As law is a blueprint of a society and is changed over time as social environments changed, analyzing histories (change provenances) of laws can reveal important facts such as legislative facts and critical events for the society. Linked Open Data (LOD) has emerged as a preferred method for publishing and sharing open data, however, there is an ontological barrier for publishing law history data as LOD. To break through the barrier, this paper proposes an ontology for law history data of the Japanese statute law. The ontology is inspired from PROV-O and SIOC ontologies. The LOD dataset based on the proposed ontology enables wide variety of analyses on the law history data by simple SPARQL queries. The analyses include simple search, visualization, temporal analysis, data mining, etc. This paper displays parts of the analyses which indicate several legislative facts behind changes of laws. The analyses demonstrate the proposed ontology and LOD dataset are useful for legal data analysis. The proposed ontology is comparable with ELI (European Legislation Identifier) which is designed for EU laws, this paper thus discusses the comparability and future directions of the proposed ontology.
Linked Open Data (LOD) [2] has become a de facto standard open data publication methodology and has been proliferated over various domains1 including the legal domain. Laws play a central role in a society and administrative activities are based on the law, therefore, constructing an LOD dataset for the law impacts on the open data movement of not only the legislative and judicial domain but also administrative domains. EUCases2 has published a legal LOD dataset of pan-EU law and EU case law, GovTrack3 offers access to US bills, members of Congress, and so on, and The National Archives, UK provides a legislation ? Copyright c 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). 1 https://lod-cloud.net/ 2 http://eucases.eu/ 3 https://www.govtrack.us/ API4 which gives access to law books at various levels for various times. ELI (European Legislation Identifier)5 is a promising ontology for EU legislations. Typically, laws are classified into statute laws and common laws, this paper focuses on statute laws (hereafter, laws for short) in Japan. The e-LAWS project in Japan attempted to realize an end-to-end open data platform which handles drafts, deliberations, promulgation, etc. of laws, which realized a support system for drafting partial amendment laws in the Japanese government and a database of currently effective laws. The database is partly opened in e-Gov. In fact, e-Gov6 (law text database), the Japanese Law Index7 (law history database) and the Japanese Law Translation Database System8 (English-translated law text database) are separately operated. Thus, LOD for Japanese laws is highly demanded to realize a universal access to legal documents in Japan.
Laws have been changed according to the changes of societies and environments, and it is necessary to manage histories of laws for enabling analyses on changes of both laws and societies. Since a non-retroactive principle for applying laws and traditional measures for enforcing, amending and repealing laws exist, former versions of laws and repealed laws are necessary for lawyers and administrative officers so that desired versions of laws are easily found [6]. There are also analytical demands for changes of laws like legislative fact discovery in legal studies and social science. Histories for individual laws are currently available online. Though histories are indeed solely useful for drafting laws and knowing amendment tendency, more advanced analyses require efforts for data acquisition, data modeling, and analytical processing. Making the law history data as an LOD dataset is beneficial for universal data accesses on law histories. To this end, designing ontologies which capture the histories is necessary. Although ontologies in the legal domain have been increasingly proposed [3, 7, 8], these are mainly for legislations and documents related to laws.
This paper proposes an ontology design for law histories by utilizing design
principles of existing ontologies, PROV-O [5] and SIOC [1]. The basic idea of the
ontology is as follows: (
The proposed ontology design and an LOD dataset based on the ontology enable various analyses for the law history data. Analyzing the original dataset 4 http://www.legislation.gov.uk/ 5 https://publications.europa.eu/en/web/eu-vocabularies/eli 6 http://elaws.e-gov.go.jp/ 7 http://hourei.ndl.go.jp/ 8 http://www.japaneselawtranslation.go.jp/ which is represented by HTML tables is laborious on extract-transform-load process for different analyses. While, the ontology enables universal access to the law history data through a standardized query interface, SPARQL endpoint, and thus makes various analyses easier. This paper reports selected analyses on the LOD dataset, namely, amendment history visualization and classificationbased enactment tendency analysis. These analyses showcase that analyses can be performed by simple SPARQL queries, and suggest prospects for more complicated analyses not only on the dataset but also on inter-connected datasets (e.g., DBpedia) with the dataset as a future vision of the dataset.
