DOCSLIB.ORG

Arxiv:1907.02679V1 [Cs.CL] 5 Jul 2019

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Arxiv:1907.02679V1 [Cs.CL] 5 Jul 2019 Improving Chemical Named Entity Recognition in Patents with Contextualized Word Embeddings Zenan Zhai1, Dat Quoc Nguyen1, Saber A. Akhondi2, Camilo Thorne2, Christian Druckenbrodt2, Trevor Cohn1, Michelle Gregory2, Karin Verspoor1 1The University of Melbourne, Australia; 2Elsevier 1fzenan.zhai,dqnguyen,trevor.cohn,[email protected] 2fs.akhondi,c.thorne.1,c.druckenbrodt,[email protected] Abstract supporting relation extraction (Wei et al., 2016), reaction prediction (Schwaller et al., 2018) and Chemical patents are an important resource for chemical information. However, few chemi- retro-synthesis (Segler et al., 2018). cal Named Entity Recognition (NER) systems However, performing NER in chemical patents have been evaluated on patent documents, due can be challenging (Akhondi et al., 2014). As le- in part to their structural and linguistic com- gal documents, patents are written in a very differ- plexity. In this paper, we explore the NER ent way compared to scientific literature. When performance of a BiLSTM-CRF model utilis- writing scientific papers, authors strive to make ing pre-trained word embeddings, character- their words as clear and straight-forward as pos- level word representations and contextual- sible, whereas patent authors often seek to pro- ized ELMo word representations for chemi- cal patents. We compare word embeddings tect their knowledge from being fully disclosed pre-trained on biomedical and chemical patent (Valentinuzzi, 2017). corpora. The effect of tokenizers optimized In tension with this is the need to claim broad for the chemical domain on NER performance scope for intellectual property reasons, and hence in chemical patents is also explored. The re- patents typically contain more details and are sults on two patent corpora show that contex- more exhaustive than scientific papers (Lupu et al., tualized word representations generated from 2011). ELMo substantially improve chemical NER performance w.r.t. the current state-of-the-art. There are a number of characteristics of patent We also show that domain-specific resources texts that create challenges for NLP in this con- such as word embeddings trained on chemical text. Long sentences listing names of compounds patents and chemical-specific tokenizers have in chemical patents are frequently used. The struc- a positive impact on NER performance. ture of sentences in patent claims is usually com- plex, and syntactic parsing in patents can be diffi- 1 Introduction cult (Hu et al., 2016). A quantitative analysis by Chemical patents are an important starting point Verberne et al.(2010) showed that the average sen- for understanding of chemical compound purpose, tence length in a patent corpus is much longer than properties, and novelty. New chemical compounds in general language use. That work also showed are often initially disclosed in patent documents; that the lexicon used in patents usually includes arXiv:1907.02679v1 [cs.CL] 5 Jul 2019 however it may take 1-3 years for these chemi- domain-specific and novel terms that are difficult cals to be mentioned in chemical literature (Senger to understand. Some patent authorities use Op- et al., 2015), suggesting that patents are a valuable tical Character Recognition (OCR) for digitizing but underutilized resource. As the number of new patents, which can be problematic when applying chemical patent applications is drastically increas- automatic NLP approaches as the OCR errors in- ing every year (Muresan et al., 2011), it is becom- troduces extra noise to the data (Akhondi et al., ing increasingly important to develop automatic 2019). natural language processing (NLP) approaches en- Most NER systems for the chemical domain abling information extraction from these patents were developed, trained and tested on either chem- (Akhondi et al., 2019). Chemical Named-Entity ical literature or only the title and abstract of Recognition (NER) is a fundamental step for in- chemical patents (Akhondi et al., 2019). There formation extraction from chemical-related texts, are substantial linguistic differences between ab- stracts and the corresponding full text publications detecting common prefixes/suffixes in chemical (Cohen et al., 2010). The performance of NER words. The obtained results show that the per- approaches on full patent documents has still not formance of CRF and SVM models can be sig- been fully explored (Krallinger et al., 2015). nificantly improved by incorporating unsupervised Hence, this paper will focus on presenting the features (e.g. word embeddings, word cluster- best NER performance achieved to date on full ing). The study also showed that the SVM model chemical patent corpus. slightly outperformed the CRF model in the chem- We use a combination of pre-trained word em- ical NER task. beddings, a CNN-based character-level word rep- To perform chemical NER on the CHEMD- resentation and contextualized word representa- NER patents corpus, Akhondi et al.