Patent standardization plays an important role in bridging innovation with industrial implementation.Most existing studies in the field of patent standardization focus on predicting whole patent standardization.They overlook partial standardization, where only certain technical concepts within a patent are adopted by standards.The main challenges in conceptlevel standardization include:1.How to effectively extract technical concepts from patent documents.2.How to accurately predict the standardization timeline of these concepts.However, existing methods still struggle to accurately extract technical concepts from patent texts and lack the capability to model their standardization timelines.To address these challenges, we propose a Patent Concept Standardization Prediction Model.This Model extracts technical concepts from patent documents and predicts their standardization distribution over time.We first extract standardized technical concepts from the standards associated with the target patents and use them as predefined categories for patent concept classification.Next, we calculate the semantic similarity between patent paragraphs and these concepts, assigning each paragraph to its most relevant category.Based on the aligned pairs, we train a multi-class classification model to enable large-scale concept-level normalization across the patent corpus.After extracting patent concepts,we use neural network to estimate the temporal distribution of standardization for each concept based on its semantic embedding.Experimental results show that our framework outperforms baselines across both patent-to-concept classification and concept-level standardization time distribution prediction.

As competition in intellectual property intensifies, patent activities have become a key indicator to measure international innovation and competitiveness. Patent events refer to events that occur throughout a patent lifecycle, and their frequency directly reflects a country’s innovation and competitive abilities. Therefore, the accurate and efficient extraction of patent events is crucial for countries to track patent development trends and optimize their patent strategies. There is relatively little research on patent event extraction, and its main challenges are: (1) the lack of annotated datasets in the field of patent event extraction restricts the training and evaluation of models. (2) Missing and ambiguous arguments during the extraction process, due to the complexity of the event text. To solve these challenges, we propose an external knowledge network-based patent event extraction method. (1) A patent event dataset is constructed, and common event types and arguments are summarized. (2) An external knowledge network is built based on patent data characteristics, and it helps us better capture event arguments in text. (3) The classification ability of event arguments is enhanced by using adversarial learning and multiple attention mechanisms. Our method is better than the baseline methods in evaluation metrics, making it more suitable for the automatic extraction of patent events.

As competition in intellectual property intensifies, countries around the world are actively engaging in patent activities.The number of patent activities has become an important indicator of international innovation and competitiveness.Patent events refer to the events related to patents, these events can include patent applications, patent grants, patent transfers, patent litigation, and more.The more frequent the patent events, the higher the country's innovation and competitiveness.Accurate and efficient extraction of patent events will help countries understand patent trends and plan their patent activities.Consequently, this has become a popular research topic.The main challenges in patent event extraction include the difficulty in handling unstructured data and the interference of noise with extraction.Additionally, existing event extraction methods have the following limitations:1.Absence of methods that utilize external knowledge related to the data to assist in event extraction, and 2. Lack of awareness in argument, leading to the omission of hidden relationships and features.To solve the above problems, we propose an external knowledge network based patent event extraction method.1.To perform event extraction in the patent domain, we constructed a patent event dataset.2. To enhance argument prompts and connections, we built an external knowledge network incorporating patent data features.3. To represent features, we used multiple attention mechanisms to enhance feature computation.Our method outperforms baseline methods in evaluation metrics, making it more suitable for the automatic extraction of patent events.

Summary How to generate summary with more novel and rich semantics is a challenging issue in the area of multi‐document automatic summary. In this paper, a core semantics extraction model (CSEM) is proposed to improve the novel and rich semantics of multi‐document summary. Firstly, for improving the rich semantics, semantic units, which are a group of association relations of keywords, are used to express texts' semantics. Secondly, for improving the novel semantics, an attenuation function is introduced to adjust the importance of semantic units according to the appearing times that semantic units in the candidate of summary sentences. Thirdly, in order to maximize the novel and rich semantics of summary, the generating process of summary is converted into the optimization process on how to find a set of sentences with a higher importance. Finally, CSEM extracts the least number of sentences to cover the most core semantics in corpus as summary. Experimental results on the benchmark DUC 2004 show that our model outperforms the state‐of‐art approaches (eg, OCCAMS_V, JS‐Gen‐2) under official metric. Especially, the recall of our model in ROUGE‐1 is 40.684%, which is better than other approaches (eg, OCCAMS_V 38.497% and JS‐Gen‐2 36.739%).