Applied_Text_Analysis_with_Python_with意思

需积分: 9 44 浏览量更新于2023-05-22 评论 1 收藏 12.17MB PDF 举报

身份认证购VIP最低享 7 折!

领优惠券(最高得80元）

资源详情

资源评论

资源推荐

Chapter1.Languageand

Computation

Applicationsthatleveragenaturallanguageprocessingtechniquestounderstand

human-generatedtextandaudiodataarebecomingfixturesofourlives.These

aretheapplicationsthatcuratethemyriadofhuman-generatedinformationon

thewebspecificallyonourbehalf,offeringnewandpersonalizedmechanismsof

human-computerinteraction.

Wehavegrownusedtotheseapplicationsservinginavarietyofroles,from

spamfiltersthatgroomouremailtraffictosearchenginesthattailorresultswith

personallyrelevantterms.Ironically,whileopportunitiestoinjectlanguage-

basedfeaturesintoapplicationscontinuetomultiply,theemergenceof

developersskilledintheirconstructiondoesnotseemtohavekeptpace.Thisis

atleastinpartbecauseasthesefeaturesbecomeincreasinglyprevalent,theyalso

becomeincreasinglyinvisible.Butitisalsobecausetherisingtideofdata

sciencehasnotyetpermeatedtheprevailingcultureofsoftwaredevelopment.

TheDataScienceParadigm

Thankstoinnovationsinmachinelearningandscalabledataprocessing,thepast

decadehasseen“datascience”and“dataproduct”rapidlybecomehousehold

terms.Theresultingnewroleof“datascientist” — onepartstatistician,onepart

computerscientist,andonepartdomainexpert — hasbecomeoneofthemost

significantjobsofthe21stcentury.

Theworkparadigmthatisgrowinguparoundthepracticeofdatascienceisone

ofresearchandexperimentation;inpartbecausemanydatascientistshave

previouslyspenttimeinpostgraduatestudies,andinpartbecauseoftheprocess

ofdatasciencedevelopmentisnecessarilyexperimental.Asaresult,data

scientistsanddatasciencedepartmentsoftenoperateautonomouslyfromthe

developmentteam,producingbusinessanalyticsforseniormanagement,which

maytheninformchangestothetechnologyorproductstrategyandeventuallybe

baseddataproducts.

Thisbookisintendedtobridgethisgapbyempoweringadevelopment-oriented

approachtotextanalytics.Initwewilldemonstratehowtoleveragethe

availableopensourcetechnologiestocreatedataproductsthataremodular,

testable,tunable,andscalable.Togetherwiththesetools,wehopetheapplied

techniquespresentedinthisbookwillenabledatascientiststobuildthenext

generationofdataproducts.

Inthischapter,wewillbeginbyframingwhatwemeanbylanguage-awaredata

productsandtalkabouthowtobeginspottingtheminthewild.Next,we’ll

discussarchitecturaldesignpatternsthatarewellsuitedtotextanalytics

applications.Finally,we’llconsidersomeoftheuniquechallengesofworking

withtextdata,andprovideashortoverviewofsomeoftheprimaryopensource

librarieswe’llbeusingthroughoutthebook.

LanguageAwareDataProducts

Dataproductsarethosethatderivetheirvaluefromdataandgeneratenewdata

inreturn .Ourviewofappliedtextanalyticsisasthecreationof“language-

awaredataproducts”;user-facingapplicationsthatareresponsivetohuman

inputandcanadapttochange,thatarenotonlyimpressivelyaccurate,but

relativelysimpletodesign.Attheircore,theseapplicationstakeintextdataas

input,parseitintocompositeparts,computeuponthosecomposites,and

recombinetheminawaythatdeliversameaningfulandtailoredendresult.

Oneofourfavoriteexamplesofthisis“YelpyInsights”,areviewfiltering

techniquethatleveragesacombinationofsentimentanalysis,significant

collocations,andsearchtechniquestodetermineifarestaurantissuitablefor

vegetariansorparticularagegroups.Thisapplicationusesarich,application-

specificcorpusformultiplelanguageprocessingcomponentsandrevealsitself

tousersinnaturalandunexpectedways.Forexample,automaticidentification

ofsignificantsentencesinreviewsandtermhighlightingallowrestaurantgoers

todigestalargeamountoftexteasilyandmakeclearerdecisions.Although

languageanalysisisnotYelp’scorebusiness,theimpactthatithasontheir

bottomlineisundeniable.