(2016) tions generated from ELMo, trained on a patent proposed an ensemble approach combining a corpus, as input to a BiLSTM-CRF model. The gazetteer-based method and a modified version results show that contextualized word represen- of tmChem. Here, the gazetteer-based method tations help improve chemical NER performance utilized a wide range of chemical dictionaries, substantially. In addition, the impact of the choice while additional features such as stems, pre- of pre-trained word embeddings and tokenizers is fixes/suffixes, chemical elements were added to assessed. the original feature set of tmChem. In the en- The results show that word embeddings that are semble approach, tokens were predicted as chem- pre-trained on chemical patents outperform em- ical mentions if recognized as positive by either beddings pre-trained on biomedical datasets, and tmChem or the gazetteer-based method. The re- using tokenizers optimized for the chemical do- sults showed that both gazetteer-based and ensem- main can improve NER performance in chemical ble approaches were outperformed by the modi- patent corpora. fied tmChem version in terms of overall F1 score, although these two approaches can obtain higher 2 Related work recall. In this section, we summarize previous methods Huang et al.(2015) proposed a BiLSTM-CRF and empirical studies on NER in chemical patents. based on the use of a bidirectional long-short Two existing Conditional Random Field (CRF)- term memory network – BiLSTM (Schuster and based systems for chemical named entity recog- Paliwal, 1997) – to extract (latent) features for a nition are tmChem (Leaman et al., 2015) and CRF classifier. The BiLSTM encodes the input in ChemSpot (Rocktaschel¨ et al., 2012); each makes both forward and backward directions and passes use of numerous hand-crafted features includ- the concatenation of outputs from both directions ing word shape, prefix, suffix, part-of-speech and as input to a linear-chain CRF sequence tagging character N-grams in an algorithm based on mod- layer. In this approach, the BiLSTM selectively elling of tag sequences. A previous detailed encodes information and long-distance dependen- empirical study explored the generalization per- cies observed while processing input sentences in formance of these systems and their ensembles both directions, while the CRF layer globally opti- (Habibi et al., 2016). The application of the tm- mizes the model by using information from neigh- Chem model trained on chemical literature cor- bor labels. pora of the BioCreative IV CHEMDNER task The morphological structures within words are (Krallinger et al., 2015) and the ChemSpot model also important clues for identifying named enti- trained on a subset of the SCAI corpus (Klinger ties in biological domain. Such morphological et al., 2008) resulted in a significant performance structures are widely used in systematic chemical drop over chemical patent corpora. name formats (e.g. IUPAC names) and hence par- Zhang et al.(2016) compared the performance ticularly informative for chemical NER (Klinger of CRF- and Support Vector Machine (SVM)- et al., 2008). Character-level word representa- based models on the CHEMDNER-patents corpus tions have been developed to leverage information (Krallinger et al., 2015). The features constructed from these structures by encoding the character se- in that work included the binarized embedding quences within tokens. Ma and Hovy(2016) uses (Guo et al., 2014), Brown clustering (Brown et al., Convolutional Neural Networks (CNNs) to encode 1992) and domain-specific features extracted by character sequences while Lample et al.(2016) developed a LSTM-based approach for encoding CRF-based models (Section 3.3) with pre-trained character level information. word embeddings (Section 3.4), character-level Habibi et al.(2017) presented an empirical word embeddings (Section 3.5), contextualized study comparing three NER models on a large col- word embeddings (Section 3.6) and implementa- lection of biomedical corpora including the BioSe- tion details (Section 3.7). mantics patent corpus: (1) tmChem–the CRF- based model with hand-crafted features–used as 3.1 Dataset the baseline; (2) a second CRF model based on CRFSuite (Okazaki, 2007) using pre-trained word We conduct experiments on 2 patent corpora: the embeddings; (3) and a BiLSTM-CRF model with BioSemantics patent corpus (Akhondi et al., 2014) additional LSTM-based character-level word em- and Reaxys gold set (Akhondi et al., 2019). beddings (Lample et al., 2016). The performance The BioSemantics patent corpus (Akhondi of CRFSuite-
Load more

Recommended publications
  • Downloaded in MS Word Format (.Docx)
