is successful, a standard flow reference is written for the input / output item. If no matching standard flow dataset is found, an unmatched record or placeholder flow reference is generated, i.e., an unmatched identifier is generated. An unmatched record includes at least the original input / output item name, direction, value, unit, structured flow search hint, the associated process dataset, and the location of the source document. This unmatched record can be used for subsequent manual review or to create a new standard flow dataset. A placeholder flow reference keeps the process dataset exportable and marks it as unmatched in the schema validation results.

[0131] The dataset export and pattern verification steps S409 to S410 are as described above and will not be repeated here.

[0132] In some possible implementations, process parameters can be set through system configuration files or environment variables to ensure the reliable progress of the automatic extraction process. For example, a maximum number of retries can be set for a standard stream retrieval query request, such as three. If the candidate stream dataset cannot be returned after exceeding this number, an unmatched tag can be directly generated. The request timeout can be set to 30 seconds, and the retry backoff interval can be set to 0.5 seconds. The process concurrency can be set according to the computing device resources, for example, to a maximum of four concurrent tasks. The concurrency of standard stream retrieval can be set to one or more concurrent tasks.

[0133] Combination Figure 4 As shown, for each input / output item, one or more stream retrieval query texts can be constructed. For example, no more than eight deduplicated query texts can be generated for the same input / output item, and the name, short description, stream attributes, geographic information, and classification of the returned candidate stream datasets can be used for matching judgment.

[0134] Those skilled in the art can adjust the values ​​according to the size of the literature, the response speed of the remote database, computing resources, and the rigor of verification, without changing the technical concept of this application, which forms a closed loop through literature preprocessing, candidate process enumeration, process-by-process dataset generation, input / output item verification, standard flow matching, and pattern verification.

[0135] Figure 5 This is a schematic diagram of the structure of an automatic process dataset extraction system provided in an embodiment of this application. The system includes a document preprocessing module 51, a process enumeration module 52, a process dataset generation module 53, an input / output item verification module 54, a flow matching module 55, and a dataset export verification module 56.

[0136] The document preprocessing module 51 is used to preprocess the life cycle assessment document fragments to obtain cleaned text including the life cycle inventory content.

[0137] The process enumeration module 52 is used to determine at least one candidate process based on the cleaning text, and each candidate process corresponds to a set of physical lifecycle inventory data including input-side data and output-side data.

[0138] The process dataset generation module 53 is used to generate a process dataset that conforms to the TIDAS data pattern requirements for each candidate process. The process dataset includes a set of physical lifecycle inventory input and output items and structured stream retrieval hints for each input and output item. The set of physical lifecycle inventory input and output items includes at least one input item and at least one output item. The structured stream retrieval hints are used to express the domain semantics for retrieval of the corresponding input and output items.

[0139] The input / output item verification module 54 is used to verify the process dataset according to preset verification requirements.

[0140] The flow matching module 55 is used to respond to the successful verification of the process dataset, determine the standard flow dataset that matches the input and output items from the external standard life cycle assessment database according to the structured flow retrieval prompt; and for the successfully matched input and output items, construct the association relationship between the standard flow dataset and the input and output items, and for the unmatched input and output items, generate a non-matching mark.

[0141] The dataset export verification module 56 is used to export the process dataset carrying the aforementioned correlation or carrying the unmatched marker, as well as the corresponding literature source dataset, into a dataset file that conforms to the TIDAS data pattern, and to perform TIDAS data pattern verification on the exported dataset file and output the pattern verification result.

[0142] This system reduces the impact of non-lifecycle inventory content and identification noise on subsequent data extraction by preprocessing document fragments. By enumerating candidate processes and generating process datasets one by one, multiple processes within the same document or document fragment can be processed separately, reducing the risk of process granularity confusion. Actively excluding lifecycle impact assessment indicators during process dataset generation reduces the risk of mistakenly writing lifecycle impact assessment indicators as physical lifecycle inventory data. Generating structured flow search prompts for each input and output item allows information such as the name, purpose, source, attributes, or synonyms of input and output items to participate in standard flow matching, improving matching accuracy. Timely verification, retry correction, and pattern verification after standard flow matching after process dataset generation result in a final output with better structural integrity, traceability, and verifiability. This enables the automatic extraction of verifiable, traceable process datasets that meet TIDAS pattern requirements from lifecycle assessment literature.

[0143] For example, based on this system, an English full text about the life cycle evaluation of a 180ml coffee cup, along with its image descriptions and table content, is taken as input. The document preprocessing module 51 preprocesses the input document fragment to generate clean text, and the process enumeration module 52 can identify multiple candidate processes from the clean text.

[0144] The process dataset generation module 53 generates the following six process datasets: usage process of an edible coffee cup (180ml), usage process of a polylactic acid disposable coffee cup (180ml), usage process of a polyethylene paper-lined coffee cup (180ml), usage process of a expanded polystyrene disposable coffee cup (180ml), single-use process of a reusable polypropylene coffee cup, and single-use process of a reusable stainless steel coffee cup. These six process datasets correspond to different cup types or usage scenarios from the same literature and are not combined into a single master list.