Contributions of this paper are summarized as follows: – ELI-comparable Ontology Design for Japanese Law History: A main contribution of this paper is to regard a law change as an event and a series of changed laws as versioned entities. The proposed ontology is comparable with European legislation ontology ELI with advantages, thus this paper discusses the comparability and future directions of the proposed ontology. – Analytical Use Cases: Another contribution of this paper is to showcase the usability of the law history LOD dataset for practical analyses and to suggest prospects for more complex analyses. The practical analyses include a simple analysis (i.e., amendment history visualization) and a data miningbased analysis (i.e., classification-based enactment tendency analysis).
The rest of this paper is organized as follows: Section 2 shows an overview of the Japanese law history and the target data. Section 3 introduces the proposed ontology design for the law history data, and Section 4 showcases selected analyses via a SPARQL endpoint. Section 5 discusses the comparability of the proposed ontology with ELI. Finally, Section 6 concludes this paper. 2
Overview of Law History in Japanese Statute Law This paper targets on Japanese statute laws rather than case laws. Since a statute law is a written law that provisions are described as a body of the law, it is necessary to change the body when to change the provisions. A series of changes in the chronological order of a law is called a history of the law.
In Japanese national statute laws, there are six types of laws currently:
constitution, act, cabinet order, cabinet office ordinance, ministerial ordinance, and
regulation of various governmental organizations. Distinctions among them are
jurisdictions. Japanese statute laws are basically enacted through following five
steps; (
The National Diet Library (NDL) services the Japanese Law Index which
provides a law history search interface. The interface allows users to search laws
by keywords, time periods and categories. Users can observe a law history for
each result law. The history is provided as an HTML table for each law,
therefore, information extraction techniques [4] can be applied to obtain law histories.
(Right of transfer)
Article 95-2 (
Each law is associated with metadata such as a law number, a title, a promulgation date, an enforcement date, and lapse and repeal dates if exist. The law number consists of a law type (e.g., act and cabinet order), a year and a serial number in the year. For example, the law number of the Copyright Act9 is “Act No.48, 1970” 10 in Japan. The title is a name of a law. The promulgation date and the enforcement date are dates when a law is publicized and becomes effective. Note that enforcement dates can be different from promulgation dates when enforcements are delayed for dissemination. The lapse date and the repeal date are dates when the law becomes null and void and is repealed.
A law history is a provenance of a law. Changes in the provenance includes enactment, amendment, repeal, suspend, and lapse. Enactment is to make a new law, which is firstly promulgated and enforced later on. Amendment is to change the body (title, provisions, enforcement date, etc.) of a law. Amendment which fully changes the body of a law is called total amendment and that which 9 English translation of the Japanese Copyright Act in the Japanese Law Translation Database System at http://www.japaneselawtranslation.go.jp/law/detail/?ia= 03&ky=copyright+act&page=24 10 LawID of this act is s45a048. effective period not effective period effective period not effective period Law ℓ" Law ℓ#" Law ℓ$ Law ℓ" Law ℓ#"
Law ℓ$ Promulgation Enforcement Promulgation of Enforcement time of ℓ" of ℓ" amendment law ℓ$ of ℓ$
Promulgation Promulgation of Enforcement Enforcement time of ℓ" amendment law ℓ$ of ℓ$ of ℓ" (a) Amendment after enforcement.
(b) Amendment before enforcement. partially changes the body is called partial amendment. Repeal is to put an end to a law and the law is no more effective. Suspend is to temporally stop the effect of a law. Lapse is that a law becomes null and void and the law is no more effective but the law still exists.
The most complicated change is the amendment which occurs 268 times per year on average. Amendment changes not only texts of laws but also titles and associated dates. Amendment can change titles of laws so as to make titles suited for the bodies of laws. Amendment can change enforcement dates of laws to delay (or hasten) the effects of the laws. Similarly, if the lapse date of a law is described, the lapse date can also be changed. There are, in general, two ways for amending laws: enlargement and consolidation. The former is to add new provisions to the tail of a former version. The latter is to revise provisions in a former version word by word according to provisions in an amendment law, and, in Japan, the latter is adopted as in many countries. Amendment laws can include descriptions for replacement, addition, deletion and so on [6]. Table 1 showcases examples of consolidations (addition of a statement) on the Copyright Act (Act No.48, 1970), where the first row represents an amendment law, the second row shows the related part of the act and the last row shows the amended text of the act (changed parts are highlighted by underlines). The table shows an article about rights to transfer is added to the next of Article 95 in Section 2 of Chapter 4. Note that amendments can be performed repeatedly, that is a new version of a law can be amended by another amendment law.