    Wang and Tseng J Cheminform (2019) 11:78 https://doi.org/10.1186/s13321-019-0401-4 Journal of Cheminformatics SOFTWARE Open Access IntelliPatent: a web-based intelligent system for fast chemical patent claim drafting Pei‑Hua Wang1 and Yufeng Jane Tseng1,2* Abstract The frst step of automating composition patent drafting is to draft the claims around a Markush structure with sub‑ stituents. Currently, this process depends heavily on experienced attorneys or patent agents, and few tools are avail‑ able. IntelliPatent was created to accelerate this process. Users can simply upload a series of analogs of interest, and IntelliPatent will automatically extract the general structural scafold and generate the patent claim text. The program can also extend the patent claim by adding commonly seen R groups from historical lists of the top 30 selling drugs in the US for all R substituents. The program takes MDL SD fle formats as inputs, and the invariable core structure and variable substructures will be identifed as the initial scafold and R groups in the output Markush structure. The results can be downloaded in MS Word format (.docx). The suggested claims can be quickly generated with IntelliPatent. This web‑based tool is freely accessible at https ://intel lipat ent.cmdm.tw/. Keywords: Web server, Markush structure, Pharmaceutical patent Introduction of visualizing and analyzing Markush structures from a Claims are the most important sections in composition composition patent [7]. Its web based application, iMa- patents [1]. In the pharmaceutical industry, claims should rVis, revised the underlying R group numbering system include key compounds and all structural derivatives that to deal with nested R group presentation [8].
    [Show full text]
  • The Impact of Academic Patenting on the Rate, Quality, and Direction of (Public) Research Output
    Methodology Data & Measurement Results Summary The Impact of Academic Patenting on the Rate, Quality, and Direction of (Public) Research Output Pierre Azoulay1 Waverly Ding2 Toby Stuart3 1Sloan School of Management MIT & NBER [email protected] 2Haas School of Business University of California — Berkeley 3Graduate School of Business Harvard University January 23, 2007 — Northwestern University Azoulay, Ding, Stuart The Impact of Academic Patenting Methodology Data & Measurement Results Summary Outline 1 Motivation(s) 2 Methodology Problems with existing approaches Selection on observables with staggered treatment decisions Implementing IPTCW estimation 3 Data & Measurement Data sources Measuring “patentability” Descriptive statistics 4 Results The determinants of selection into patenting The impact of academic patenting on the rate of publications The impact of academic patenting on the quality of publications The impact of academic patenting on the content of publications 5 Caveats, Summary & Future Directions Azoulay, Ding, Stuart The Impact of Academic Patenting Methodology Data & Measurement Results Summary Outline 1 Motivation(s) 2 Methodology Problems with existing approaches Selection on observables with staggered treatment decisions Implementing IPTCW estimation 3 Data & Measurement Data sources Measuring “patentability” Descriptive statistics 4 Results The determinants of selection into patenting The impact of academic patenting on the rate of publications The impact of academic patenting on the quality of publications The impact
    [Show full text]
  • Construing Patentability of Chemical Technology Inventions, with Focus
    Construing Patentability of Chemical Technology Inventions, with Focus on their Patent Eligibility and Industrial Applicability: A Comparison on the Patent Examination Approach in the Philippines and in Japan By Anthea Kristine Y. Paculan Intellectual Property Office of the Philippines Supervised by Yorimasa Suwa, PhD., MBA Senior Researcher, Asia-Pacific Industrial Property Center, Japan Institution of Promoting Invention and Innovation Advised by Dr. Kazukiyo Nagai Professor, Department of Applied Chemistry, School of Science and Technology, Meiji University Tokiko Mizuochi, Ph.D Project Assistant Professor, Keio University Office for Open Innovation Final Report In fulfillment of the Study Cum Research fellowship program Sponsored by Japan Patent Office 27 June – 25 October 2019 The views and findings in this report are those of the author and do not necessarily reflect the views and policy of the organization or sponsor of this study. i ○c JPO 2020 Abstract The Intellectual Property Office of the Philippines (IPOPHL) must cope up with the emerging challenges in the patent examination of various fields of technology. The Chemical Technology field at IPOPHL covers a wide array of chemical-related subject-matters which in turn has resulted in handling concerns to the examiners assigned to perform substantive examination on such diverse technologies. Japan Patent Office (JPO) has provided comprehensive guidelines addressing various patentability issues, especially that of patent eligibility and industrial applicability of subject matters in the chemical field. By introducing conceptual aspects of the Japanese patent system as a model, this study allowed the investigation of the similarities and differences in JPO’s and IPOPHL’s examination procedure and assessment of patentability requirements, with focus on patent eligibility and industrial applicability of chemical technology inventions.
