Solving the Problems of Finding Law on the Web: World Law and DIAL

Graham Greenleaf ,
 project leader
graham@austlii.edu.au

Daniel Austin , Philip Chung
and Andrew Mowbray
Software Authors & System Developers

Madeleine Davis and Jill Matthews
Legal Indexers

Australasian Legal Information Institute ( AustLII )

This is a Refereed Article published on 29 February 2000.

Citation: Greenleaf G et al, 'Solving the Problems of Finding Law on the Web: World Law and DIAL', 2000 (1) The Journal of Information, Law and Technology (JILT). <http://elj.warwick.ac.uk/jilt/
00-1/greenleaf.html>.New citation as at 1/1/04: <http://www2.warwick.ac.uk/fac/soc/law/elj/jilt/ 2000_1/greenleaf/>.

1. The Potential - A World-Wide Law Library on the Internet

Despite its recent development, the Web already contains an astonishing variety of legal materials from dozens of countries[ 1 ]. Significant collections of legislation are already available on the Web from over 50 countries. The full text is available on the Web of all legislation from almost all the jurisdictions of the USA, Canada, Australasia, many Latin American countries and some European countries (such as Norway and Germany), and extensive collections from many other European counties (such as the United Kingdom, France, Spain, Portugal). Substantial collections of legislation are available from many developing countries, including India, Turkey, Kazakhstan, South Africa, Vietnam, Zambia, China, Mexico and Israel.

There are also extensive collections of case law from about 20 countries, particularly from North America and Australasia and some European courts, but also courts from India, Korea, Brazil and other countries. The Parliaments of dozens of countries have Web pages, and these contain many significant resources concerning legislation and law reform. Law reform commissions and similar bodies are starting to make their reports and working papers available via the Web. There are specialist university and other centres which provide very large specialist collections of materials in areas such as constitutional law, trade law, the law of the sea and human rights.

Despite the abundance of valuable legal materials already on the Web, and the rapidity with which these materials are expanding, these materials are often very difficult to find, since they are scattered across thousands of Web sites located all around the world. As we shall see, the tools we have used for internet legal research until now do not serve us well enough.

1.1 Importance to Developing Countries

The Internet is still dominated, in terms of both location of Web sites and location of users, by the developed world of North America, Europe and Australasia. The preponderance of English-language information on the Web is in part a reflection of this.

However, the availability of legal information on the World-Wide-Web is of considerable importance to developing counties, in Asia and elsewhere. Law libraries with reasonably comprehensive and up-to-date collections of legislation, case law and law reform reports, are virtually unknown in the developing world, and the costs of maintaining them are prohibitive. Access to large online commercial services such as Lexis is also prohibitive for most lawyers in developing countries. This is the principal reason that the Asian Development Bank has funded the development of Project DIAL (Development of the Internet for Asian Law) (one subject of this paper): to provide a means by which legislative and law reform personnel can obtain access to comparative legislative and law reform models. It is of equal importance for superior Courts in developing countries to have access to the decisions of similar Courts throughout the world. It is not only the legislation or case law of the developed countries that is of interest and importance: a legislator in Mongolia might find their best model for a Buy-Operate-Transfer (BOT) law from a Kazakhstan Web site, and the Supreme Court of a small Pacific Islands country is likely to find the most important precedents for its decisions in the decisions (currently almost completely unobtainable) of similar Pacific Island states.

Legal information on the Internet is also important to developing countries for a quite different reason: the provision via the Internet of comprehensive and up-to-date information about a country's laws and legal system can be a striking demonstration of the transparency of a country's legal system. This is likely to be of considerable importance to potential foreign investors, both in symbolic terms and in facilitating efficient and low cost legal advice necessary for investment decisions to be made. A recent example is the proliferation of government agency Web sites in Mongolia which provide the legislation that they administer[ 2 ]: it would be very difficult and expensive to obtain that legislation by any other means.

2. The Problem - Internet Legal Research is Difficult

2.1 Two Types of Tools - Catalogs and Search Engines

There are essentially only two types of tools which help users find legal materials on the internet, commonly called catalogs and search engines.

Catalogs are where individual Web sites are classified by hand according to various classificatory schemes. Because of the human intelligence and effort involved, they are also called 'intellectual' indexes. Usually, such indexes only provide the title, URL[ 3 ] and perhaps a brief description of each site indexed. Yahoo! is a well known example of an internet-wide catalog or intellectual index of the Web (i.e. one which is not law-specific).

