Resource timeliness identification method and device, electronic equipment and storage medium

Abstract

The invention provides a resource timeliness identification method and device, electronic equipment and a storage medium. The invention relates to the technical field of artificial intelligence, in particular to the technical fields of deep learning, large models, natural language processing, knowledge maps, data analysis and the like, and can be used for application scenes such as resource timeliness analysis, intelligent search, intelligent recommendation and the like. According to the specific implementation scheme, the method comprises the steps of obtaining multi-scale timeliness sensitive features of target resources; obtaining resource vertical class features of the target resource; acquiring author droop features of the target resource; and determining a timeliness identification result of the target resource based on the timeliness sensitive word list, the multi-scale timeliness sensitive feature, the resource droop feature and the author droop feature.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, especially to technical fields such as deep learning, large models, natural language processing, knowledge graphs, and data analysis. It can be used in application scenarios such as resource timeliness analysis, intelligent search, and intelligent recommendation, and specifically relates to a resource timeliness identification method, device, equipment, and storage medium. Background Art

[0002] In the Internet age, information updates very quickly, with a large amount of new content being generated every day. In order to keep pace with the rapidly developing society, users are usually not satisfied with outdated content, but pursue real-time updated information, so they often have high requirements for resource timeliness. Summary of the invention

[0003] The present disclosure provides a resource timeliness identification method, device, electronic device and storage medium.

[0004] According to the first aspect of the present disclosure, a resource timeliness identification method is provided, the method comprising: obtaining multi-scale timeliness sensitive features of a target resource; obtaining resource vertical category features of the target resource; obtaining author vertical category features of the target resource; and determining a timeliness identification result of the target resource based on a timeliness sensitive vocabulary, multi-scale timeliness sensitive features, resource vertical category features, and author vertical category features.

[0005] According to the second aspect of the present disclosure, a resource timeliness identification device is provided, which includes: a first acquisition module for acquiring multi-scale timeliness sensitive features of a target resource; a second acquisition module for acquiring resource vertical category features of the target resource; a third acquisition module for acquiring author vertical category features of the target resource; and an identification module for determining a timeliness identification result of the target resource based on a timeliness sensitive vocabulary, multi-scale timeliness sensitive features, resource vertical category features, and author vertical category features.

[0006] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0007] at least one processor; and

[0008] a memory communicatively connected to the at least one processor; wherein,

[0009] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute any method in the embodiments of the present disclosure.

[0010] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute any method according to the embodiments of the present disclosure.

[0011] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program, which implements any method according to the embodiments of the present disclosure when executed by a processor.

[0012] By adopting the solution disclosed in the present invention, the timeliness of resources is judged based on the timeliness-sensitive vocabulary and different dimensional characteristics of resources, which can improve the accuracy and efficiency of resource timeliness identification.

[0013] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0015] Figure 1 is a flowchart of a method for identifying resource timeliness according to an embodiment of the present disclosure;

[0016] Figure 2 is an architectural diagram of resource timeliness identification according to an embodiment of the present disclosure;

[0017] Figure 3 is a schematic diagram of obtaining a time-sensitive word list according to an embodiment of the present disclosure;

[0018] Figure 4 is a schematic diagram of obtaining time-sensitive features at a sentence scale according to an embodiment of the present disclosure;

[0019] Figure 5 is a schematic diagram of obtaining time-sensitive features at a vocabulary scale according to an embodiment of the present disclosure;

[0020] Figure 6 is a schematic diagram of obtaining time-sensitive features on a time scale according to an embodiment of the present disclosure;

[0021] Figure 7 is a structural diagram of a resource timeliness identification device according to an embodiment of the present disclosure;

[0022] Figure 8 is a scenario diagram of a method for identifying resource timeliness according to an embodiment of the present disclosure;

[0023] Fig. 9 It is a structural diagram of an electronic device used to implement the resource timeliness identification method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0024] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0025] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there may be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The term "at least one" in this article means any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C. The terms "first" and "second" in this article refer to multiple similar technical terms and distinguish them. They do not mean to limit the order or to limit them to only two. For example, the first feature and the second feature refer to two types / two features. The first feature can be one or more, and the second feature can also be one or more.

[0026] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present disclosure.

[0027] Before introducing the technical solutions of the embodiments of the present disclosure, the technical terms that may be used in the present disclosure are further explained:

[0028] Resource timeliness: Resource timeliness refers to the measure of resource effectiveness over a certain period of time. It reflects the value and usage of resources over a specific period of time.

[0029] Multi-scale time-sensitive characteristics: Multi-scale time-sensitive characteristics refer to the time-sensitive characteristics of resources at different time scales.

[0030] Resource vertical characteristics: Resource vertical characteristics refer to the characteristics of the field or category to which the resource belongs. These characteristics may include the resource's theme, content type, audience group, etc.

[0031] Author vertical characteristics: Author vertical characteristics refer to the characteristics of the author's field or professional background.

[0032] In the related technologies, the current resource timeliness classification has the following technical difficulties:

[0033] 1. Resource timeliness is often strongly correlated with specific keywords in the resources, such as today, recent, and review. Classification only at the semantic level of resource content will lose detailed information, resulting in reduced classification accuracy. In addition, there are many ways to express keywords that are strongly related to timeliness, which are difficult to list exhaustively.

[0034] 2. Keywords that are strongly related to resource timeliness often have different meanings in different contexts and need to be accurately judged based on the context.

[0035] 3. For specific resources, there is often a certain amount of noise in the process of obtaining its content, or the resource content is inconsistent with the actual meaning of the resource (such as video background music, etc.). In this case, it is difficult to accurately determine the timeliness of the resource using only the resource content.

[0036] The present disclosure provides a method for identifying resource timeliness. Figure 1 It is a flow chart of a resource timeliness identification method according to an embodiment of the present disclosure, and the resource timeliness identification method can be applied to a resource timeliness identification device. The resource timeliness identification device is located in an electronic device. The electronic device includes but is not limited to fixed devices and / or mobile devices. For example, fixed devices include but are not limited to servers, and the server can be a cloud server or an ordinary server. For example, mobile devices include but are not limited to: mobile phones, tablet computers. In some possible implementations, the resource timeliness identification method can also be implemented by a processor calling computer-readable instructions stored in a memory.