    [Show full text]
  • Intellectual Property in the Context of the WTO TRIPS Agreement: Challenges for Public Health MINISTER of HEALTH Humberto Costa
    Intellectual Property in the Context of the WTO TRIPS Agreement: challenges for public health MINISTER OF HEALTH Humberto Costa PRESIDENT OF OSWALDO CRUZ FOUNDATION Paulo Marchiori Buss DIRECTOR OF NATIONAL SCHOOL OF PUBLIC HEALTH Jorge A. Z. Bermudez COORDINATOR OF NUCLEUS FOR PHARMACEUTICAL POLICIES Jorge A. Z. Bermudez Intellectual Property in the Context of the WTO TRIPS Agreement: challenges for public health EDITORS: Jorge A. Z. Bermudez Maria Auxiliadora Oliveira WHO/PAHO Collaborating Center for Pharmaceutical Policies National School of Public Health Sergio Arouca Oswaldo Cruz Foundation A S L U O T R E P O P A P H S O N I O D V I M U N Rio de Janeiro, September 2004 Copyrigth © 2004 ENSP/WHO – Oswaldo Cruz Foundation – FIOCRUZ, Jorge A.Z. Bermudez. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior permission of the National School of Public Health Sergio Arouca. ISBN: 85-88026-16-3 Cover and Graphic Design: Lucia ReginaPantojo de Brito English Translation: Laura Anne Krech Revisions: Claudia Garcia Serpa Osório de Castro Gabriela Costa Chaves Thiago Botelho Azeredo NOTE FROM THE AUTHORS This book presents information produced or compiled by Nucleus for Pharmaceutical Policies professionals and external collaborators regarding the process of implementation of the TRIPS Agreement in developing countries, focusing on its implications for public health policies, particularly those related to access to medicines. Interpretations and views expressed in this publication are solely the responsibility of the authors, not representing any official or institutional position, as to the issues here approached.
    [Show full text]
  • Elsevier Research Platforms
    | 1 | Scopus Trainer : Nattaphol Sisuruk Elsevier Training Consultant, Research Solutions E-mail : [email protected] | 2 | ELSEVIER is a leading Science & Health Information Provider CONTENTPROVISION ‘E’ CONTENT PROVISION RESEARCH MGMT SEARCH & DISCOVERY /PROMOTION TOOLS Niels Louis Alexander Albert George F. John C. Roger D. Craig C Mello Smoot Medicine Bohr Physics Pasteur Fleming Einstein Mather Kornberg Physics Physics (Chemistry) Medicine Physics Chemistry 2 | 3 | Globally recognised high impact content Disseminate Global Elsevier Citations Total STM reference, publication & citations Elsevier share Coverage: Approximately 5,000 publishers Other Get cited Publisher A Publisher B CertifyCertify Investigate Global Elsevier Publications Global References to Elsevier Elsevier Publish Cite Elsevier Other Publisher A Other Publisher A Publisher B 24 Citations Per Paper: Publisher B 27% of all references Global team 2010-2014 74 offices in 24 countries Publisher References Publications Citations 7,000 Journal Editors Elsevier 56,304,346 1,888,115 45,990,748 Publisher A 15,738,334 1,221,036 22,374,220 70,000 Editorial Board Members Publisher B 23,064,330 747,976 18,298,048 600,000 authors Other 116,371,011 5,261,600 95,192,376 Totals 211,478,021 9,118,727 181,855,392 | 4 | Elsevier Research Platforms : Researchers seek a digital environment where ideas can be exchanged, examined, and applied with tools that empower STM knowledge. To find and analyze data from over 5000 publishers Access the leading eBooks and journal articles published by Elsevier Manage your research and showcase your profile via free services :These platforms make data and content easier to search, access, analyze, and share.
