<> "The repository administrator has not yet configured an RDF license."^^ . <> . . "Neural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata"^^ . "Intellectual property violations involve substantial litigation and license costs, because of\r\nwhich patent search is of utmost importance. Over the years, patent corpora have amassed\r\nmillions of patents, making manual searches impractical. Patent classification techniques\r\nhelp domain experts to search and analyze patents. On submission to an examination office,\r\na patent application is assigned with labels from pre-defined patent taxonomies, e.g., Cooperative\r\nPatent Classification (CPC) and International Patent Classification (IPC). CPC/IPC\r\nclassification helps to route patent applications to the correct department and assists in\r\nperforming prior art searches. In addition to CPC/IPC classification, we address the classification\r\ntask associated with the Patent Landscape Study (PLS), a process that allows\r\norganizations to search patents, categorize them into custom labels, and analyze them to\r\nderive crucial insights. This thesis significantly contributes to the improvement of patent\r\nclassification systems by addressing the key challenges described below.\r\n\r\nMost of the existing CPC/IPC classification datasets provide only limited texts of the\r\nincluded patents and are, therefore, insufficient for our experiments. In response to this\r\nissue, we release a CPC classification dataset that includes the full texts of patents. Further,\r\nthe unavailability of open-source datasets is a major bottleneck for the automation of PLS.\r\nTo address this challenge, we curate, enrich, and release three open-source datasets from\r\ntwo diverse domains.\r\n\r\nDespite CPC/IPC classification being a hierarchical multi-label classification task, most\r\nprior neural models have not considered the hierarchical taxonomy when designing model\r\narchitectures and have often predicted labels only for a single level. We make a major contribution\r\nwith our memory-efficient model architecture, which shares a single transformerbased\r\nlanguage model across multiple classification heads, one for each label in the taxonomy,\r\nand leverages hierarchical links in the model architecture. We demonstrate that the\r\nproposed technique consistently outperforms baselines, particularly for infrequent labels.\r\n\r\nOur analysis shows that the sentences and abstracts of patents are often duplicated,\r\nillustrating the relevance of the full texts of patents to perform classification. However,\r\ntransformer-based language models that take 512 or 4,096 tokens as input are insufficient\r\nfor patents, which contain 12.5k tokens on average. Motivated by these factors, we make a\r\nmajor contribution with our document representation technique, which combines truncated\r\nsection text embeddings using vector summation, performing better than baselines. In\r\naddition, we propose a sentence ranker and demonstrate that the extractive summarization\r\ntechniques are effective in selecting informative sentences for neural representation in the\r\ncontext of patent classification.\r\n\r\nUnlike CPC/IPC classification, in the case of PLS, the CPC/IPC labels are known during\r\ninference. As a major contribution, we enrich the document representation by combining\r\nCPC/IPC labels with patent text to predict PLS-oriented categories, often representing\r\nconcepts different from CPC/IPC labels. To demonstrate the broader applicability of the\r\nproposed technique, we apply it to a similar task: classifying research publications into\r\ntarget categories using text and author-provided keywords as input."^^ . "2024" . . . . . . . "Subhash Chandra"^^ . "Pujari"^^ . "Subhash Chandra Pujari"^^ . . . . . . "Neural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata (Other)"^^ . . . . . . "indexcodes.txt"^^ . . . "Neural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata (PDF)"^^ . . . "thesis_Subhash_Chandra_Pujari_2024-07-30.pdf"^^ . . . "Neural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata (Other)"^^ . . . . . . "lightbox.jpg"^^ . . . "Neural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata (Other)"^^ . . . . . . "preview.jpg"^^ . . . "Neural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata (Other)"^^ . . . . . . "medium.jpg"^^ . . . "Neural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata (Other)"^^ . . . . . . "small.jpg"^^ . . "HTML Summary of #35223 \n\nNeural Patent Classification beyond Title and Abstract: Leveraging Patent Text and Metadata\n\n" . "text/html" . . . "004 Informatik"@de . "004 Data processing Computer science"@en . . . "020 Bibliotheks- und Informationswissenschaft"@de . "020 Library and information sciences"@en . . . "600 Technik, Medizin, angewandte Wissenschaften"@de . "600 Technology (Applied sciences)"@en . .