Anothersimpleexampleofbolt-onlanguageanalysiswithoversizedeffectsis

the“suggestedtag”featureaddedtositeslikeStackOverflow,Netflix,Amazon,

YouTube,andothers.Tagsaremetainformationaboutaquestionorapostthat

areessentialforsearchandrecommendationsandplayalargerolein

determiningwhatcontentisviewedbyspecificusers.Theyalsoactby

identifyingpropertiesofthecontenttheydescribeandcanbeusedtogroup

contentwithsimilar,existingcontentandproposenamedtopicsfortheuserto

selectduringediting.SuchfeatureshighlightthebasicmethodologyofNLP

applications:clusteringsimilartextintomeaningfulgroupsorclassifyingtext

withspecificlabels,orsaidanotherway — unsupervisedandsupervised

machinelearning.

Thegoalofmachinelearningistofitexistingdatatosomemodel(oftencalled

training),creatingarepresentationoftherealworldthatisabletomake

decisionsorgeneratepredictionsonnewdatabasedondiscoveredpatterns.In

practice,thisisdonebyselectingamodelfamilythatdeterminestherelationship

betweenthetargetdataandtheinput,specifyingaformthatincludesparameters

andfeatures,thenusingsomeoptimizationproceduretominimizetheerrorof

themodelonthetrainingdata.Thefittedmodelcannowbeintroducedtonew

dataonwhichitwillmakeaprediction-returninglabels,probabilities,

membership,orvaluesbasedonthemodelform.Thechallengeistostrikea

balancebetweenbeingabletopreciselylearnthepatternsintheknowndataand

beingabletogeneralizesothemodelperformswellonexamplesithasnever

seenbefore.

Machinelearningiscentraltoapplieddatascience.Whileapplicationsthat

performnaturallanguageprocessinghavebeenaroundforseveraldecades

(rememberClippy?),theadditionofmachinelearningenablesadegreeof

flexibility,andthus,responsiveness,thatwouldnototherwisebepossible.By

trainingamachinelearningmodelonanapplication-specificcorpuslike

restaurantreviews,themodelisabletobetterhighlighttermsandmeaningsthat

areassociatedwiththatdomainratherthantheambiguityimpliedbygeneral

language.Modelscanberetrainedonnewdata,targetnewdecisionspaces,and

evenbefitonaper-userbasiswithoutmuchaddedprogrammingeffortsuchthat

theycancontinuetobeusedanddevelopedastheapplicationchanges.

Toadaptandrespondtonewanddifferentcontextsandusers,applicationsthat

implementpredictivemodelsmustbegeneralizableandadaptable.Dataproducts

剩余396页未读，继续阅读

lssc4205

粉丝: 14
资源: 5

会员权益专享

Applied_Text_Analysis_with_Python

评论0

会员权益专享

最新资源

Applied_Text_Analysis_with_Python

评论0

text analytics with python

Text.Analytics.with.Python

Text Analytics with Python 无水印pdf 0分

请用pl/sql 写一个ebs 应付发票核销的脚本

applied spatial data analysis with r second

plot_confusion_matrix python代码

oracle adg查看同步延时

custom_ref_applied: false

运行 python manage.py makemigrations 后，出现 django.db.migrations.exceptions.InconsistentMigrationHistory: Migration users.0001_initial is applied before its dependency auth.0012_alter_user_first_name_max_length on database 'default'. 是哪里问题？如何处理？

matrix analysis and applied linear algebra pdf

Error-[SV-MWC] Missing with clause ../sve/common/common_parser.sv, 365 $unit, "this.u_ai_single_rd_wr_msg.base_addr.find(this.single_rd_data_msg[i][54:48])" Array method 'find' requires a with clause as it cannot be applied directly to the element type.

getBytes in Text cannot be applied

applied plotting, charting data representation in python(umich)| assignment3

会员权益专享

最新资源