    [Show full text]
  • The Concept of Novel Compositions of Matter: a Theoretical Analysis
    rty Right e s: op O r p P e l n a Robson, Intel Prop Rights 2013, 2:1 u A t c c c e l e DOI: 10.4172/2375-4516.1000108 l s e t s n Intellectual Properties Rights: Open Access I ISSN: 2375-4516 ReviewResearch Article Article OpenOpen Access Access The Concept of Novel Compositions of Matter: A Theoretical Analysis Barry Robson* University Director of Research, St Matthews University School of Medicine, The Dirac Foundation, Oxford shire UK and Quantal Semantics Inc., North Carolina, USA Abstract Here is discussed in the manner of a review the nature and uses of information measures in the discipline of patenting. From one perspective, the information content in a patent diminishes rapidly as the broadness of the claims increases. Claims made by Markush representations facilitate the quantification of that. The equations will approach yielding zero information if a massive number of chemical themes were implied. Importantly, a more detailed examination of these equations have implications that allow discussion of various aspects of novelty, reasonable consistency with a specific purpose, and perhaps even how many arguments and counterarguments there should be between examiner and assignee. Keywords: Novel compositions; Patents; Similarity a biological target molecule that provides a complementary surface (albeit influenced by the surrounding environment); if favorable, it Introduction allows specific binding. In principle, the focus on formula hugely helps the discipline of patenting. It reflects the fact that adding, removing, Information from patents changing, or rearranging atoms in a molecule occurs in discrete jumps, By serving to protect intellectual property in exchange for disclosing computationally a matter of integers employed in a clear cut graph- new insight and data for the benefit of science and engineering, patents theoretic description (in contrast, the above field involves continuous are well recognized to be a rich information source [1].
    [Show full text]
  • Bulk Processing of Molecule Patent Associations“
    DIPLOMARBEIT / DIPLOMA THESIS Titel der Diplomarbeit / Title of the Diploma Thesis „Bulk Processing of Molecule Patent Associations“ verfasst von / submitted by Patrick Penner angestrebter akademischer Grad / in partial fulfilment of the requirements for the degree of Magister der Pharmazie (Mag. Pharm.) Wien, 2017 / Vienna, 2017 Studienkennzahl lt. Studienblatt / A 449 degree programme code as it appears on the student record sheet: Studienrichtung lt. Studienblatt / Diplomstudium Pharmazie degree programme as it appears on the student record sheet: Betreut von / Supervisor: Univ.-Prof. Mag. Dr. Thierry Langer Mitbetreut von / Co-Supervisor: Acknowledgements First and foremost, I would like to thank Prof. Dr. Thierry Langer for the opportunity of this project and the many things I learned in the course of it. It was an enlightening experience and a very welcome chance to apply knowledge. Furthermore, I would like to thank Gökhan Ibis for his continued support and guidance in software development. I would also like to thank Dr. Thomas Seidel for his expertise in cheminformatics and the specific pointers he gave me along the way. Miriam Penner deserves mentioning for her graphical design work on the KNIME Node icon. Lastly I would like to thank my patient proofreaders: Katharina Penner, Miriam Penner, Arthur Garon, Clara van Hoey and Markus Wieder. iii Kurzfassung Die Suche nach Molekülen in chemischen Patenten ist schon seit Jahrzehnten eine Heraus- forderung. Die Ungenauigkeit von Moleküldarstellungen, die Extraktion von Strukturen aus Patenten und die große Anzahl veröffentlichter Patente erschweren dieses Unterfangen. Dieses Projekt widmet sich einem bis jetzt wenig beachteten Aspekt der Patentsuche, nämlich einer automatisierten Suche nach Patenten für größere Molekülmengen.