[0037] like Figure 1 As shown, the resource timeliness identification method includes:

[0038] S101, obtain multi-scale time-sensitive features of target resources;

[0039] S102, obtaining resource vertical characteristics of the target resource;

[0040] S103, obtaining the author vertical category characteristics of the target resource;

[0041] S104. Determine the timeliness identification result of the target resource based on the timeliness sensitive vocabulary, multi-scale timeliness sensitive features, resource vertical category features and author vertical category features.

[0042] In the disclosed embodiments, the target resource is a resource to be identified, and these resources may include various types, such as videos, articles, audio, pictures, and various documents. Video resources may include teaching videos, news reports, movie trailers, etc. Article resources may include news reports, academic papers, blog posts, user comments, etc. Audio resources may include podcasts, music clips, voice commands, etc. Image resources may include photos, charts, illustrations, advertising posters, etc. Document resources may include emails, spreadsheets, presentations, etc.

[0043] In the disclosed embodiment, multi-scale timeliness sensitive features refer to features related to the timeliness of target resources extracted at different scales (such as content, sentence, vocabulary, and time, etc.). These features can reflect the timeliness and information value of resources at different levels. By integrating multi-scale timeliness sensitive features, the timeliness of resources can be preliminarily judged. Here, the embedding model at the corresponding scale can be used to obtain the timeliness sensitive features at the corresponding scale.

[0044] In the disclosed embodiments, the resource vertical category characteristics refer to the characteristics of the field or category to which the resource belongs. Resources in different fields may have different timeliness requirements. For example, the timeliness of news reports is usually extremely high, while academic papers may focus more on long-term value. Therefore, when identifying the timeliness of resources, it is necessary to obtain the vertical category characteristics of the resources in order to more accurately evaluate their timeliness. Here, the pre-trained first vertical category model can be used to obtain the resource vertical category characteristics.

[0045] In the disclosed embodiments, the author's vertical category features refer to the characteristics of the author's field or professional background. Authors in different fields may also have different timeliness of published resources. For example, resources published by industry experts or well-known media people often have high timeliness and authority. Therefore, when identifying the timeliness of resources, the author's vertical category features also need to be considered. For example, a pre-trained second vertical category model can be used to obtain the author's vertical category features.

[0046] In some embodiments, after obtaining the multi-scale timeliness-sensitive features, resource vertical category features, and author vertical category features of the target resource, machine learning algorithms or expert systems and other methods can be used to comprehensively analyze and evaluate these features based on a timeliness-sensitive vocabulary to determine the timeliness identification result of the target resource.

[0047] Figure 2 The architecture diagram of resource timeliness identification is shown as Figure 2As shown, the multi-scale timeliness sensitive features of the target resource are determined based on the resource content, the resource vertical category features of the target resource are determined based on the resource vertical category data, and the author vertical category features of the target resource are determined based on the author vertical category data; the multi-scale timeliness sensitive features, resource vertical category features and author vertical category features are aligned, the aligned features are fused, and the timeliness recognition result of the target resource is determined based on the fused multi-dimensional features.

[0048] For example, suppose there is a news report, which was published on November 20, 2024, and its content is about a new policy recently issued by a certain city. When identifying the timeliness of this news report, you can follow the following steps: Extract multi-scale timeliness sensitive features: The news report was published on November 20, 2024, and the policy mentioned in the content was also issued recently. Therefore, this news report has high timeliness. Obtain resource vertical characteristics: This news report belongs to the news report category, which usually requires extremely high timeliness. Therefore, its timeliness requirements are high. Obtain author vertical characteristics: Assuming that the author of this news report is a well-known media person, the news reports he publishes are usually highly authoritative and timely. Comprehensive evaluation: Combining the above information, it can be judged that the news report has a high timeliness and is a timely and effective source of information.

[0049] The technical solution of the disclosed embodiment determines the timeliness identification result of the target resource based on multi-scale timeliness sensitive features, resource vertical category features and author vertical category features, which can improve the accuracy of identification and thus improve the accuracy of resource recommendation. By automatically identifying the timeliness of resources, outdated or invalid information can be quickly screened out, improving the efficiency of information screening.

[0050] In some embodiments, obtaining multi-scale time-sensitive features of the target resource includes at least one of the following:

[0051] Obtain the timeliness-sensitive features of the target resources at the content scale;

[0052] Obtain time-sensitive features of target resources at the sentence scale;

[0053] Acquire the time-sensitive features of the target resource at the lexical scale;

[0054] Obtain the time-sensitive characteristics of target resources on a time scale.

[0055] In the disclosed embodiment, content scale refers to the timeliness and relevance of the overall content of the resource. The timeliness-sensitive feature of content scale mainly focuses on the relevance and importance of the topic or event discussed by the resource at the current time point.

[0056] In the disclosed embodiment, the sentence scale refers to the timeliness and information value of each sentence in the resource. The timeliness sensitive features on the sentence scale focus on the tense, information content and importance of key sentences in the resource.

[0057] In the disclosed embodiment, the vocabulary scale refers to the timeliness and popularity of the words and phrases used in the resource. The timeliness sensitive feature at the vocabulary scale focuses on the popularity and usage of time-sensitive words and phrases related to time, events or trends in the resource.

[0058] In the embodiments of the present disclosure, the time scale refers to the time information and time nodes involved in the resource. The time-sensitivity feature on the time scale mainly focuses on the accuracy and timeliness of the time information in the resource.

[0059] In some implementations, the timeliness-sensitive features at the content scale mainly focus on the timeliness and relevance of the overall content of the resource. Specifically, extracting the timeliness-sensitive features at the content scale of the target resource may include:

[0060] Analyze resource topics: Determine the topics or events that the resource discusses and their relevance and importance at the current point in time.

[0061] Evaluate content update frequency: For resources that are continuously updated (such as news reports, blog posts, etc.), evaluate their update frequency and content changes to determine their timeliness.

[0062] Compare with similar resources: Compare the target resource with similar resources and analyze the uniqueness and timeliness of their content.

[0063] In some implementations, the timeliness sensitive features at the sentence scale focus on the timeliness and information value of each sentence in the resource. Specifically, extracting the timeliness sensitive features at the sentence scale of the target resource may include:

[0064] Identify key sentences: Find the key sentences in the source that describe time, events, or trends.

[0065] Analyze sentence tense: Determine the timeliness of a sentence based on its tense (such as present tense, past tense, future tense, etc.).

[0066] Evaluate the information value of sentences: Evaluate the timeliness of sentences based on the amount of information they provide, their novelty, and their importance.