Due to the presence of the time lag between the promulgation date and the enforcement date of a law, amendment can performed in between these dates. Therefore, there exist provisions in laws which are promulgated but are never enforced. Such provisions are useful for those who analyze legislation. Figure 1 illustrates two amendment situations. The figure assumes that law `1 is enacted and is amended (renamed to `01 for the sake of convenience) by amendment law `2. Figure 1(a) represents `2 amends `1 after `1 is enforced. While, Figure 1(b) shows `1 amends `2 between the promulgation date and enforcement date of `1. In the former case, `1 becomes effective when it is enforced and `01 becomes effective when `2 is enforced. Note that during the period between the promulgation date and the enforcement date of `2, there exists `01 with no effect. In the latter case, `1 has never been effective because `2 is promulgated before `1 is enforced, therefore, there exits a period that the body of `1 is known in public even though `1 has
Regulation s u b c l a s s O f
Law • lawTitle • lawNum • promulgateDate • enforceDate • lapseDate • lawID • ndl:shortTitle • ndl:classification • hasVersion • latestVersion • partialAmend • totalAmend • repeal • suspend • lapse
• versionBeginDate • versionEndDate • previousVersion • generate no effect. Obviously, more complicated amendments can occur (e.g., amendment on amendment law `2), but the explanation for the situation is omitted for the sake of superfluousness. 3
Ontology Design for Law History A basic idea of the proposed ontology for the law history is that changes of laws are regarded as versioning of laws. Keeping versions in a single graph is beneficial, which enables analysing law history data by simple SPARQL queries, meaning that it does not require to specify graphs corresponding with specific versions. However, it is not convenient if versions of laws separately exist in a graph, due to the large number of changes for querying. For instance, when aggregating the number of laws amending specified two laws, grouping law versions by corresponding laws is required. To cope with this problem, the proposed ontology includes an conceptual class for laws which connect with all versions of individual laws. In the ontology, laws for abstraction are defined as class law:Law and versions of laws are defined as class law:LawVersion11.
Figure 2 illustrates a graphical view of the proposed ontology. According to the previous section, there are six classes (i.e., Constitution, Act, CabinetOrder, etc.) which are subclasses of law:Law class. Versioning properties are inspired from PROV-O [5] and SIOC [1] ontologies. PROV-O ontology is designed for data provenance, therefore, it includes generating events (i.e., prov:Activity). In the proposed ontology, changing laws are corresponding with the events. That is, the event that a law changes another law is regarded as generation of a new version (law:LawVersion) of the changed law. SIOC ontology includes version access properties like sioc:previous_version and sioc:latest_version. These are useful to access individual versions of laws and sequences of law versions. For the sake of consistency of naming rules, the proposed ontology include properties equivalent to those in the other ontologies with different names and they are marked by owl:equivalentProperty to represent the equivalences. 11 Namespace law: is under discussion, therefore, concrete URIs will be decided in the near future. hasVersion latestVersion h24a056_v1
generate version BeginDate version
EndDate hasVersion hasVersion latestVersion previousVersion h19a128_v2 previousVersion h19a128_v3
h30a071 2012-08-10 hasVersion versionBeginDate latestVersion versionEndDate 2012-08-10 h30a071_v1
In principle, each law change generates a version of the changed law. For instance, promulgation generates a version of the enacted law itself and amendment generates a version of the amended law. In the proposed ontology, an instance of law:Law is associated with an instance of law:LawVersion which is a version of the law, where the association is represented by law:hasVersion for each version and law:latesrVersion for the latest version. When law `1 which partially amends another law `2 is promulgated, instances of law:Law and law:LawVersion for `1 are generated, then instances of law:Law of `1 and `2 are connected by law:partialAmend and a new instance of law:LawVersion for `2 is generated by the latest version of `1. To maintain the effective periods (discussed in Section 2) and temporal orders of law versions of a law, law:versionBeginDate and law:versionEndDate are associated with versions, and law:previousVersion connects two temporally consecutive versions. The effective period and non-effective period of a law can be calculated using these predicates as follows:
e ective_period = non_e ective_period =
EFD to VED if EFD < VED
none otherwise
VBD to EFD if EFD < VED
VBD to VED otherwise
(