    [Show full text]
  • Patenting Nature: a Problem of History Christopher Beauchamp Brooklyn Law School, [email protected]
    Brooklyn Law School BrooklynWorks Faculty Scholarship Winter 2013 Patenting Nature: A Problem of History Christopher Beauchamp Brooklyn Law School, [email protected] Follow this and additional works at: https://brooklynworks.brooklaw.edu/faculty Part of the Intellectual Property Law Commons, Legal History Commons, and the Litigation Commons Recommended Citation 16 Stan. Tech. L. Rev. 257 (2013) This Article is brought to you for free and open access by BrooklynWorks. It has been accepted for inclusion in Faculty Scholarship by an authorized administrator of BrooklynWorks. STANFORD TECHNOLOGY LAW REVIEW VOLUME 16, NUMBER 2 WINTER 2013 PATENTING NATURE: A PROBLEM OF HISTORY Christopher Beauchamp* CITE AS: 16 STAN. TECH. L. REV. 257 (2013) http://stlr.stanford.edu/pdf/patentingnature.pdf ABSTRACT The practice of patenting genetic material is currently under sharp attack. Recent litigation has forced the courts to grapple with the doctrinal basis for patenting DNA sequences identical to those found in nature. Faced with conflicting authorities and difficult policy questions, courts have leaned heavily on history to guide—or at least to justify—their decisions. This article explores the history in question. It traces the patent law’s changing treatment of “products of nature” in an attempt to untangle the origins of present-day patentability arguments. The evidence suggests that the historical foundations of the bar on patenting products of nature are surprisingly shaky. The article also reveals how isolated biological materials first came to be patented. This task, I argue, requires looking not only to court decisions, but also to the history of patent practice. My principal vehicle for doing so is the case of Parke-Davis & Co.
    [Show full text]
  • Cheminformatics and the Semantic Web: Adding Value with Linked Data and Enhanced Provenance
    Advanced Review Cheminformatics and the Semantic Web: adding value with linked data and enhanced provenance Jeremy G. Frey∗ and Colin L. Bird Cheminformatics is evolving from being a field of study associated primarily with drug discovery into a discipline that embraces the distribution, management, ac- cess, and sharing of chemical data. The relationship with the related subject of bioinformatics is becoming stronger and better defined, owing to the influence of Semantic Web technologies, which enable researchers to integrate hetero- geneous sources of chemical, biochemical, biological, and medical information. These developments depend on a range of factors: the principles of chemical identifiers and their role in relationships between chemical and biological enti- ties; the importance of preserving provenance and properly curated metadata; and an understanding of the contribution that the Semantic Web can make at all stages of the research lifecycle. The movements toward open access, open source, and open collaboration all contribute to progress toward the goals of integration. C " 2013 John Wiley & Sons, Ltd. How to cite this article: WIREs Comput Mol Sci 2013. doi: 10.1002/wcms.1127 INTRODUCTION cipline of bioinformatics evolved more recently, in heminformatics is usually defined in terms of response to the vast amount of data generated by the application of computer science and infor- molecular biology, applying mathematical, and com- mationC technology to problems in the chemical sci- putational techniques not only to the management ences. Brown1 introduced the term chemoinformatics of that data but also to understanding the biological in 1998, in the context of drug discovery, although processes, pathways, and interactions involved.
    [Show full text]
  • IP Analytics and Machine Learning Applied to Create Process Visualization Graphs for Chemical Utility Patents
    processes Article IP Analytics and Machine Learning Applied to Create Process Visualization Graphs for Chemical Utility Patents Amy J. C. Trappey 1,* , Charles V. Trappey 2 , Chih-Ping Liang 1 and Hsin-Jung Lin 1 1 Department of Industrial Engineering and Engineering Management, National Tsing Hua University, Hsinchu 300, Taiwan; [email protected] (C.-P.L.); [email protected] (H.-J.L.) 2 Department of Management Science, National Yang Ming Chiao Tung University, Hsinchu 300, Taiwan; [email protected] * Correspondence: [email protected]; Tel.: +886-3-5742651 Abstract: Researchers must read and understand a large volume of technical papers, including patent documents, to fully grasp the state-of-the-art technological progress in a given domain. Chemical research is particularly challenging with the fast growth of newly registered utility patents (also known as intellectual property or IP) that provide detailed descriptions of the processes used to create a new chemical or a new process to manufacture a known chemical. The researcher must be able to understand the latest patents and literature in order to develop new chemicals and processes that do not infringe on existing claims and processes. This research uses text mining, integrated machine learning, and knowledge visualization techniques to effectively and accurately support the extraction and graphical presentation of chemical processes disclosed in patent documents. The computer framework trains a machine learning model called ALBERT for automatic paragraph text classification. ALBERT separates chemical and non-chemical descriptive paragraphs from a patent for effective chemical term extraction. The ChemDataExtractor is used to classify chemical Citation: Trappey, A.J.C.; Trappey, terms, such as inputs, units, and reactions from the chemical paragraphs.
