3 d2@sdZddlZddlZddlZddlZddlZddlZddlmZddl m Z ddl m Z m Z ddlmZdaddd d d gZdd d ZddZGddde ZGdddZedkrddlmZmZddZeeeZdS)z; Classifiers that make use of the external 'Weka' package. N)stdin) ClassifierI) config_javajava)DictionaryProbDist.z/usr/share/wekaz/usr/local/share/wekaz /usr/lib/wekaz/usr/local/lib/wekacCst|dk r|atdkrt}dtjkr:|jdtjdxh|D]`}tjjtjj|dr@tjj|dat t}|rt dtd|dn t dtt tq@Wtdkrt ddS) NZWEKAHOMErzweka.jarz [Found Weka: z (version z)]z[Found Weka: %s]zUnable to find weka.jar! Use config_weka() or set the WEKAHOME environment variable. For more information about Weka, please see https://www.cs.waikato.ac.nz/ml/weka/) r_weka_classpath _weka_searchosenvironinsertpathexistsjoin_check_weka_versionprint LookupError) classpathZ searchpathr versionr2/tmp/pip-build-v9q4h5k9/nltk/nltk/classify/weka.py config_weka"s$    rcCsfytj|}Wn"ttfk r(Yn dSz$y |jdStk rRdSXWd|jXdS)Nzweka/core/version.txt)zipfileZipFile SystemExitKeyboardInterruptreadKeyErrorclose)jarzfrrrrCs  rc@sbeZdZddZddZddZddZd d Zd d Zd ddddddZ e dgdfddZ dS)WekaClassifiercCs||_||_dS)N) _formatter_model)self formattermodel_filenamerrr__init__TszWekaClassifier.__init__cCs|j|dddgS)Nz-p0z -distribution)_classify_many)r$ featuresetsrrrprob_classify_manyXsz!WekaClassifier.prob_classify_manycCs|j|ddgS)Nz-pr()r))r$r*rrr classify_many[szWekaClassifier.classify_manyc Csttj}ztjj|d}|jj||dd|jd|g|}t |t t j t j d\}}|r| rd|krtt dn t d||j|jtjjd Sx&tj|D]}tjtjj||qWtj|XdS) Nz test.arffz!weka.classifiers.bayes.NaiveBayesz-lz-T)rstdoutstderrzIllegal options: -distributionzOThe installed version of weka does not support probability distribution output.z"Weka failed to generate output: %s )rtempfilemkdtempr r rr"writer#rr subprocessPIPE ValueErrorparse_weka_outputdecoderencodingsplitlistdirremovermdir) r$r*optionstemp_dirZ test_filenamecmdr-r.frrrr)^s2  zWekaClassifier._classify_manycCs2ddtjd|D}tt|jj|}t|S)NcSsg|]}|jrt|qSr)stripfloat).0vrrr sz:WekaClassifier.parse_weka_distribution..z[*,]+)rer9dictzipr"labelsr)r$sZprobsrrrparse_weka_distributionsz&WekaClassifier.parse_weka_distributioncsx0t|D]$\}}|jjdr ||d}Pq W|djdddddgkrbdd|d dDS|djddddd gkrfd d|d dDStjd |drd d|DSx|ddD] }t|qWtd|ddS)Nzinst#ractualZ predictederrorZ predictioncSs*g|]"}|jr|jdjddqS):)rAr9)rClinerrrrEsz4WekaClassifier.parse_weka_output..rP distributioncs&g|]}|jrj|jdqS)rP)rArKr9)rCrQ)r$rrrEsz^0 \w+ [01]\.[0-9]* \?\s*$cSs g|]}|jr|jdqS)rP)rAr9)rCrQrrrrEs zRUnhandled output format -- your version of weka may not be supported. Header: %s) enumeraterA startswithr9rFmatchrr5)r$linesirQr)r$rr6s*    z WekaClassifier.parse_weka_outputz!weka.classifiers.bayes.NaiveBayeszweka.classifiers.trees.J48z#weka.classifiers.functions.Logisticzweka.classifiers.functions.SMOzweka.classifiers.lazy.KStarzweka.classifiers.rules.JRip) naivebayeszC4.5Zlog_regressionZsvmZkstarZripperrZTc Csttj|}tj}ztjj|d}|j||||j krJ|j |} n ||j j kr^|} n t d|| d|d|g} | t |7} |rt j} nd} t| t| dt||Sx&tj|D]} tjtjj|| qWtj|XdS)Nz train.arffzUnknown classifier %sz-dz-t)rr-)rARFF_Formatter