[0145] The aforementioned six process datasets comprise a total of 63 input / output items. Some of these items are inputs, corresponding to input-side data, while others are outputs, corresponding to output-side data. Each input / output item includes an input / output item name, input / output direction, value, unit, and a structured stream retrieval hint. The structured stream retrieval hint includes a base name, processing method, mixed or geographical location, source path, English synonyms, state purity, stream attributes, and usage context fields. The stream matching module 55 performs standard stream matching based on the input / output item names and structured stream retrieval hints, and writes a reference to the standard stream dataset into the input / output items.

[0146] The dataset export verification module 56 exports six process dataset files and one source dataset file, and performs schema verification on the exported files. The verification results show that all the process dataset files and the source dataset file passed the schema verification.

[0147] In other words, the system can convert multiple processes in the same LCA document into dataset files with input / output items, structured stream search hints, stream citations, and source citations.

[0148] In an exemplary embodiment, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in any of the above embodiments.

[0149] For the device embodiments and computer program product embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. Furthermore, the device embodiments described above are illustrative; the modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, i.e., they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without any inventive effort.

[0150] This application can be implemented by executing several computer program code flows using an electronic device. The electronic device loads the computer program into non-volatile memory and uses the processor of the electronic device to read these computer program instructions into memory for execution.

[0151] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The following is in conjunction with... Figure 6 The electronic devices described in some embodiments of this application will be explained.

[0152] Reference Figure 6 Electronic devices 600 can be mobile phones, computers, digital broadcasting terminals, messaging devices, game consoles, tablets, medical devices, fitness equipment, personal digital assistants, etc.

[0153] Device 600 may include one or more of the following components: processing component 601, memory 602, power supply component 603, multimedia component 604, audio component 605, input / output (I / O) interface 606, sensor component 607, and communication component 608.

[0154] Processing component 601 typically controls the overall operation of device 600, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 601 may include one or more processors 609 to execute instructions to perform all or part of the steps of the methods described above. Furthermore, processing component 601 may include one or more modules to facilitate interaction between processing component 601 and other components. For example, processing component 601 may include a multimedia module to facilitate interaction between multimedia component 604 and processing component 601.

[0155] Memory 602 is configured to store various types of data to support the operation of device 600. Examples of this data include instructions for any application or method operating on device 600, contact data, phonebook data, messages, pictures, videos, etc. Memory 602 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0156] Power supply component 603 provides power to various components of device 600. Power supply component 603 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to device 600.

[0157] Multimedia component 604 includes a screen that provides an output interface between the device 600 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 604 includes a front-facing camera and / or a rear-facing camera. When the device 600 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0158] Audio component 605 is configured to output and / or input audio signals. For example, audio component 605 includes a microphone (MIC) configured to receive external audio signals when device 600 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 602 or transmitted via communication component 608. In some embodiments, audio component 605 also includes a speaker for outputting audio signals.

[0159] Input / output (I / O) interface 606 provides an interface between processing component 601 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0160] Sensor assembly 607 includes one or more sensors for providing state assessments of various aspects of device 600. For example, sensor assembly 607 may detect the on / off state of device 600, the relative positioning of components such as the display and keypad of device 600, changes in the position of device 600 or a component of device 600, the presence or absence of user contact with device 600, the orientation or acceleration / deceleration of device 600, and temperature changes of device 600. Sensor assembly 607 may also include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 607 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 607 may also include an accelerometer, a gyroscope, a magnetometer, a pressure sensor, or a temperature sensor.

[0161] Communication component 608 is configured to facilitate wired or wireless communication between device 600 and other devices. Device 600 can access wireless networks based on communication standards, such as WiFi, 2G or 3G, 4G or 5G, or combinations thereof. In one exemplary embodiment, communication component 608 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 608 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0162] In an exemplary embodiment, device 600 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components.

[0163] In an exemplary embodiment, this application also provides a non-transitory computer-readable storage medium including instructions, such as a memory 602 including instructions that can be executed by a processor 609 of a device 600. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

Claims

1. A method for automatically extracting process datasets from life cycle assessment literature, characterized in that, include: Preprocess the fragments of life cycle assessment literature to obtain clean text that includes the contents of the life cycle inventory; Based on the cleaning text, at least one candidate process is determined, and each candidate process corresponds to a set of physical lifecycle inventory data including input-side data and output-side data. Generate a process dataset that conforms to the TIDAS data model requirements for each candidate process. The process dataset includes a set of physical lifecycle inventory input and output items and structured stream retrieval hints for each input and output item. The set of physical lifecycle inventory input and output items includes at least one input item and at least one output item. The structured stream retrieval hints are used to express the domain semantics of the corresponding input and output items for retrieval. The process dataset is validated according to preset validation requirements; In response to the successful verification of the process dataset, a standard flow dataset matching the input and output items is determined from an external standard lifecycle assessment database based on the structured flow retrieval prompts. For successfully matched input / output items, establish the association between the standard stream dataset and the input / output items; for unmatched input / output items, generate unmatched tags. The process dataset carrying the aforementioned association or the aforementioned unmatched tag, along with the corresponding literature source dataset, are exported as dataset files conforming to the TIDAS data pattern. The exported dataset files are then subjected to TIDAS data pattern validation, and the pattern validation results are output.