[0067] In some implementations, the timeliness-sensitive features at the vocabulary scale focus on the timeliness and popularity of the words and phrases used in the resource. Specifically, extracting the timeliness-sensitive features at the vocabulary scale of the target resource may include:

[0068] Identify time-sensitive words: Look for time-sensitive words and phrases in your resources that refer to time, events, or trends.

[0069] Analyze vocabulary popularity: Use natural language processing tools or dictionary resources to analyze the popularity and usage of these words at the current point in time.

[0070] Assessing word relevance: Assessing the currency of words based on their relevance, novelty, and importance to the resource topic.

[0071] In some implementations, the time-sensitive features on a time scale mainly focus on the time information and time nodes involved in the resource. Specifically, extracting the time-sensitive features of the target resource on a time scale may include:

[0072] Extract time information: Extract all time-related information from resources, such as date, time period, time node, etc.

[0073] Analyze time nodes: Determine the timeliness of the events or trends discussed by the resource based on the time nodes. For example, recent events have a high timeliness.

[0074] Compare time context: Compare the time information in the resource with the current time point to analyze the timeliness and relevance of the resource content.

[0075] Assuming there is a news report about the "2024 World Cup", its multi-scale time-sensitive features can be extracted as follows:

[0076] Content scale: This news report discusses the 2024 World Cup, which is a current hot topic and therefore has a high degree of timeliness.

[0077] Sentence scale: Key sentences in news reports, such as "The 2024 World Cup will open on XX day of XX month in XX city" and "XX teams have confirmed their participation", provide specific information about the opening time of the World Cup and the participating teams. These sentences are highly timely.

[0078] Lexical scale: Time-sensitive words such as "2024 World Cup", "opening", and "participating teams" are used in news reports. These words have high popularity and frequency of use at the current time point.

[0079] Time scale: The time information involved in news reports, such as "2024", "XX month XX day", etc., is closely related to the current time point and therefore has high timeliness.

[0080] By extracting these multi-scale time-sensitive features, we can have a more comprehensive understanding of the timeliness and information value of the target resources.

[0081] In this way, by extracting multi-scale time-sensitive features, we can quickly identify information related to the target time node or with timeliness, and improve the efficiency of information screening. By extracting multi-scale time-sensitive features, we can recommend more accurate and valuable resources to users.

[0082] In some embodiments, the resource timeliness identification method may also include: obtaining at least two candidate keywords by performing word segmentation processing on the text data included in the target resource; determining the frequency of occurrence of each candidate keyword in the target resource and its timeliness identifier; based on the timeliness identifier, screening out timeliness-sensitive keywords from the candidate keywords; and determining a timeliness-sensitive word list based on the timeliness-sensitive keywords and timeliness identifiers.

[0083] In the disclosed embodiment, word segmentation is the process of dividing a continuous text into independent vocabulary units.

[0084] In the disclosed embodiment, the candidate keywords are words that have important meanings in the text and are usually used to represent the theme or content of the text.

[0085] In the disclosed embodiment, the timeliness flag is a classification level (or level) used to describe the timeliness characteristics of a keyword, such as “high timeliness”, “medium timeliness” and “low timeliness”. The timeliness flag can be pre-defined according to the timeliness characteristics of the keyword.

[0086] In the disclosed embodiment, the timeliness-sensitive keyword is a keyword that appears frequently in resources and has significant timeliness characteristics.

[0087] In some implementations, word segmentation processing of text data includes: using natural language processing (NLP) technology or word segmentation tools to perform word segmentation processing on text data in the target resource, and dividing continuous text into independent vocabulary units, namely keywords.

[0088] In some implementations, determining the frequency of occurrence of each candidate keyword in the target resource and its corresponding timeliness identifier includes: for each candidate keyword, counting its frequency of occurrence in the target resource using a word frequency counting algorithm; and assigning a timeliness identifier to it based on the frequency of occurrence.

[0089] In some embodiments, time-sensitive keywords are screened out from multiple candidate keywords, including: determining screening rules based on the frequency of occurrence and timeliness flag of each candidate keyword. For example, a threshold can be set to select only candidate keywords with a frequency of occurrence higher than a certain value and a timeliness flag of "high timeliness" as timeliness-sensitive keywords. Through the screening rules, sensitive keywords with significant timeliness are screened out from all candidate keywords.

[0090] In some implementations, constructing a timeliness sensitive word list includes: integrating the screened timeliness sensitive keywords and timeliness identifiers into a timeliness sensitive word list. The timeliness sensitive word list can be used for subsequent resource timeliness evaluation or classification tasks.

[0091] Figure 3 A schematic diagram of obtaining a time-sensitive word list is shown, such as Figure 3 As shown, a word segmentation operation is performed on the multiple contents included in the input data to obtain multiple candidate keywords; the word frequency (frequency) of each candidate keyword is counted, and the timeliness mark of each candidate keyword is determined based on the frequency of occurrence of each candidate keyword; a timeliness sensitive word list is generated based on the timeliness mark of each candidate keyword, wherein the timeliness sensitive word list includes timeliness sensitive keywords and their corresponding timeliness marks. For example, the timeliness mark corresponding to "today" is "super timeliness", the timeliness mark corresponding to "Dragon Boat Festival" is "strong timeliness", and the timeliness mark corresponding to "world" is "weak timeliness".

[0092] Assume that the target resource is an article about the latest technology news, and the content contains multiple keywords, such as "artificial intelligence", "machine learning", "blockchain", etc. These keywords can be obtained through word segmentation.

[0093] Keyword frequency statistics: Assume that "artificial intelligence" appears 10 times in the article, "machine learning" appears 5 times, and "blockchain" appears 2 times.

[0094] Determination of timeliness label: According to historical data and manual labeling, "artificial intelligence" and "machine learning" are hot topics in the current science and technology field and have high timeliness, so they are classified as "high timeliness" labels; although "blockchain" is also an important technology, its timeliness is relatively low, so it is classified as "medium timeliness" labels.

[0095] Screening of time-sensitive keywords: According to the set threshold (such as the frequency of occurrence is greater than 5 times), screen out "artificial intelligence" as a time-sensitive keyword.

[0096] Construction of time-sensitive vocabulary: Integrate "artificial intelligence" and its "high timeliness" label into the time-sensitive vocabulary.

[0097] In this way, through word segmentation and word frequency statistics, keywords in the text and their timeliness characteristics can be more accurately identified, thereby improving the accuracy of resource timeliness identification. The entire identification process is based on natural language processing and statistical analysis technology, which realizes automated processing, reduces manual intervention, and improves work efficiency.