'Search engine', in the Web context, has become a shorthand way of referring to the combination of a 'Web robot' and the data it gathers, and the text retrieval software used to search that data. A program (variously called a 'Web robot' or 'Web spider') traverses the Web, downloading every page it encounters, so that every word on every page can be converted into one very large word occurrence index ('concordance') which can be searched by the text retrieval software. They can also be called 'automatic indexes'. When the search engine displays a URL as a result of a search, that URL is to the original site, not to a mirror on the remote site. Alta Vista , Northern Light , Hot Bot , Excite and Infoseek are well-known 'general purpose' search engines that search an index created by a Web spider, and where the Web spider goes to all types of subject matter. They have only existed since Alta Vista 's creation in 1996. The principle advantage of this approach that it is possible to search every word that has been indexed, not just the titles and brief summary of what is on the site. About 85% of internet users use search engines to locate information ( Lawrence and Giles 1999 ).

Combinations of catalogs and search engines on the one site are now becoming more common, and such combinations are often referred to as 'portals'. Some catalogs are now attempting the automated or semi-automated classification of new Web sites located by a Web spider (for example, Excite Australia ). This approach is potentially promising but has not yet been shown to produce useful systems for legal research.

Despite the existence of these research aids, finding legal information on the internet is surprisingly difficult, partly because neither catalogs nor search engines used alone can provide a satisfactory solution. This is particularly in relation to internet-wide (i.e. not law-specific) catalogs and search engines. These difficulties will now be summarised[ 4 ].

2.2 Catalogs Alone are Insufficient

Good catalogs (intellectual indexes) for law are hard to find[ 5 ]. While there are many multi-country intellectual indexes to law on the internet[ 6 ], none are even remotely comprehensive, and many are US-oriented with only a slight international gloss. Many are updated only rarely if at all. Some very good indexes do exist for particular countries (e.g. Canada, the USA, Germany and Australia), and many exist for particular subject matter areas, but they are often difficult to find from the multi-country indexes. It is therefore difficult to find a good place to start. The coverage of legal materials in general-purpose internet indexes is no more helpful, as an inspection of the limited coverage of legal materials in an index such as Yahoo! (the largest general-purpose catalog) will show .

Catalogs are hard to maintain. As the quantity of legal material on the internet grows, the sites that contain significant legal information grow so numerous, and some are so large, that it is difficult to maintain catalogs at all, and particularly to maintain them with any depth of indexing of each site. The best that can be hoped for is that sites with significant legal materials are identified in the index, even though there is no detailed description of their content. For example, it soon becomes impossible to include in a catalog the content of each piece of legislation, each case, or each journal article included on a large site.

As a result, we can say that catalogs are inherently shallow - even when they are good at identifying important law sites, they cannot index very deeply into those sites.

2.3 General Purpose Search Engines Alone are Insufficient

The main problem of general purpose search engines for legal research is their lack of comprehensiveness, but there are numerous other problems as well[ 7 ].

General search engines are not comprehensive. There are very good internet-wide search engines, but they are nowhere near as comprehensive as people often assume. A July 1999 report in Nature by Lawrence and Giles ( 1999 ) shows that of eleven major search engines tested (including Alta Vista , Northern Light , Hot Bot , Excite and Infoseek ) the search coverage they provided of the estimated 800M pages on the Web was 16% at best (Northern Light) and down to 5.6% ( Excite ), with some not named here even lower.

There are a number of reasons, other than the rapidly expanding size of the Web, for the lack of comprehensiveness of search engines based on automated Web spiders. They include the following[ 8 ]:

• Some robots only index a sample of pages on a particular site (at least at any one time), and do not continue indexing until they complete all pages on a site in one session[ 9 ].
   

• Well-behaved robots[ 10 ] adhere to the robot exclusion standard[ 11 ], by which Web servers tell robots which pages they may not index on a site. Because of the effects of some robots on server performance, and for other reasons, many servers exclude robots. All major search engines observe robot exclusions.
   

• 'Non-Web' databases are often accessible on the Web behind search forms, but the underlying databases cannot be indexed by robots as they are not in the form of HTML pages.
   

• There are some technical problems with frames and with dynamically created Web pages that mean that they cannot be included in Web spider indexing.

To make matters worse, the percentage of the available Web pages that the search engines are indexing is declining, not increasing, due to the rapid expansion of the Web from an estimated 320 M pages in December 1997 to 800 M pages in February 1999[ 12 ]. The decline in the best search engine was from about 33% to 16% of the estimated total size of the Web, a dramatic drop. It seems that the Web spiders of general purpose search engines simply cannot keep up with the expansion of the Web. The technical problems caused by such expansion may also be exacerbated by diminishing financial returns from the costs of trying to keep up. Lawrence and Giles suggest:

The combined coverage of the search engines tested by Lawrence and Giles was estimated at 42% of the total number of Web pages, a decline from 60% in their 1997 study. They agree that use of meta-search engines such as MetaCrawler will provide a more comprehensive search. However, it seems from their research that the best possible coverage would still only be about 42%, and even then this would depend on maximum coverage and sorting efficiency of the meta-search engine, so the real figure is likely to be somewhere between 16% and 42%.

