Semantic Gravitor

Purpose

Finds the semantic gravity for different Part of Speech (POS) tags within a given window. This means that the relatedness value is calculated for all words with other words within the window (relative to the word). Then these values are summed per pos-tag. If instead of a tagrow with !POS-tags a row of words is given then a sum per word (type as in type versus token) will be made.

The configfile is configsemanticgravitor.pl

Synopsis

./semanticgravitor.pl

[-c <corpus>] [-f <fromsubdir>] | [-fd <fromdir>] [-a <tagrowsubdir>] | [-ad <tagrowdir>] [-e <sensedtagrowsubdir>] | [-ed <sensedtagrowdir>] [-tc <tocorpus>] [-t <tosubdir>] | [-td <targetdir>] [-tf <tofile>] [-w <windowsize>] [-ce <ceiling>] [-lo] [-ao [-aw]] [-ps <partsubdir>] [-p <part>] | [-pd <partdir>] [-pf <divprtfile> [...]] [-dr] [-? = -h = -help = --help] [-v [<verboselvl>]]

-c: the corpus whose wordrow files are to be used

-f: subdir below corpus where the wordrow files can be found (defaults to reduced if none is specified and if it's not changed in the configfile)

-fd: full path to the wordrows dir

-a: the tagrowsubdir. Defaults to tagrow. This is the dir in which the !POS- tagrows are to be found for which the semantic gravity-properties should be calculated

-ad: full path to the tagrow dir

-e: the sensedtagrowsubdir. Defaults to sensetagrow. This dir contains the wordnet-sense-tags (they look like #v#1, #n#12, etc...)

-ed: full path to the sensedtagrow dir

-tc: target-corpus. If none is set the source corpus is used

-t: the subdir relative to the corpusdir in which the semantic gravity list should be stored. Note that this option is only possible if the corpus is given with the -c or -tc option (not the full path with the -td option)

-td: full path to the place where the semantic gravity list should be stored

-tf: the file in which the list of semantic gravity data per !POS-tag should be stored

-w: the window-size for disambiguating (same as in wordcombinationfinder.pl, must be even)

-ce: the ceiling to use for relatedness (to prevent skews)

-lo: enable the logarithmic scale (ceiling is applied befor this if enabled)

-ao: all relatedness-values are set to one, usefull for finding the distribution of terms

-aw: all words, not just those for which senses exist. Should only be called with -ao

-ps: subdir in which the divisions are

-p: the name of a division to use as a part. The file-names in all the divprt -files in the division-dir are added and used as the list of files to use unless you specify one or more explicitly using the -pf option

-pd: full path to the division-dir to use as a part. This option causes the -p option to be ignored

-pf: one or more divprt files to use as the part instead of all files in the division

-dr: dry-run. Nothing is written or deleted, only reading and reporting is done

-v: the level of verbosity, default verboselevel = 2, available levels: 0,1,2,3

-?: (and equivalents) prints help: the purpose and the synopsys

NOTE:

If no from and to-dirs are given the defaults in the config file are used

You can download (or look at the sources of) SemanticGravitor [here]. To run it you will also need [the config file] and the [fiauimenre library]. You can also get the entire tool-package (containing the newest version of all fiauimenre tools and the library) [in one download]