from_trainr0r1r r rr2_CLASSIFIER_CLASSvaluesr5listr3r4rrr!r:r;r<) clsr&r* classifierr=quietr%r>Ztrain_filenameZ javaclassr?r-r@rrrtrains*        zWekaClassifier.trainN) __name__ __module__ __qualname__r'r+r,r)rKr6r] classmethodrcrrrrr!Ss ,1r!c@sVeZdZdZddZddZddZdd Zed d Z d d Z dddZ ddZ dS)r[z Converts featuresets and labeled featuresets to ARFF-formatted strings, appropriate for input into Weka. Features and classes can be specified manually in the constructor, or may be determined from data using ``from_train``. cCs||_||_dS)a) :param labels: A list of all class labels that can be generated. :param features: A list of feature specifications, where each feature specification is a tuple (fname, ftype); and ftype is an ARFF type string such as NUMERIC or STRING. N)_labels _features)r$rIfeaturesrrrr'szARFF_Formatter.__init__cCs|j|j|S)zBReturns a string representation of ARFF output for the given data.)header_section data_section)r$tokensrrrformat szARFF_Formatter.formatcCs t|jS)zReturns the list of classes.)r_rh)r$rrrrIszARFF_Formatter.labelscCs0t|dst|d}|j|j||jdS)z.Writes ARFF data to a file for the given data.r2wN)hasattropenr2rnr)r$outfilermrrrr2s  zARFF_Formatter.writecCsdd|D}i}x|D]\}}x|jD]\}}tt|trFd}nFtt|tttfr`d}n,tt|trtd}n|dkrq*n td||j|||krtd||||<q*WqWt |j}t ||S) z Constructs an ARFF_Formatter instance with class labels and feature types determined from the given data. Handles boolean, numeric and string (note: not nominal) types. cSsh|] \}}|qSrr)rCtoklabelrrr !sz,ARFF_Formatter.from_train..z {True, False}ZNUMERICSTRINGNzUnsupported value type %rzInconsistent type for %s) items issubclasstypeboolintrBstrr5getsortedr[)rmrIrjrsrtfnamefvalftyperrrr\s$   zARFF_Formatter.from_traincCsXd dtj}|d7}x"|jD]\}}|d||f7}q W|dddj|jf7}|S) z#Returns an ARFF header as a string.z% Weka ARFF file z"% Generated automatically by NLTK z%% %s z@RELATION rel z@ATTRIBUTE %-30r %s z@ATTRIBUTE %-30r {%s} z-label-,z3% Weka ARFF file % Generated automatically by NLTK )timectimerirrh)r$rJrrrrrrk9s zARFF_Formatter.header_sectionNcCs|dkr|ot|dttf}|s0dd|D}d}xN|D]F\}}x*|jD] \}}|d|j|j|7}qJW|d|j|7}q:W|S)a Returns the ARFF data section for the given data. :param tokens: a list of featuresets (dicts) or labelled featuresets which are tuples (featureset, label). :param labeled: Indicates whether the given tokens are labeled or not. If None, then the tokens will be assumed to be labeled if the first token's value is a tuple or list. NrcSsg|] }|dfqS)Nr)rCrsrrrrE]sz/ARFF_Formatter.data_section..z @DATA z%s,z%s ) isinstancetupler_ri _fmt_arff_valr})r$rmZlabeledrJrsrtrrrrrrlNs zARFF_Formatter.data_sectioncCs@|dkr dSt|ttfr"d|St|tr4d|Sd|SdS)N?z%sz%r)rrzr{rB)r$rrrrrhs zARFF_Formatter._fmt_arff_val)N) rdrerf__doc__r'rnrIr2 staticmethodr\rkrlrrrrrr[s   r[__main__)binary_names_demo_features names_democCstjd|dS)Nz/tmp/name.modelzC4.5)r!rc)r*rrrmake_classifiervsr)N)rr rFr3r0rrsysrZnltk.classify.apirZnltk.internalsrrZnltk.probabilityrrr rrr!r[rdZnltk.classify.utilrrrrarrrr s2    !$}