There are other significant problems with general purpose search engines which stem from the ambitious nature of their task of indexing all information on the Web irrespective of its subject matter:

• General search engines are often out of date Lawrence and Giles concluded from their tests that there was 'evidence that indexing of new or modified pages can take several months or longer'. As a result, there may be new pages added in that period that are not indexed and searchable, even if the site is included in the particular search engine's coverage.
   

• General search engines contain too much 'noise' It is difficult to make searches precise enough to find only legal materials using general search engines, because they index predominantly non-legal material. It is usually necessary to try to impose some ad hoc search limitation (in addition to the real search terms) such as 'law or legislation or code or court' or some such, to try to stem the flood of irrelevant information (or more likely, to fool the relevance ranking into putting legally oriented material first).
   

• General search engines are difficult to search for law from particular countries It is difficult for most users to limit searches to materials concerning laws of particular countries[ 13 ], and failure to do so will usually result in the search being flooded with material from North America and other 'content rich' parts of the internet.
   

• General search engines are biased in favour of 'popular' pages The way in which the Web spiders for general search engines work is that they find many of the pages they index simply by following hypertext links from other pages. This likely bias that search engines would have in favour of finding pages which have numerous hypertext links to them has been confirmed: there is a direct relationship between the probability of a page being indexed and the number of links to that page from other pages ( Lawrence and Giles 1999 ). New types of search engines are exacerbating the problem as they take into account in their measurement of relevance ranking the number of other pages that link to the page concerned[ 14 ]. Lawrence and Giles conclude 'For ranking based on popularity, we can see a trend where popular pages become more popular, while new, unlinked pages have an increasingly difficult time becoming visible in search-engine listings. This may delay or even prevent the widespread visibility of new high-quality information.'
   

• General search engines do not provide unbiased world-wide coverage Given the limited coverage of general search engines, their bias in favour of materials which are hypertext linked from other pages, and the fact that existing catalogs of legal materials on the internet are very heavily biased in favour of materials from developed countries (particularly North America) it seems very likely that general search engines are likely to be biased in favour of materials in English, and materials from developed countries.
   

• Some general search engines are selling priority places in their relevance ranking . This approach distorts the reliability of search engines, as the 'top ranked' sites may not be those which contain the most relevant materials.

The conclusion we may draw from all of these problems is that it would be unrealistic to expect general purpose internet-wide search engines to provide a very effective method of a task as specialised and 'non-popular' as internet legal research.

2.4 Research Problems with Individual Web sites

Many significant law sites can't be searched When you do find a site containing valuable legal information it will often not have a search engine at all, so searching at word level is not possible. Of the more than 30 internet sites around the world containing significant quantities of legislation, less than half have any search engine[ 15 ]. It requires considerably greater technical ability to run a search engine than it does to simply put pages of legal material onto the internet where they can be browsed.

Using different search engines can be confusing Even if a law site does have its own search engine, users who wish to find legal materials on different sites can also be easily confused by the need to use different search engines with different search commands.

2.5 An Unsatisfactory Result

So the problems of finding legal materials world-wide are that it is both difficult to find which useful sites exist for a particular country or subject, and also difficult to find what is on such sites as are known. These research problems are very substantial even for the most expert 'internet savvy' lawyers and law librarians. They are much worse for inexperienced users.

3. AustLII's New Solution - World Law and DIAL

The challenge is to find a new approach to legal research on the internet which will provide an internet-wide (which means world-wide) method of effectively providing access to legal materials available on the Internet, no matter where they are located.

The answer we propose is in part a technical solution, a limited area search engine for law, but to a large extent the success of the technical component will depend on an organisational element, the creation of a multi-national group of collaborators who are willing to make joint use of the technical tools we have developed in order to create and sustain a world-wide legal research facility.

3.1 A New Technical Solution - A Limited Area Search Engine for Law

Our approach to reducing the problems of internet legal research rests on these propositions:

• A catalog created by intellectual indexing effort is essential to identify high value law sites and legal resources, but cannot and should not aim to be comprehensive in its depth of indexing particular sites once they have been identified.
   

