Text Representations for Patent Classification

D’Hondt, Eva; Verberne, Suzan; Koster, C. H. A.; Boves, L.W.J.

doi:10.1162/coli_a_00149

Cited by 41 publications

(12 citation statements)

References 22 publications

(29 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The bag-of-words (BOW) model [8,18] is a typical, statistically-based text representation approach, which is almost always used in patent analysis studies [1,18]. After stemming, filtering and stop-word removal, the BOW represents each document by the words' occurrences, ignoring their ordering and grammar in the original document.…”

Section: Feature Extraction From Textmentioning

confidence: 99%

A Hierarchical Feature Extraction Model for Multi-Label Mechanical Patent Classification

et al. 2018

Sustainability

View full text Add to dashboard Cite

Abstract:Various studies have focused on feature extraction methods for automatic patent classification in recent years. However, most of these approaches are based on the knowledge from experts in related domains. Here we propose a hierarchical feature extraction model (HFEM) for multi-label mechanical patent classification, which is able to capture both local features of phrases as well as global and temporal semantics. First, a n-gram feature extractor based on convolutional neural networks (CNNs) is designed to extract salient local lexical-level features. Next, a long dependency feature extraction model based on the bidirectional long-short-term memory (BiLSTM) neural network model is proposed to capture sequential correlations from higher-level sequence representations. Then the HFEM algorithm and its hierarchical feature extraction architecture are detailed. We establish the training, validation and test datasets, containing 72,532, 18,133, and 2679 mechanical patent documents, respectively, and then check the performance of HFEMs. Finally, we compared the results of the proposed HFEM and three other single neural network models, namely CNN, long-short-term memory (LSTM), and BiLSTM. The experimental results indicate that our proposed HFEM outperforms the other compared models in both precision and recall.

show abstract

Section: Feature Extraction From Textmentioning

confidence: 99%

A Hierarchical Feature Extraction Model for Multi-Label Mechanical Patent Classification

et al. 2018

Sustainability

View full text Add to dashboard Cite

show abstract

“…The functionality of n-gram graphs was provided by the open source library of JInsect. 9 To derive the performance of the representation models, we applied them to two established classification algorithms that are typically used for TC in conjunction with the bag models: Naive Bayes Multinomial (NBM) and Support Vector Machines (SVM) [29,44]. The former classifies instances based on the conditional probabilities of their feature values, while the latter uses optimization techniques to identify the maximum margin decision hyperplane.…”

Section: Setupmentioning

confidence: 99%

“…In [9], the authors examine four content-based representation models: lemmatized token unigrams, lemmatized token bigrams and lemmatized "dependency triples" 15 that were obtained from the Stanford and the AEGIR parsers. The comparative analysis was performed over a set of curated documents that contained patent abstracts in English.…”

Section: Related Workmentioning

confidence: 99%

Graph vs. bag representation models for the topic classification of web documents

2015

View full text Add to dashboard Cite

Text classification constitutes a popular task in Web research with various applications that range from spam filtering to sentiment analysis. In this paper, we argue that its performance depends on the quality of Web documents, which varies significantly. For example, the curated content of news articles involves different challenges than the usergenerated content of blog posts and Social Media messages. We experimentally verify our claim, quantifying the main factors that affect the performance of text classification. We also argue that the established bag-of-words representation models are inadequate for handling all document types, as they merely extract frequent, yet distinguishing terms from the textual content of the training set. Thus, they suffer from low robustness in the context of noisy or unseen content, unless they are enriched with contextual, application-specific information. In their place, we propose the use of n-gram graphs, a model that goes beyond the bag-of-words representation, transforming every document into a graph: its nodes correspond to character or word n-grams and the co-occurring ones are connected by weighted edges. Individual document graphs can be combined into class graphs and graph similarities are employed to position and classify documents into the vector space. This approach offers two advantages with respect to bag models: first, classification accuracy increases due to the contextual information that is encapsulated in the edges of the n-gram graphs. Second, World Wide Web it reduces the search space to a limited set of robust, endogenous features that depend on the number of classes, rather than the size of the vocabulary. Our thorough experimental study over three large, real-world corpora confirms the superior performance of n-gram graphs across the main types of Web documents.

show abstract

“…Other parts are narrative text providing information regarding the patent and are given under the headings: Title, Abstract, Field, Background, Detailed description, and Claims. There are many ways to represent the whole patent document in previous patent automatic classification and retrieval applications 15,20,21 . Some researchers believe that the human generated abstracts of patent documents are very precise and the most important section for patent classification.…”

Section: Automatic Patent Classificationmentioning

confidence: 99%

A Supervised Requirement-oriented Patent Classification Scheme Based on the Combination of Metadata and Citation Information

Zhu¹,

Wang²,

Zhu³

et al. 2015

IJCIS

View full text Add to dashboard Cite

Patent classification systems are applied extensively in innovative analysis. Existing patent classification schemes are either technology-dependent or TRIZ-based. The former ones, such as the IPC and UPC, are normally developed by different patent offices in the world mainly for the purpose of patentability examination and patent retrieval, while the latter is for TRIZ users and analysts with no more than 40 categories. These static classifications are too complex and general to meet the in-depth patent classification requirements of a specific technology area or organization. To tackle these drawbacks, in this paper, we propose an automatic requirement-oriented patent classification scheme as a complementary method using supervised machine learning techniques to classify patent dataset into a user-defined taxonomy. The requirement-oriented patent taxonomy can be technology-dependent, application-dependent or a mixture of both tailored to specific business objectives. It is more comprehensible and adaptable to various patent management requirements. Through a set of experiments on a collection of 14,414 patents in a case study in the technology area of system on a chip (SoC), we recommend using the combination of the metadata and citation information as the document representation for the new method since it can obtain relatively high classification accuracy with a dramatically simplified document preprocessing process.

show abstract

Text Representations for Patent Classification

Cited by 41 publications

References 22 publications

A Hierarchical Feature Extraction Model for Multi-Label Mechanical Patent Classification

A Hierarchical Feature Extraction Model for Multi-Label Mechanical Patent Classification

Graph vs. bag representation models for the topic classification of web documents

A Supervised Requirement-oriented Patent Classification Scheme Based on the Combination of Metadata and Citation Information

Contact Info

Product

Resources

About