[0098] In some embodiments, by performing word segmentation on the text data included in the target resource, at least two candidate keywords are obtained, including: performing word segmentation on the text data to obtain at least two words; obtaining the word frequency and the length of each word; and obtaining at least two candidate keywords from the words according to the word frequency and the length of the word.

[0099] In the embodiments of the present disclosure, the word frequency is the number of times a vocabulary appears in the text.

[0100] In the embodiments of the present disclosure, the length of a vocabulary is the number of characters in the vocabulary.

[0101] In some embodiments, counting the word frequency and the length of the vocabulary includes: performing word frequency statistics on the result of word segmentation, that is, calculating the number of times each vocabulary appears in the text. At the same time, calculating the length of each vocabulary, that is, the number of characters in the vocabulary. The length of the vocabulary can be used as an auxiliary index for screening keywords because usually shorter vocabularies (such as stop words) may not contain much useful information, while longer vocabularies (such as technical terms or phrases) may be more specific and information-rich.

[0102] In some embodiments, screening according to the word frequency and the length of the vocabulary includes:

[0103] Setting a word frequency threshold: Only retain the vocabularies with a word frequency higher than the set threshold because high-frequency vocabularies often better reflect the theme of the text.

[0104] Setting a vocabulary length threshold: According to specific circumstances, a reasonable range of vocabulary lengths can be set, and only the vocabularies within this range are retained. For example, vocabularies with too short lengths (such as 1-2 characters) can be excluded because they may be meaningless characters or stop words; for example, some high-frequency but meaningless vocabularies, such as "de" (的), "le" (了), etc., are usually called stop words. At the same time, vocabularies with too long lengths (such as exceeding a certain number of characters) can also be excluded because they may be sentence fragments or long phrases.

[0105] Screening by comprehensively considering the word frequency and the length of the vocabulary: After setting the thresholds, the vocabularies that meet both the word frequency and the vocabulary length requirements are extracted as keywords.

[0106] Suppose the text data of the target resource is an article about "the application of artificial intelligence in the medical field", and part of the content is as follows: "The application of artificial intelligence in the medical field is becoming more and more extensive. It can assist doctors in disease diagnosis, treatment plan formulation, etc."

[0107] Word segmentation result: artificial intelligence / in / medical / field / of / application / more and more / extensive / , / it / can / assist / doctors / in / disease / diagnosis / , / formulate / treatment plan / , etc.

[0108] Word frequency statistics: Assume that "artificial intelligence" appears twice, "medical" appears twice, "field" appears once, "application" appears once, and other words appear once or less.

[0109] Vocabulary length: Assume that all vocabulary words are within a reasonable range (e.g., 2-6 characters).

[0110] Set threshold: Set the word frequency threshold to 2 and the word length range to 2-6 characters.

[0111] Keyword extraction: According to the set threshold, "artificial intelligence" and "medical" are selected as keywords. These two words not only have a high frequency, but also have a reasonable length, which can better reflect the theme of the text.

[0112] In this way, by comprehensively considering the two factors of word frequency and word length, keywords that reflect the text theme can be screened out more accurately. By setting reasonable word frequency and word length thresholds, meaningless words and lengthy sentence fragments can be effectively excluded, reducing the impact of noise on keyword extraction. The entire keyword extraction process is based on natural language processing technology and statistical analysis methods, which improves recognition efficiency.

[0113] In some embodiments, extracting time-sensitive features of a target resource at a sentence scale includes: obtaining sentences containing time-sensitive words by segmenting the target resource; and obtaining time-sensitive features output by the text embedding model by inputting the sentences containing time-sensitive words into a text embedding model.

[0114] Here, the text embedding model is a pre-trained model. The text embedding model is used to output the time-sensitive features of the target resource at the sentence scale based on the input resource title and the sentences containing time-sensitive words in the resource. The text embedding model is a model that maps text (such as sentences, paragraphs, etc.) into a low-dimensional vector space, which can capture the semantic content of the text.

[0115] Figure 4 A schematic diagram of obtaining time-sensitive features at the sentence scale is shown, such as Figure 4As shown, the content contained in the input data is processed into sentences to obtain multiple sentences; sentences containing time-sensitive words are cut out from the multiple sentences; the title of the target resource and the sentences containing time-sensitive words in the target resource are input into the text embedding model, and the text embedding model extracts semantic features, calculates the similarity between the resource title and the sentence feature vector, selects the top K features in similarity, and performs weighted summation, and obtains the time-sensitive features on the sentence scale after mapping by a multilayer perceptron (MLP), where K is an integer greater than 1. The multilayer perceptron is a feedforward neural network composed of multiple fully connected layers stacked together, which can learn complex nonlinear relationships and map the input feature vector to a higher-level feature space.

[0116] In the disclosed embodiment, similarity is a measurement method for measuring the similarity between two vectors, and the value range is between [-1, 1]. The more similar the two vectors are, the closer their similarity is to 1.

[0117] In some implementations, the target resource (such as subtitles in a video, articles, press releases, etc.) is segmented into multiple independent sentences using sentence segmentation techniques in NLP. According to the timeliness sensitive word list, sentences containing these timeliness sensitive words are screened out. These sentences are more likely to contain information related to the timeliness of the target resource.

[0118] In some implementations, a text embedding model is used to obtain semantic features of the resource title and each filtered sentence. These feature vectors can capture the semantic content of the sentence or title for subsequent similarity calculation and feature weighting.

[0119] In some embodiments, the similarity between each sentence and the resource title can be calculated by calculating the cosine similarity or other similarity metrics between the sentence feature vector and the title feature vector. According to the similarity sorting, the K sentences with the highest similarity to the title are screened out. The K value can be adjusted according to specific needs. These sentences are usually more able to summarize or reflect the main content and timeliness characteristics of the resource.

[0120] In some embodiments, the feature vectors of the K selected sentences are weighted and summed. The weight can be determined based on the similarity between the sentence and the title, and the sentence with higher similarity is given a higher weight. The feature vector after weighted summation is used as the input of MLP, and the time-sensitive features of the target resource at the sentence scale are obtained through the mapping of MLP. These features can be used for subsequent tasks such as time series analysis, classification, and clustering.