As a result of converting the dataset shown in Section 2 based on the ontology, 571,132 triples are generated. Figure 3 displays a part of the converted data. Due to the space limitation, attributive and systematic properties are omitted. There are three laws, law h19a128, law h24a056 and law h30a071, where the first law is amended by the latter two laws. The following gives an explanation 1 SELECT ?lawVersion ?versionBeginDate ?enforceDate ?versionEndDate 2 WHERE { 3 <lawURI> law:hasVersion ?lawVersion. 4 ?lawVersion law:enforceDate ?enforceDate. 5 FILTER (?enforceDate >= "target_date") 6 OPTIONAL { 7 ?lawVersion law:versionEndDate ?versionEndDate. 8 FILTER (?versionEndDate <= "target_date") 9 } } Listing 1: SPARQL query for determining an effective version of a law at a given date. <lawURI> and target_date are resectvely replaced by a specifiec law URI and a specific date. of the history in a chronological order. Firstly, law h19a128 is promulgated and law version h19a128_v1 is generated. Secondly, law h24a056 is promulgated and amends h19a128. This amendment is represented by the directed arcs, (h24a056, partialAmend, h19a128) and (h24a056_v1, generate, h19a128_v2). Similarly, law h30a071 amends h19a128. All law versions are associated with individual law:versionBeginDate and law:versionEndDate. 4
Law is recognized as a blueprint of a society, analyzing law history data is promising for knowing various aspects of the society. The aspects include states of a society at specific time periods, changes of social blueprints over time, and legislative facts related with law enactments, amendments, etc. The following subsections showcase example analyses of law history data based on the proposed ontology, namely, effective law version detection, amendment history visualization and classification-based enactment tendency analysis. The tail of this section discusses and indicates future analyses with external data sources. 4.1
One of main objectives of law amendment analysis is to determine the effective laws at the specified moment. Laws are retroactive to the moment when an activity relevant to laws is done. For a criminal example, suppose that a robber violates the Penal Code at December 10, 2016 and the robber is arrested two years later, the effective version of the Penal Code at December 10, 2016 must be applied to the robber. Therefore, it is necessary to determine a version of a law that should be applied. The effective version of a law can be obtained by the SPARQL query in Listing 1, where <lawURI> and target_date are placeholders for a URI for the law and a date for examining, respectively. 1 SELECT ?lawVersion 2 WHERE { 3 <lawURI> law:hasVersion ?lawVersion. 4 ?lawVersion law:versionBeginDate ?versionBeginDate; 5 law:enforceDate ?enforceDate; 6 law:versionEndDate ?versionEndDate.
7 } Listing 2: SPARQL query for amendment hisotry visualizations. <lawURI> is a specifiec law URI.
y n60 c e u q e50 r F nt40 e m d n30 e m A e20 v i a10 t l u m Cu0 1952 1962 1972 Y1e9a8r2 1992 2002 2012 (a) Salary of judges.
cyn500 e u q re400 F t emn300 d n e Am200 e v tia100 l u m Cu 0 y c n e50 u q e r F40 t n e m30 d n e m A20 e v i t10 a l u m
Cu0 1906 1926 1946Year1966 1986 2006
1956 1966 1976Yea1r986 1996 2006 2016 (b) Income tax.
(c) Public assistant. A law is a blueprint of a society and the content of the law has been changed over time with respect to the changes of social environments. For example, a Japanese law that minors cannot buy alcohol is enacted because alcohol is considered harmful for the minors. After recognizing an issue that sellers had provided alcohol to the minors without age confirmation, the law is amended so that sellers must confirm ages of buyers with regardless of ages. This example indicates that changes of laws are closely related with social environments and issues.
Analyses with amendment history visualization can indicate social changes in terms of laws. This section include three example visualizations, namely laws about salary of judges (lawID is s23a075) in Figure 4(a), income tax (m32a017) in Figure 4(b), and public assistant (s25a144) in Figure 4(c). The figures illustrate cumulative amendment frequencies over years, where horizontal axes represent years and vertical axes represent cumulative amendment frequencies over years. The frequencies can obtained using the SPARQL query in Listing 2, where <lawURI> specifies a target law.
These examples show that simple visualizations are useful for knowing changes of social environments over time. The first example (Figure 4(a)) represents salaries of judges are constantly changed. This is because the salaries are determined (almost) yearly. The second example (Figure 4(b)) illustrates that the amendment frequency of the law about income tax has drastically changed after World War II (1945). Before the period, the law determines taxes for only people with high income. After the period, the law is changed for applying income tax for all persons, and, due to the increased number of applied persons, amendment frequency has been drastically increased. The last example (Figure 4(c)) is about welfare aids. The example indicates that there are a few amendments on ages for steady economically growth (1973-1991), and the number of amendments has been rapidly increased after the collapse of the bubble economy (1991). This analysis suggests amendment frequencies and economic situations are correlated. 4.3
In contrast to the previous analysis, the classification-based enactment analysis indicates attentions to laws of specific topics. As previously mentioned, laws are enacted based on legislative facts. In the proposed LOD, classifications on laws are included (but limited). The classification is hierarchized, for instance, “Construction/CounterDisaster” represents a two-level law classification about counter disaster in terms of construction, where the top level class is “Construction” and the second-level is “CounterDisaster”. With the associated classifications, the analysis in this section can reveal tendency of enactments in terms of the classifications. The query in Listing 3 obtains the number of enacted laws on a specified classification for each year.
