会员体验
专利管家(专利管理)
工作空间(专利管理)
风险监控(情报监控)
数据分析(专利分析)
侵权分析(诉讼无效)
联系我们
交流群
官方交流:
QQ群: 891211   
微信请扫码    >>>
现在联系顾问~
热词
    • 31. 发明专利
    • Data processing system
    • GB2589631A
    • 2021-06-09
    • GB201917877
    • 2019-12-06
    • ERNST & YOUNG GMBH
    • TILL RAINER WOBBEMICHAEL WIESE
    • G06F16/36G06N20/00G06Q40/00
    • Disclosed is a data processing system comprising a database arrangement (fig.1, 102) and at least one processor (fig.1, 104) in communication with the database. The processor is configured to obtain S1, from the database arrangement, labelled training data 202; process the labelled training data using at least one machine learning algorithm S2, which can be one of decision tree, gradient boosting or neural network based algorithms, to generate a data processing model 204; obtain, from the database arrangement or a user device, at least one test document 206; process the at least one test document using the data processing model to extract key information from the at least one test document S4; and compare the extracted key information 208 with reference key information 210 to identify whether or not there exist inconsistencies between the extracted key information and the reference key information S5, and can optionally produce a summary of this comparison 212. The labelled training data can be made up of a plurality of training documents 202A and a training target document 202B which contains key information of the plurality of training documents. The training data can be financial data, and the test document a financial document.
    • 32. 发明专利
    • A system and method for identifying online rumors based on propagation influence
    • AU2021102006A4
    • 2021-06-03
    • AU2021102006
    • 2021-04-19
    • UNIV BEIJING TECHNOLOGY & BUSINESS
    • DUAN DAGAOHAN ZHONGMINGBAI CHENYU
    • G06F40/30G06F16/36G06N3/08
    • The present invention discloses a method for identifying network rumors based on propagation influence, specifically comprising the steps of: training a neural network on a plurality of texts containing rumor information to obtain a text feature matrix; obtaining a first user node information matrix; constructing a propagation graph structure; constructing a model of publishing user passing attention nodes; training a second user node information matrix and said text feature matrix based on the updated neural network model to obtain a rumor identification model; identify network rumors according to said rumor identification model. The present invention also discloses a network rumor recognition system based on propagation influence, including: a text feature matrix acquisition module, a user node information acquisition module, a first construction module, a second construction module, a training module, and a recognition module; in summary, the present invention effectively captures node information of different dimensions in the forwarding relationship of rumor information and can effectively identify rumor information. - - *4 rumor information. cosrutngaprptsiot grap strctre the nodes insirptutr b*eing a plrlt 4f sai tse nod inomtn an teajcnym riinsi grap stScur bei t Sa fowrdn tn comntn reStosi *Swen plualty f aidtetsby uer f aidruorinfraio.uliain
    • 35. 发明专利
    • ITEM RECOGNITION
    • AU2019201132B2
    • 2020-08-27
    • AU2019201132
    • 2019-02-18
    • ACCENTURE GLOBAL SOLUTIONS LTD
    • HAWKINS CHRISYOUNGDAHL ALLISONNGUYEN DAVID HDANG LEEANN CHAU TUYETNGUYEN DAVID TNAYLOR CHARLOTTEAGASKAR MAYA
    • G06F16/36