[0121] Assume that the target resource is a news report about the latest technological developments, titled "Artificial Intelligence Leads a New Technological Trend". Divide the news report into multiple sentences, and filter out sentences containing time-sensitive words such as "artificial intelligence". Use the text embedding model to encode the title and each filtered sentence to obtain their semantic feature vectors. Calculate the cosine similarity between each sentence and the title, and filter out the three sentences with the highest similarity to the title (K = 3). Perform a weighted summation of the feature vectors of the three filtered sentences based on the similarity, and input the weighted feature vectors into the MLP for mapping to obtain the time-sensitive features of the news report at the sentence scale. These features can be used for subsequent tasks such as technology development trend analysis and news report classification.

[0122] In this way, by combining the timeliness-sensitive vocabulary and text embedding model, we can more accurately extract sentences and features related to the timeliness of the target resource, and select the K sentences with the highest similarity to the title, which not only provides rich timeliness information, but also enhances the interpretability of the model, making it easier for users to understand the decision basis of the model. Using multi-layer perceptrons for feature mapping can learn the complex relationship between sentences and improve the performance of the model in tasks such as time series analysis and classification, such as execution accuracy and efficiency.

[0123] In some embodiments, obtaining timeliness-sensitive features of a target resource at a lexical scale includes: obtaining the frequency of occurrence of timeliness-sensitive keywords corresponding to different timeliness identifiers in the target resource based on a timeliness-sensitive word list; determining a timeliness identifier category vector based on the frequency of occurrence; and obtaining the timeliness-sensitive features output by the lexical embedding model by inputting the timeliness identifier category vector into a lexical embedding model.

[0124] Here, the vocabulary embedding model is a pre-trained model. The vocabulary embedding model is used to output vocabulary-scale time-sensitive features based on the frequency of occurrence of time-sensitive keywords in the target resource. The frequency of occurrence of different identifier keywords in the resource content is counted according to the time-sensitive word list, and then the vocabulary-scale embedding vectors of different time-sensitive identifiers are obtained using the vocabulary embedding model, and the identifier embedding vectors are weighted and summed according to the frequency of occurrence of time-sensitive keywords in the resource. After MLP mapping, the vocabulary-scale time-sensitive features are obtained.

[0125] Figure 5 A schematic diagram of obtaining time-sensitive features at the vocabulary scale is shown, such as Figure 5As shown, the content contained in the input data is segmented, the frequency of occurrence of different identification keywords is counted, and the timeliness identification category vector is obtained; the timeliness identification category vector (such as a one-hot vector) is input into the vocabulary embedding model to obtain the timeliness identification embedding vector, and the timeliness identification embedding vector is weighted averaged, and after MLP mapping, the vocabulary scale timeliness sensitive feature is obtained.

[0126] In some implementations, counting the frequency of occurrence of time-sensitive keywords includes: determining keywords with different time-sensitive identifiers according to a pre-built time-sensitive word list; and counting the frequency of occurrence of these keywords in the target resource. This can be achieved through a simple word frequency counting algorithm, or more efficiently processed using NLP tools.

[0127] In some embodiments, constructing a timeliness identification category vector includes: creating a category vector for each timeliness sensitive keyword according to the statistical results. The category vector may be a binary vector, in which each position represents a timeliness identification, and if the keyword belongs to the identification, the corresponding position is 1, otherwise it is 0.

[0128] It should be noted that the category vector here is constructed based on the timeliness identifier of the keyword, rather than the text content of the keyword itself.

[0129] In some implementations, a vocabulary embedding model is used to map each keyword into a low-dimensional vector space to obtain an embedding vector for each keyword. According to the frequency of occurrence of each keyword in the target resource, a weighted sum is performed on the corresponding timeliness identification category vector. In this way, each timeliness identification will obtain a weighted vector representation.

[0130] For example, suppose the target resource is an article about the latest technological developments, which contains multiple time-sensitive keywords, such as "artificial intelligence", "blockchain", etc. These keywords belong to different timeliness labels, such as "high timeliness", "medium timeliness", etc. The frequency of occurrence of these keywords in the article is counted, such as "artificial intelligence" appears 5 times, and "blockchain" appears 3 times. Create a category vector for each keyword, such as the category vector of "artificial intelligence" is [1, 0] (assuming it belongs to the high timeliness label), and the category vector of "blockchain" is [0, 1] (assuming it belongs to the medium timeliness label). Use the vocabulary embedding model to map each keyword into the vector space, and perform weighted summation of the embedded vectors according to the frequency of occurrence. The weighted summed vector is used as the input of the MLP, and the time-sensitive features of the target resource at the vocabulary scale are obtained through the MLP mapping. These features can be used for subsequent time series analysis, prediction and other tasks.

[0131] In this way, by combining the time-sensitive vocabulary and the vocabulary embedding model, the time-sensitive features of the target resource at the vocabulary scale can be extracted more accurately. Using MLP for feature mapping can learn complex nonlinear relationships and enhance the generalization ability of the model. The extracted time-sensitive features can provide strong support for subsequent time series analysis, prediction and other tasks, help to better understand the time-sensitive features of the target resources, and thus improve the accuracy of resource time-sensitive identification. The recognition of the time-sensitive features of the target resource at the vocabulary scale is completed based on the vocabulary embedding model, which realizes automated processing, reduces manual intervention, and speeds up the recognition speed, thereby helping to improve the efficiency of resource time-sensitive identification.

[0132] In some embodiments, obtaining a timeliness-sensitive feature of a target resource on a time factor scale includes: obtaining the content display time of the target resource; obtaining the resource publishing time of the target resource; determining a display time factor based on the content display time and the resource publishing time, the display time factor including the time closest to the resource publishing time in the content display time; classifying the display time factor according to the time difference between the display time factor and the resource publishing time to obtain a display time category vector of the target resource; and obtaining the timeliness-sensitive feature output by the time embedding model by inputting the display time category vector into a time embedding model.

[0133] In the embodiment of the present disclosure, the content display time can be obtained from the content of the target resource by using regular matching. Regular matching is a string matching technology that searches for strings that meet a specific pattern by writing a regular expression.

[0134] In the disclosed embodiment, the content display time (also referred to as the display time word) is a word or phrase that clearly indicates the time in the text, such as "October 1, 2023", "Today", "Next Wednesday", etc.

[0135] In the embodiment of the present disclosure, the display time factor is the display time extracted from the target resource that is closest to the resource publishing time.

[0136] In the embodiment of the present disclosure, the display time category vector is a category vector obtained by classifying the display time factor according to the time difference between the resource publishing time.

[0137] In the disclosed embodiment, the time embedding model is a pre-trained neural network model for mapping the display time category vector into a low-dimensional feature space and learning the relationship between different time classifications and time-sensitive features. Here, the time embedding model is a pre-trained model. The time embedding model is used to output the time-sensitive features of the target resource on the time factor scale according to the display time category vector.