Figure 5 displays three classification-based enactment analyses. The first analysis (Figure 5(a)) is of counter disaster. As soon as disaster occurs, laws to support people adversely affected by the disaster are enacted. A notorious disaster in Japan is the Great East Japan Earthquake in 2011, and new five laws are enacted immediately. Similarly, just after the Great Hanshin-Awaji Earthquake occurred in 1995, two laws are enacted in the same year. Figure 5(b) displays enactment frequencies for laws about urban development. After World War II, Japanese cities need restoration works and the government supports their restorations by a municipal enterprise. However, a few years later, the government decided to shrink the enterprise due to the large financial burden. To resist the decision, several cities appeal to the government to support their restoration, and, as a result, 14 cities are selected for restoration supports. Each law is drafted for each of the 14 cities, therefore, 14 laws are enacted during 1949 to 1951. The last example (Figure 5(c)) is about the government bonds in Japan. The special bond act is a one-year effective law and the act is yearly enacted with different name after 1975. However, after 2012 (right-most part of the figure), the act is no more enacted. This is because the act in 2012 has become a three-year effective law. 4.4
The analyses in this section introduce the capability of the proposed LOD dataset via simple SPARQL queries and the dataset is potentially useful for finding effective versions of laws, analyzing laws themselves, indicating legislative facts and society situations. Analytical results in this section can be associated with known facts like disasters and economic situations. Since the association is done 1 SELECT year(?date) count(?law) 2 WHERE { 3 ?law rdf:type law:Act; 4 ndl:classification "placeholder"; 5 law:promulgateDate ?date. 6 } 7 GROUP BY (year(?date)) 8 ORDER BY (year(?date)) Listing 3: SPARQL query for classification-based enactment analyses. The placeholder is replaced by a specifiec classification name.
Construction/CounterDisaster 5 y c n4 e u q r3 e F t n e2 m t c 1 a n E 0 1950 1960 1970 1980 1990 2000 2010
Year
Construction/UrbanDevelopment 8 7 y c n e6 u q e5 r F tn4 t23 e m c a n E1 0 1950 1955 1960 1965 1970 1975
Year
Finance/GovernmentBonds 4 y c n e3 u q e r F tn2 e m t1 c a n E 0 1900 1920 1940 1960 1980 2000
Year (a) Counter disaster.
(b) Urban development.
(c) Government bonds. manually in this experiment, the experiment shows a prospect that automatic association is promising for more advanced analyses. 5
Discussion: Comparing with ELI ELI (European Legislation Identifier)12 is a legislation ontology for EU laws, which can be regarded as a meta-schema of legal documents. ELI abstracts legal documents by three levels (LegalResource, LegalExpression and Format), and ELI provides meta data for each level as well as relationships among these levels. The top-level, LegalResource, describes abstract concepts of legal documents, the second level, LegalExpression, expresses abstracted contents of the legal documents corresponding with the top level, and Format enables to connect the conceptual models of laws with concrete documents like files and URLs. ELI defines several relationships among legal documents such as amendment, consolidation, citation, etc., and it is possible to extend to include dedicated relationships for different legislation systems.
The proposed ontology in this paper and ELI are comparable. The main
part of the ontologies are similar in terms of the following two points: (
This paper proposes an ontology for law histories in Japan, which can handle multi-versions of laws with change events like amendments. The design of the ontology is based on PROV-O and SIOC ontologies. LOD dataset for Japanese law history based on the proposed ontology enables various analyses in wide range. The analyses shown in this paper illustrates the ability for fundamental analyses (e.g., finding effective versions of laws on specific dates) and advanced analyses (e.g., finding legislative facts based on temporal analyses). These analyses indicate more advanced analyses are expectable. To realize universal ontology design for legal documents, the proposed ontology in compared with a promising ontology, ELI, and the comparison realizes that the interchangeability of the ontologies is an important discussion.
This work was partly supported by JSPS KAKENHI Grant Number JP18H03492.