mc2skos: A Python repository from Scriptoteket

Python script for converting MARC 21 Classification and MARC 21 Authority records (serialized as MARCXML) to SKOS concepts.

Initially developed to support the project "Felles terminologi for klassifikasjon med Dewey", for converting Dewey Decimal Classification (DDC) records. Issues and suggestions for generalizations and improvements are welcome!

See mapping schema for MARC21 Classification and for MARC21 Authority below.

Installation

Releases can be installed from the command line with pip:

$ pip install --upgrade mc2skos             # with virtualenv or as root
$ pip install --upgrade --user mc2skos      # install to ~/.local

Works with both Python 2.7 and 3.4+. See Travis for details on tested Python versions.
If lxml fails to install on Windows, try the windows installer from from PyPI.
If lxml fails to install on Unix, install system packages python-dev and libxml2-dev
Make sure the Python scripts folder has been added to your PATH.

To directly use a version from source code repository:

$ git clone https://github.com/scriptotek/mc2skos.git
$ cd mc2skos
$ pip install -e .

Usage

mc2skos infile.xml outfile.ttl      # from file to file
mc2skos infile.xml > outfile.ttl    # from file to standard output

Run mc2skos --help or mc2skos -h for options.

URIs

URIs are generated automatically for known concept schemes, identified from 084 $a for classification records and from 008[11] / 040 $f for authority records. To list known concept schemes:

$ mc2skos -l

To add more vocabularies, you can edit vocabularies.yml. Pull requests for adding more vocabularies are very welcome!

URIs can be also be generated on the fly from an URI template specified with option --uri. The following template parameters are recognized:

{control_number} is the control number from 001, 010 or 016. The current approach is to use 010 or 016 if defined, otherwise 001. If you find examples where this approach fails, please add them to [#42](#42).
{collection} is "class", "table" or "scheme"
{object} is a member of the classification scheme and part of a {collection}, such as a specific class or table. Spaces in the URI are replaced by hyphens or another character configured with option --whitespace.
{edition} is taken from 084 $c (with language code stripped)

To add skos:inScheme statements to all records, an URI template can be specified with option --scheme. Otherwise, it will be derived from a default template if the concept scheme is known.

To add an additional skos:inScheme statement to table records, an URI template can be specified with option --table_scheme. Otherwise, it will be derived from a default template if the concept scheme is known.

The following example is generated from a DDC table record:

<http://dewey.info/class/6--982/e21/> a skos:Concept ;
    skos:inScheme <http://dewey.info/scheme/edition/e21/>,
                  <http://dewey.info/table/6/e21/> ;
    skos:notation "T6--982" ;
    skos:prefLabel "Chibchan and Paezan languages"@en .

Mapping schema for MARC21 Classification

Only a small part of the MARC21 Classification data model is converted, and the conversion follows a rather pragmatic approach, exemplified by the mapping of the 7XX fields to skos:altLabel.

MARC21XML	RDF
`001` Control Number (see note above on 001, 010 & 016)	`dcterms:identifier`
`005` Date and time of latest transaction	`dcterms:modified`
`008[0:6]` Date entered on file	`dcterms:created`
`008[8]="d" or "e"` Classification validity	`owl:deprecated`
`010` Control Number (see note above on 001, 010 & 016)	`dcterms:identifier`
`016` Control Number (see note above on 001, 010 & 016)	`dcterms:identifier`
`153 $a`, `$c`, `$z` Classification number	`skos:notation`
`153 $j` Caption	`skos:prefLabel`
`153 $e`, `$f`, `$z` Classification number hierarchy	`skos:broader`
`253` Complex See Reference	`skos:editorialNote`
`353` Complex See Also Reference	`skos:editorialNote`
`680` Scope Note	`skos:scopeNote`
`683` Application Instruction Note	`skos:editorialNote`
`684` Auxiliary Instruction Note	`skos:editorialNote`
`685` History Note	`skos:historyNote`
`700` Index Term-Personal Name	`skos:altLabel`
`710` Index Term-Corporate Name	`skos:altLabel`
`711` Index Term-Meeting Name	`skos:altLabel`
`730` Index Term-Uniform Title	`skos:altLabel`
`748` Index Term-Chronological	`skos:altLabel`
`750` Index Term-Topical	`skos:altLabel`
`751` Index Term-Geographic Name	`skos:altLabel`
`753` Index Term-Uncontrolled	`skos:altLabel`
`765` Synthesized Number Components	`mads:componentList` (see below)

Synthesized number components

Components of synthesized numbers explicitly described in 765 fields are expressed using the mads:componentList property, and to preserve the order of the components, we use RDF lists. Example:

@prefix mads: <http://www.loc.gov/mads/rdf/v1#> .

<http://dewey.info/class/001.30973/e23/> a skos:Concept ;
    mads:componentList (
        <http://dewey.info/class/001.3/e23/>
        <http://dewey.info/class/1--09/e23/>
        <http://dewey.info/class/2--73/e23/>
    ) ;
    skos:notation "001.30973" .