    • A computer-implemented method including receiving, from a first device, device identification information for the first device, an image that was captured by the first device, and an account identifier identifying a user account that is associated with the image, determining, using the account identifier, account data for the user account, determining, based on the first device identification information, a particular set of one or more items that are available to patrons of a particular establishment with which the first device is registered, determining, using image recognition which compares the image only with the particular set of one or more items, that the image likely shows a first item in the particular set of one or more items, using the account data and in response to the determination that the image likely shows the first item in the particular set of one or more items, identifying, from among the particular set of one or more items, a second item in the particular set of one or more items that is different from the first item for which a first attribute value for the first item is the same as a second attribute value for the second item, and providing, to a second device, instructions for presentation of information about the second item. Accenture Global Solutions Limited 12587-0612001 1/3____________ _ co C14 Ej -0E I- 0 2 c: o o a) E o = C-) " C ) E w C.) L) c4 0 2 VI cI 2 ---------------- I
    • 37. 发明专利
    • СПОСОБ И УСТРОЙСТВО ИЗВЛЕЧЕНИЯ ТЕМАТИЧЕСКИХ ПРЕДЛОЖЕНИЙ ВЕБ-СТРАНИЦ
    • RU2729227C2
    • 2020-08-05
    • RU2016145396
    • 2016-11-18
    • GUANCHZHOU SHENMA MOBAJL INFORMEJSHN TEKNOLODZHI KO LTD
    • LI CHENYAOTSZEN KHUNLEJ
    • G06F16/36G06F16/951
    • Изобретениеотноситсяк средствамизвлечениятематическихпредложенийвеб-страниц. Техническийрезультатзаключаетсяв повышенииточноститематическихпредложений, извлеченныхс веб-страниц. Получаютвозможныевеб-страницыи предварительнопостроеннуюмодельмашинногообучения, причемкаждаявозможнаявеб-страницасодержитмножествопредварительноотобранныхвозможныхтематическихпредложений, причемкаждоевозможноетематическоепредложениесодержитнесколькословесныхсегментов. Определяютзначениясловесныххарактеристик, указывающиенауровниважностисловесныхсегментовв каждойвозможнойвеб-страницесоответственно, ивводятупомянутыезначениясловесныххарактеристикв модельмашинногообучениядляполучениязначенияважностидлякаждогословесногосегмента. Длякаждойвозможнойвеб-страницыопределяютзначениечастичногопорядкадлякаждоговозможноготематическогопредложенияв соответствиисозначениямиважностисловесныхсегментов, содержащихсяв возможномтематическомпредложении. Длякаждойвозможнойвеб-страницы, выбираютодноизмножествавозможныхтематическихпредложений, связанноесозначениемчастичногопорядка, превышающимпредварительнозаданноепороговоезначение, вкачествецелевоготематическогопредложениявозможнойвеб-страницы. 3 н. и 17 з.п. ф-лы, 6 ил.
    • 38. 发明专利
    • AKTUALISIEREN VON TRAININGSDATEN
    • DE112018005167T5
    • 2020-06-25
    • DE112018005167
    • 2018-09-13
    • IBM
    • YANAGAWA TAKUMIKOMINE HIROAKIMARUYAMA KAORI
    • G06F16/36G06F16/35G06F40/35
    • Trainingsdaten, die eine erste und eine zweite Gruppe von Fragen enthalten, die jeweils einer Antwort zugehörig sind, werden durch einen Computer in einen Speicher gelesen. Es wird eine Klasse von Antworten ermittelt, die für jede Frage die Antwort auf eine Frage enthält, wobei jede Klasse von Antworten eine Klassenbezeichnung hat, die jeder der Fragen zugehörig ist, und jede der Fragen dementsprechend in eine entsprechende Klasse von Antworten klassifiziert wird. Es werden erste und zweite Trainingsdaten erzeugt, die die erste und die zweite Gruppe von Fragen und entsprechende Klassen von Antworten zum Verwenden in ersten bzw. zweiten Klassifikatoren enthalten. Jede Frage aus der ersten und zweiten Gruppe von Fragen wird durch die ersten bzw. zweiten Klassifikatoren klassifiziert, wobei das Klassifizieren entsprechende Klassifizierungsergebnisse erzeugt. Die ersten oder zweiten Trainingsdaten werden auf der Grundlage der Klassifizierungsergebnisse aktualisiert, um entsprechende erste bzw. zweite Trainingsdaten zu erzeugen.