2. The method according to claim 1, characterized in that, The preprocessing of life cycle assessment literature includes at least one of the following operations: Based on paragraph characteristics and / or paragraph position of the document content, filter out document fragments that do not carry lifecycle inventory content; Numerical artifact correction is performed on non-table and non-image description content based on the format type of the document content; Convert characters into a preset standard format.

3. The method according to claim 1, characterized in that, The step of determining at least one candidate process based on the cleaning text includes: The candidate process is determined based on a set of physical lifecycle inventory data rows belonging to the same process boundary in the cleaning text; The physical lifecycle inventory data rows include data rows used to quantify the resource consumption, product output, and / or environmental emissions of the candidate process, and the physical lifecycle inventory data rows have identifiable physical units.

4. The method according to claim 1 or 3, characterized in that, The candidate process includes at least one of the following: Evidence information, used to record the original content information of the candidate process in the literature; The information used to determine the data availability of the candidate process in life cycle assessment modeling is used to identify the specific process.

5. The method according to claim 1, characterized in that, The process dataset also includes the name, input / output direction, and unit information for each input / output item; The verification of the process dataset according to preset verification requirements includes at least one of the following verification contents: Verify whether the process dataset includes both input and output items, and verify whether the names of the input and output items and the field content of the structured stream retrieval prompt information are complete; Verify whether the input / output item names, input / output directions, and / or the unit information represent life cycle impact assessment indicators.

6. The method according to claim 1, characterized in that, The process dataset also includes process classification information; The method further includes: Based on the preset process classification hierarchy, the process dataset corresponding to the candidate process is classified level by level. The step of validating the process dataset according to preset validation requirements includes: Perform process classification verification on the process dataset.

7. The method according to claim 1, characterized in that, The process dataset also includes geographic information and reference years; The process dataset generated for each candidate process in accordance with the TIDAS data model includes: The standardized description of the geographic information is determined according to the preset geographic coding method; In response to the fact that the reference year is not explicitly given in the cleaned text, the reference year is determined based on the data coverage year of the document; The step of validating the process dataset according to preset validation requirements includes: The process dataset is subjected to geographic information normalization verification.

8. The method according to claim 1, characterized in that, The step of determining the standard flow dataset matching the input and output items from an external standard lifecycle assessment database based on the structured flow retrieval prompts includes: Based on the stream type to which the input and output items belong, a stream retrieval query request is constructed based on the input and output item names and / or the structured stream retrieval prompt information; Based on the stream retrieval query request, a preset number of candidate stream datasets are determined from the external standard lifecycle assessment database; In response to a successful match between the description information of the target candidate stream dataset and the input / output item, the target candidate stream dataset is determined to be the standard stream dataset.

9. The method according to claim 8, characterized in that, For successfully matched input / output items, the process of constructing the association between the standard stream dataset and the input / output items includes: In response to a successful match, the identification information of the standard stream dataset is written into the stream reference data field of the input / output item; The process of generating a non-matching flag for unmatched input / output items includes: In response to a matching failure, generate a non-matching record for the input / output item and / or associate the input / output item with a placeholder stream reference.

10. The method according to claim 1, characterized in that, The method further includes: In response to the failure of the process dataset verification, a new process dataset is generated for the candidate process. In response to the number of times the process dataset is regenerated reaching a preset value, the process dataset verification failure information of the candidate process is output, and the candidate process is associated with the manual review mark.

11. An automatic data extraction system for process datasets in life cycle assessment literature, characterized in that, include: The document preprocessing module preprocesses the life cycle assessment document fragments to obtain clean text including the life cycle inventory content. The process enumeration module is used to determine at least one candidate process based on the cleaning text. Each candidate process corresponds to a set of physical lifecycle inventory data including input-side data and output-side data. The process dataset generation module is used to generate process datasets that conform to the TIDAS data model requirements for each candidate process. The process dataset includes a set of physical lifecycle inventory input and output items and structured stream retrieval hints for each input and output item. The set of physical lifecycle inventory input and output items includes at least one input item and at least one output item. The structured stream retrieval hints are used to express the domain semantics of the corresponding input and output items for retrieval. The input / output item verification module is used to verify the process dataset according to preset verification requirements; The flow matching module is used to determine, in response to the successful verification of the process dataset, a standard flow dataset that matches the input and output items from an external standard lifecycle assessment database based on the structured flow retrieval prompts. as well as For successfully matched input / output items, establish the association between the standard stream dataset and the input / output items; for unmatched input / output items, generate unmatched tags. The dataset export and verification module is used to export the process dataset carrying the aforementioned correlation or the aforementioned unmatched marker, as well as the corresponding literature source dataset, into a dataset file that conforms to the TIDAS data pattern, and to perform TIDAS data pattern verification on the exported dataset file and output the pattern verification result.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-10.

13. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method as described in any one of claims 1-10.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-10.

Images