Retrieving list members in order is surprisingly hard with SPARQL. Retrieving ordered pairs is the best solution I've come up with so far:

PREFIX mads: <http://www.loc.gov/mads/rdf/v1#>
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX skos: <http://www.w3.org/2004/02/skos/core#>

SELECT ?c1_notation ?c1_label ?c2_notation ?c2_label
WHERE { GRAPH <http://localhost/ddc23no> {

    <http://dewey.info/class/001.30973/e23/> mads:componentList ?l .
        ?l rdf:rest* ?sl .
        ?sl rdf:first ?e1 .
        ?sl rdf:rest ?sln .
        ?sln rdf:first ?e2 .

        ?e1 skos:notation ?c1_notation .
        ?e2 skos:notation ?c2_notation .

        OPTIONAL {
            ?e1 skos:prefLabel ?c1_label .
        }
        OPTIONAL {
            ?e2 skos:prefLabel ?c2_label .
        }
}}

c1_notation	c1_label	c2_notation	c2_label
"001.3"	"Humaniora"@nb	"T1--09"	"Historie, geografisk behandling, biografier"@nb
"T1--09"	"Historie, geografisk behandling, biografier"@nb	"T2--73"	"USA"@nb

Additional conversion rules for WebDewey data

The script comes with a few extra rules for distinguishing between different types of notes in WebDewey records and extract entities from these. The entity extraction rules (marked with [*] below) utilizes a non-standard namespace and are not enabled by default. Specify the --webdewey flag to use them.

MARC21XML	RDF
`680` having `$9 ess=ndf` Definition note	`skos:definition`
`680` having `$9 ess=nvn` Variant name note	`wd:variantName` [*] for each subfield `$t`
`680` having `$9 ess=nch` Class here note	`wd:classHere` [*] for each subfield `$t`
`680` having `$9 ess=nin` Including note	`wd:including` [*] for each subfield `$t`
`680` having `$9 ess=nph` Former heading	`wd:formerHeading` [*] for each subfield `$t`
`694` having `$9 ess=nml` ???	`SKOS.editorialNote`
`7XX` having `$9 ess=isCaption` Relative index term to use as caption	`skos:prefLabel`

Notes that are currently not treated in any special way:

253 having $9 ess=nsx Do-not-use.
253 having $9 ess=nce Class-elsewhere
253 having $9 ess=ncw Class-elsewhere-manual
253 having $9 ess=nse See.
253 having $9 ess=nsw See-manual.
353 having $9 ess=nsa See-also
683 having $9 ess=nbu Preference note
683 having $9 ess=nop Options note
683 having $9 ess=non Options note
684 having $9 ess=nsm Manual note
685 having $9 ess=ndp Discontinued partial
685 having $9 ess=nrp Relocation
689 having $9 ess=nru Sist brukt i...

Mapping schema for MARC21 Authority

Only a small part of the MARC21 Authority data model is converted.

MARC21XML	RDF
`001` Control Number	`dcterms:identifier`
`005` Date and time of latest transaction	`dcterms:modified`
`008[0:6]` Date entered on file	`dcterms:created`
`065` Other Classification Number	`skos:exactMatch` (see below)
`080` Universal Decimal Classification Number	`skos:exactMatch` (see below)
`083` Dewey Decimal Classification Number	`skos:exactMatch` (see below)
`1XX` Headings	`skos:prefLabel`
`4XX` See From Tracings	`skos:altLabel`
`5XX` See Also From Tracings	`skos:related`, `skos:broader` or `skos:narrower` (see below)
`667` Nonpublic General Note	`skos:editorialNote`
`670` Source Data Found	`skos:note`
`677` Definition	`skos:definition`
`678` Biographical or Historical Data	`skos:note`
`680` Public General Note	`skos:note`
`681` Subject Example Tracing Note	`skos:example`
`682` Deleted Heading Information	`skos:changeNote`
`688` Application History Note	`skos:historyNote`
`7XX` Heading Linking Entries	`skos:xxxMatch` (see below)

Notes:

Mappings are generated for 065, 080 and 083 only if an URI pattern for the classification scheme has been defined in the config.
SKOS relations are generated from 5XX fields if the fields contain a $0 subfield containing either a control number or an URI for the related record. The relationship type is skos:broader if $w=g, skos:narrower if $w=h, and skos:related otherwise. If $w=r and $4 contains an URI, that URI is used as the relationship type. Note that $4 must precede $0 (since both subfields can be repeated).
Mappings/relationships are generated for 7XX headings if the fields contain a $0 subfield containing either the control number or the URI of the related record. If $0 contains a control number, an URI pattern for the vocabulary (found in indicator 2 or $2) must be defined in mc2skos.record.CONFIG. If $4 contains an URI, that URI is used as the relationship type. Otherwise, if $4 contains one of the ISO 25964 relations, the corresponding SKOS relation is used. Otherwise, the default value skos:closeMatch is used. Note that $4 must precede $0 (since both subfields can be repeated).