[0138] Figure 6A schematic diagram showing the acquisition of time-sensitive features on a time scale is shown, such as Figure 6 As shown in the figure, for a resource, firstly, the display time words (such as April 8, 2024, September 1, 2023, etc.) in the resource content are extracted by regular matching, and the time words closest to the resource publishing time are selected as the display time factor of the current resource. Then, for the display time factor, the display time factor is classified according to the time difference between the display time factor and the resource publishing time to obtain the display time category vector. Finally, the display time category vector is used as input, and the time embedding model is used to obtain the timeliness sensitive features of the target resource on the time scale.

[0139] In some implementations, extracting the display time words includes: using a regular matching method to search for a character string that conforms to the time format in the target resource. This can be achieved by writing a regular expression, which should be able to match various common time formats, such as "YYYY-MM-DD-YYYY", "YYYY-MM-DD", "MM / DD / YYYY", etc. Through regular matching, all possible display time words are extracted and stored in a list for future use.

[0140] In some embodiments, screening the display time factor includes: obtaining the resource posting time of the target resource from the metadata or release information of the resource; calculating the time difference between each extracted display time word and the resource posting time. This can be achieved by converting the time word into a timestamp and calculating the difference between the timestamps. Screening the content display time closest to the resource posting time as the display time factor of the target resource. If there are multiple time words that are very close to the posting time (for example, a difference of a few minutes or hours), one of them can be selected as a representative, or further judged based on context information.

[0141] In some embodiments, the display time factor is classified, including: classifying the display time factor according to the time difference between the display time factor and the resource posting time. For example, the time difference within one day can be classified as "today", the time difference within one week can be classified as "this week", the time difference within one month can be classified as "this month", and so on. A category vector is created for each classification, in which each position represents a time classification. If the display time factor belongs to a certain classification, the corresponding position is 1, otherwise it is 0. In this way, each display time factor can be converted into a display time category vector.

[0142] In some embodiments, obtaining time-sensitive features includes: using a time embedding model, taking a display time category vector as input, and obtaining time-sensitive features of the target resource on a time factor scale. This model can be a pre-trained neural network model that has been trained to learn the relationship between different time categories and time-sensitive features.

[0143] Assume that the target resource is a news report titled "The Latest Scientific and Technological Advances in October 2024" and published on October 5, 2024. Using regular matching, "October 2024" is extracted from the news report as the display time word. The time difference between "October 2024" and the publication time October 5, 2024 is calculated, and it is found that they are very close (within a few days), so "October 2024" is selected as the display time factor. According to the time difference, "October 2024" is classified as "this month". Then create a display time category vector, where the position corresponding to "this month" is 1, and the positions corresponding to other time categories are 0. Take the display time category vector as input, and use the time embedding model to obtain the timeliness-sensitive features of news reports on the time factor scale. These features can be used for subsequent tasks such as time series analysis and news report classification.

[0144] In this way, by regular matching and filtering the time words closest to the publication time, the display time factor of the target resource can be accurately extracted, thereby improving the accuracy of time-sensitive feature extraction. By using the time embedding model to map the display time category vector to a low-dimensional feature space, the complex relationship between different time classifications and time-sensitive features can be learned, thereby enhancing the generalization ability of the model. The extracted time-sensitive features can provide strong support for subsequent time series analysis, prediction and other tasks, help better understand the time characteristics of the target resource on the time factor scale, and thus improve the accuracy of resource timeliness identification. Based on the time embedding model, the recognition of the time-sensitive features of the target resource on the time scale is completed, which realizes automated processing, reduces manual intervention, speeds up the recognition speed, and thus helps to improve the efficiency of resource timeliness identification.

[0145] In some embodiments, the timeliness identification result of the target resource is determined based on multi-scale timeliness sensitive features, resource vertical category features and author vertical category features, including: aligning the multi-scale timeliness sensitive features, resource vertical category features and author vertical category features to obtain aligned multi-dimensional features; by inputting the aligned multi-dimensional features into a timeliness identification model, obtaining the timeliness identification result output by the timeliness identification model, wherein the timeliness identification model is used to perform feature fusion on the aligned multi-dimensional features, and obtain the timeliness identification result based on the fused features.

[0146] In the disclosed embodiment, multi-scale timeliness-sensitive features are used to represent features related to the timeliness of target resources, such as sentence scale, paragraph scale, article scale, etc.

[0147] In the disclosed embodiment, resource vertical category features are used to represent the features of the category to which the target resource belongs, such as news, technology, entertainment, etc.

[0148] In the disclosed embodiment, the author vertical category features are used to represent the features of the category to which the author of the target resource belongs, such as professional journalists, self-media authors, ordinary users, etc.

[0149] In the disclosed embodiment, the timeliness identification model is a model for fusing the aligned multi-dimensional features and obtaining the timeliness identification result of the target resource based on the fused features.

[0150] In some implementations, the multi-scale time-sensitive features, resource vertical features, and author vertical features are preprocessed to ensure that their dimensions and formats are consistent for alignment, which may require vectorization or normalization of certain features.

[0151] In some implementations, these features are combined in a specific order or structure to form an aligned multi-dimensional feature vector, which contains time-sensitive information of the target resource at multiple scales, information about the resource category to which it belongs, and information about the category to which the author belongs.

[0152] In some embodiments, the aligned multidimensional feature vector is used as input to a timeliness recognition model. This model can be a deep learning model, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or a variant thereof, or other types of machine learning models, such as a support vector machine (SVM), a random forest, etc.

[0153] In some embodiments, after receiving the input, the timeliness recognition model performs feature fusion, that is, combines and interacts each feature in the aligned multi-dimensional feature vector to extract more useful information.

[0154] In some implementations, the model makes decisions based on the fused feature vectors and outputs a timeliness identification result for the target resource. This result can be a binary label (such as "strong timeliness" and "weak timeliness"), a multi-class label (such as "very strong", "strong", "medium", "weak", "very weak"), or a continuous timeliness score.