    [Show full text]
  • Cancer, Patents, and Women's Health." American University Journal of Gender, Social Policy & the Law
    Journal of Gender, Social Policy & the Law Volume 15 | Issue 2 Article 9 2007 Symposium: Molecules and Conflict: Cancer, Petents, and Women's Health Eileen M. Kane Follow this and additional works at: http://digitalcommons.wcl.american.edu/jgspl Part of the Health Law Commons, Intellectual Property Commons, and the Women Commons Recommended Citation Kane, Eileen M. "Symposium: Molecules and Conflict: Cancer, Patents, and Women's Health." American University Journal of Gender, Social Policy & the Law. 15, no. 2 (2007): 305-335. This Article is brought to you for free and open access by the Washington College of Law Journals & Law Reviews at Digital Commons @ American University Washington College of Law. It has been accepted for inclusion in Journal of Gender, Social Policy & the Law by an authorized administrator of Digital Commons @ American University Washington College of Law. For more information, please contact [email protected]. Kane: Symposium: Molecules and Conflict: Cancer, Petents, and Women's H MOLECULES AND CONFLICT: CANCER, PATENTS, AND WOMEN’S HEALTH EILEEN M. KANE* Effective health care for women relies on a nexus of scientific, medical, and legal regimes. Intellectual property law offers incentives for creative accomplishments, and patent law, in particular, offers incentives for the development of medical innovations. This Article examines an intersection of women’s health and the patent system. Breast cancer is the most common cancer in women and has been the focus of sustained basic and clinical scientific research. The Article presents an analysis of patent- related issues that have accompanied the development of leading compounds for the prevention, diagnosis, and treatment of breast cancer, in particular, Taxol, Tamoxifen, Herceptin, and the BRCA1 and BRCA2 genes.
    [Show full text]
  • Topics of Prominence, an Insight Into Identifying New, Emerging Research Trends, Setting the Scene for Technologies of the Future
    6th International Conference on Future-Oriented Technology Analysis (FTA) – Future in the Making Brussels, 4-5 June 2018 TOPICS OF PROMINENCE, AN INSIGHT INTO IDENTIFYING NEW, EMERGING RESEARCH TRENDS, SETTING THE SCENE FOR TECHNOLOGIES OF THE FUTURE Abstract Our methodology aims at identifying the most dynamic research areas worldwide, understanding research trends thanks to a bottom-up approach, and driving to enhanced foresight, providing policy-makers with a well-informed base regarding research trends. ‘Topics of Prominence’ in Science can help answer the following questions: • Which are the research fields with high momentum? • Which stakeholders are leading the way in those Topics? Countries, institutions, researchers, sources? • What are the related Topics adjacent to ones with high momentum? Underlying methodology The creation of Topic Prominence in Science is based upon years of research by Kevin Boyack and Richard Klavans and extensive institutional feedback and testing of SciVal product, User Experience and development teams. Due to recent advances and improvements in quality of the underlying data, the development of better algorithms to cluster scientific papers and greater computing technology, we have now created a highly accurate model able to analyze the whole of the Scopus dataset at once. What is a Topic? Our technology takes into consideration 95% of the articles in ScopusTM and clusters them into 96,000 global, unique research topics based on citation patterns. Scopus publications are clustered into Topics based on a direct citation analysis. A Topic can be defined as a collection of documents with a common intellectual interest. Topics are dynamic and many Topics are multidisciplinary.
    [Show full text]