[0155] Suppose the target resource is a news report about the latest scientific and technological progress, and its author is a well-known science and technology journalist. Extract the multi-scale timeliness-sensitive features of the news report, such as the latest release date mentioned in the sentence, the time node of the scientific and technological progress, etc. Extract the resource vertical category features of the news report, such as the "technology" category. Extract the author vertical category features of the news report, such as the "well-known science and technology journalist" category. Preprocess and align these features to form a multi-dimensional feature vector containing multi-scale timeliness-sensitive features, resource vertical category features, and author vertical category features, and input the aligned multi-dimensional feature vector into the timeliness recognition model. The model performs feature fusion and decision-making, and outputs the timeliness recognition results of the news report, such as "strong timeliness". In this way, the timeliness recognition results of the target resource can be accurately determined based on the multi-scale timeliness-sensitive features, resource vertical category features, and author vertical category features.

[0156] In this way, by combining multi-scale time-sensitive features, resource vertical features, and author vertical features, the timeliness of the target resource can be more comprehensively reflected, thereby improving the accuracy of recognition. The aligned multi-dimensional feature vector contains rich information, enabling the timeliness recognition model to better cope with various complex situations and enhance the robustness of the model. By considering resource vertical features and author vertical features, personalized timeliness recognition results can be provided for different types of resources and authors to meet diverse needs.

[0157] It should be understood that Figures 2 to 6 The schematic diagram shown is only exemplary and not restrictive, and it is scalable, and those skilled in the art can Figures 2 to 6 Various obvious changes and / or substitutions can be made to the examples, and the resulting technical solutions still fall within the scope of the disclosure of the embodiments of the present disclosure.

[0158] The present disclosure provides a resource timeliness identification device, such as Figure 7 As shown, the resource timeliness identification device may include: a first acquisition module 701, used to obtain multi-scale timeliness sensitive features of the target resource; a second acquisition module 702, used to obtain the resource vertical category features of the target resource; a third acquisition module 703, used to obtain the author vertical category features of the target resource; an identification module 704, used to determine the timeliness identification result of the target resource based on the timeliness sensitive vocabulary, multi-scale timeliness sensitive features, resource vertical category features and author vertical category features.

[0159] In some embodiments, the first acquisition module 701 is specifically used to: obtain the timeliness sensitive features of the target resource at the content scale; obtain the timeliness sensitive features of the target resource at the sentence scale; obtain the timeliness sensitive features of the target resource at the vocabulary scale; obtain the timeliness sensitive features of the target resource at the time scale.

[0160] In some embodiments, the resource timeliness identification device may also include: a word segmentation module, which is used to obtain at least two candidate keywords by performing word segmentation processing on the text data included in the target resource; a first determination module, which is used to determine the frequency of occurrence of each keyword in the target resource and its timeliness identification; a screening module, which is used to screen out timeliness-sensitive keywords from the candidate keywords based on the timeliness identification; and a second determination module, which is used to determine a timeliness-sensitive word list based on the timeliness-sensitive keywords and timeliness identification.

[0161] In some embodiments, the word segmentation module is used to: obtain at least two words by performing word segmentation processing on text data; obtain the word frequency and word length of each word; and obtain multiple candidate keywords from the words based on the word frequency and word length.

[0162] In some embodiments, the first acquisition module 701 is further used to: obtain sentences containing timeliness-sensitive words by segmenting the target resource; and obtain timeliness-sensitive features output by the text embedding model by inputting the sentences containing timeliness-sensitive words into the text embedding model.

[0163] In some embodiments, the first acquisition module 701 is also used to: obtain the frequency of occurrence of timeliness-sensitive keywords corresponding to different timeliness identifiers in the target resources according to the timeliness-sensitive word list; determine the timeliness identifier category vector of the target resource according to the frequency of occurrence; and obtain the timeliness-sensitive features output by the word embedding model by inputting the timeliness identifier category vector into the word embedding model.

[0164] In some embodiments, the first acquisition module 701 is also used to: obtain the content display time of the target resource; obtain the resource publishing time of the target resource; determine the display time factor based on the content display time and the resource publishing time, the display time factor including the time closest to the resource publishing time in the content display time; classify the display time factor according to the time difference between the display time factor and the resource publishing time to obtain the display time category vector of the target resource; and obtain the timeliness-sensitive features output by the time embedding model by inputting the display time category vector into the time embedding model.

[0165] In some embodiments, the recognition module 704 includes: an alignment submodule, which is used to align multi-scale timeliness sensitive features, resource vertical category features and author vertical category features to obtain aligned multi-dimensional features; an identification submodule, which is used to input the aligned multi-dimensional features into a timeliness recognition model to obtain the timeliness recognition results output by the timeliness recognition model, wherein the timeliness recognition model is used to perform feature fusion on the aligned multi-dimensional features and obtain the timeliness recognition results based on the fused features.

[0166] For the description of specific functions and examples of each module and submodule of the device in the embodiment of the present disclosure, reference can be made to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0167] The resource timeliness identification device of the disclosed embodiment can judge the timeliness of resources based on a timeliness-sensitive vocabulary and different dimensional features of resources, thereby improving the accuracy and efficiency of resource timeliness identification.

[0168] The present disclosure provides a scenario diagram of a method for identifying resource timeliness, such as Figure 8 shown.

[0169] As mentioned above, the resource timeliness identification method provided by the embodiment of the present disclosure is applied to electronic devices. Electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices.

[0170] Specifically, the electronic device can perform the following operations:

[0171] Acquire multi-scale time-sensitive features of target resources;

[0172] Obtain resource vertical characteristics of the target resource;

[0173] Obtain the author vertical characteristics of the target resource;

[0174] The timeliness identification results of the target resources are determined based on the timeliness sensitive vocabulary, multi-scale timeliness sensitive features, resource vertical category features and author vertical category features.

[0175] It should be understood that Figure 8 The scene diagram shown is only illustrative and not restrictive. Those skilled in the art can Figure 8 Various obvious changes and / or substitutions can be made to the examples, and the resulting technical solutions still fall within the scope of the disclosure of the embodiments of the present disclosure.

[0176] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0177] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0178] Fig. 9A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0179] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 to a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0180] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0181] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a CPU, a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as the resource timeliness identification method. For example, in some embodiments, the resource timeliness identification method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the resource timeliness identification method described above may be executed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the resource timeliness identification method in any other appropriate manner (for example, by means of firmware).

[0182] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0183] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0184] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0185] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a cathode ray tube (CRT) or a liquid crystal display (LCD) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0186] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: Local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

[0187] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0188] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0189] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for identifying resource timeliness, comprising: Acquire multi-scale time-sensitive features of target resources; Obtain resource vertical characteristics of the target resource; Obtaining the author vertical characteristics of the target resource; The timeliness identification result of the target resource is determined based on the timeliness sensitive word list, the multi-scale timeliness sensitive features, the resource vertical category features and the author vertical category features.

2. The method according to claim 1, wherein: The step of acquiring the multi-scale time-sensitive features of the target resource includes at least one of the following: Acquire time-sensitive characteristics of the target resource at the content scale; Acquire time-sensitive features of the target resource at the sentence scale; Acquiring time-sensitive features of the target resource at a lexical scale; Acquire time-sensitive characteristics of the target resource on a time scale.

3. The method according to claim 1 or 2, further comprising: Obtaining at least two candidate keywords by performing word segmentation processing on the text data included in the target resource; Determine the occurrence frequency and timeliness identifier of each candidate keyword in the target resource; Based on the timeliness identifier, filtering out timeliness-sensitive keywords from the candidate keywords; The timeliness-sensitive word list is determined according to the timeliness-sensitive keywords and the timeliness identifier.

4. The method according to claim 3, wherein: The step of performing word segmentation on the text data included in the target resource to obtain at least two candidate keywords includes: Performing the word segmentation process on the text data to obtain at least two words; Obtain the word frequency and vocabulary length of each of the words; The candidate keywords are obtained from the words according to the word frequency and the vocabulary length.

5. The method according to claim 2, wherein: The acquiring of the time-sensitive features of the target resource at the sentence scale includes: By segmenting the target resource into sentences, sentences containing time-sensitive words are obtained; By inputting a sentence containing time-sensitive words into a text embedding model, the time-sensitive features output by the text embedding model are obtained.

6. The method according to claim 2, wherein: The acquiring of the time-sensitive features of the target resource at the vocabulary scale includes: Acquire the occurrence frequency of timeliness-sensitive keywords corresponding to different timeliness identifiers in the target resource according to the timeliness-sensitive word list; Determining a timeliness identification category vector of the target resource according to the occurrence frequency; The timeliness identification category vector is input into a vocabulary embedding model to obtain the timeliness sensitive feature output by the vocabulary embedding model.

7. The method according to claim 2, wherein: The obtaining of the time-sensitive characteristics of the target resource on a time scale includes: Obtaining the content display time of the target resource; Obtain the resource release time of the target resource; Determine a display time factor based on the content display time and the resource publishing time, the display time factor including the time of the content display time that is closest to the resource publishing time; Classifying the display time factor according to the time difference between the display time factor and the resource posting time to obtain a display time category vector of the target resource; By inputting the display time category vector into the time embedding model, the time-sensitive feature output by the time embedding model is obtained.

8. The method according to claim 1, wherein: The determining of the timeliness identification result of the target resource based on the timeliness sensitive word list, the multi-scale timeliness sensitive feature, the resource vertical category feature and the author vertical category feature includes: Aligning the multi-scale timeliness-sensitive features, the resource vertical category features, and the author vertical category features to obtain aligned multi-dimensional features; The timeliness recognition result output by the timeliness recognition model is obtained by inputting the aligned multidimensional features into a timeliness recognition model, wherein the timeliness recognition model is used to perform feature fusion on the aligned multidimensional features and obtain the timeliness recognition result based on the fused features.

9. A resource timeliness identification device, comprising: The first acquisition module is used to acquire multi-scale time-sensitive features of target resources; A second acquisition module is used to acquire resource vertical category characteristics of the target resource; A third acquisition module is used to obtain the author vertical category characteristics of the target resource; An identification module is used to determine the timeliness identification result of the target resource based on the timeliness sensitive word list, the multi-scale timeliness sensitive feature, the resource vertical category feature and the author vertical category feature.

10. The device according to claim 9, wherein: The first acquisition module is used to: Acquire time-sensitive characteristics of the target resource at the content scale; Acquire time-sensitive features of the target resource at the sentence scale; Acquiring time-sensitive features of the target resource at a lexical scale; Acquire time-sensitive characteristics of the target resource on a time scale.

11. The device according to claim 9 or 10, further comprising: A word segmentation module, used for obtaining at least two candidate keywords by performing word segmentation processing on the text data included in the target resource; A first determination module is used to determine the occurrence frequency and timeliness identifier of each candidate keyword in the target resource; A screening module, configured to screen out time-sensitive keywords from the candidate keywords based on the time-sensitive identifier; The second determination module is used to determine the timeliness sensitive word list according to the timeliness sensitive keyword and the timeliness identifier.

12. The device according to claim 11, wherein The word segmentation module is used to: Performing the word segmentation process on the text data to obtain at least two words; Obtain the word frequency and vocabulary length of each of the words; The candidate keywords are obtained from the words according to the word frequency and the vocabulary length.

13. The device according to claim 10, wherein: The first acquisition module is further used for: By segmenting the target resource into sentences, sentences containing time-sensitive words are obtained; By inputting a sentence containing time-sensitive words into a text embedding model, the time-sensitive features output by the text embedding model are obtained.

14. The device according to claim 10, wherein: The first acquisition module is further used for: Acquire the occurrence frequency of timeliness-sensitive keywords corresponding to different timeliness identifiers in the target resource according to the timeliness-sensitive word list; Determining a timeliness identification category vector of the target resource according to the occurrence frequency; The timeliness identification category vector is input into a vocabulary embedding model to obtain the timeliness sensitive feature output by the vocabulary embedding model.

15. The device according to claim 10, wherein: The first acquisition module is further used for: Obtaining the content display time of the target resource; Obtain the resource release time of the target resource; Determine a display time factor based on the content display time and the resource publishing time, the display time factor including the time of the content display time that is closest to the resource publishing time; Classifying the display time factor according to the time difference between the display time factor and the resource posting time to obtain a display time category vector of the target resource; By inputting the display time category vector into the time embedding model, the time-sensitive feature output by the time embedding model is obtained.

16. The device according to claim 9, wherein: The identification module comprises: An alignment submodule, used to align the multi-scale time-sensitive features, the resource vertical category features, and the author vertical category features to obtain aligned multi-dimensional features; The recognition submodule is used to obtain the timeliness recognition result output by the timeliness recognition model by inputting the aligned multidimensional features into the timeliness recognition model, wherein the timeliness recognition model is used to perform feature fusion on the aligned multidimensional features and obtain the timeliness recognition result based on the fused features.

17. An electronic device comprising: at least one processor; as well as a memory communicatively connected to at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 8.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are for causing a computer to execute a method according to any one of claims 1-8.

19. A computer program product comprising a computer program stored on a storage medium, the computer program implementing the method according to any one of claims 1 to 8 when executed by a processor.

Images