• Web robot indexing of remote law sites, and a sufficiently powerful search engine, are necessary to provide the depth of search capacity that intellectual indexing cannot provide;
   

• This is particularly so when many significant law sites do not have search engines at all, and where there is no consistency among the search engines used by those that do.
   

• Searching robot indexed sites will work much better if (i) only law sites are indexed (to remove non-legal 'noise' and improve precision); and (ii) such sites are indexed comprehensively (to improve recall). We call the combination of such a Web spider (a 'targeted' Web spider) and search engine dedicated to legal materials only a 'limited area search engine'.
   

• Significant law sites which normally exclude robots may allow a 'targeted' law-oriented Web spider to make them searchable, by request. The number of requests may be manageable.
   

• A comprehensive catalog is needed to identify the law sites worth indexing, and therefore to 'target' the robot. The intellectual index therefore serves the double function of a useful resource in itself, and the essential means of 'feeding' the search engine.
   

• A versatile method of effectively limiting the scope of searches to particular types of subject matter (legislation, case law, laws from a particular country, laws on a particular subject) is needed to improve the precision of legal research, as the amount of legal materials over which searches will be conducted is still very large.
   

• Once a law-oriented Web spider has created a searchable index of key law sites, specific searches over that index for various types of subject matter can be 'embedded' in the intellectual index, thereby making the intellectual index 'self-updating' to a certain extent, and so reducing its maintenance costs. Such 'embedded searches' also cater for inexperienced users who have difficulty in formulating searches.
   

• Maintenance of both the catalog and the validity of the web spider targeting (both of which can degrade due to page movements) must be automated as far as possible.

The key to effective legal research on the internet may therefore be a tight integration of an intellectually created catalog and a search engine based on a Web spider, a symbiotic relationship in which each builds on the features provided by the other[ 16 ]. Intellectual indexing and automated indexing can feed off each other.

The following diagram explains this relationship, from the perspective of the indexer and the user.

Interactions in the use of a World Law/DIAL's limited area search engine

3.2 AustLII's Tools - Web Indexing, Web Spider and Search Engine Software

AustLII personnel[ 17 ] have developed software and systems to implement this approach:

• internet indexing software ('Feathers') which allows remote updating by multiple contributors to an index, full search facilities over the index entries, and a facility to 'target' a robot to fully index specific sites identified in the index[ 18 ];
   

• a robot or 'Web spider' (called 'Gromit'), and a 'harness' or means of controlling it (called 'Wallace'[ 19 ]), which only indexes those sites or parts of sites to which it is sent (by the indexing software), and does not stray 'off site' to index unwanted material[ 20 ].
   

• a search engine (SINO), which has the full range of Boolean and proximity search commands, optional relevance ranking of search results, and a facility for limiting the scope of searches to specific databases or collections of databases[ 21 ].

3.3 International Indexing Partnerships - A Necessary Element

The approach we are taking requires the development and maintenance of a world-wide catalog created by intellectual indexing effort, which at least catalogs significant national law sites in all countries world-wide and the major subject-oriented resources. A free access law facility such as AustLII, even if it does obtain significant funding support for the task, is unlikely to be able to support more than a couple of legal indexing staff to undertake the task of adding content. The human languages that such staff are conversant with will be necessarily limited. While it is possible to provide a basic level of world-wide coverage with such resources, as can already be seen in the World Law / DIAL ' Countries ' page, this will not be fully satisfactory unless regional, national or subject experts are also involved in the indexing process.

Consequently, the way in which we are now envisaging the World Law / DIAL project is that, while most of the organisation of the content indexing and much of the substantive indexing will be carried out by AustLII staff, wherever possible we will invite appropriate authorities from other countries, and other language and subject specialists, to participate in the cataloging process as 'indexing partners' for particular parts of the catalog.

There are a number of ways in which such partnerships can proceed, ranging from a partner periodically emailing lists of proposed URLs to be added to our pages to AustLII's indexers (who would add them to the relevant pages and send the Web spider); to our indexers checking a partner site regularly for any additional links which should be added to our pages; to a remotely located indexer being provided with password access to the editing interface to World Law such that they can add links and change sub-categories in 'their' parts of the index but not elsewhere. The editing interface of World Law is via the Web, so contributing editors can be located anywhere with Web access.

Wherever collaboration occurs, it will be acknowledged on those pages of the catalog where the collaboration is based, by inclusion under the 'Contributor' heading of the logo (or name) of the contributing organisation or person and a link to their Web site. For example, Australia's Department of Foreign Affairs and Trade contributes content to the creation of our 'Treaties and International Agreements' pages, which is acknowledged as shown below.

Some of the advantages of this approach for our indexing partners are that they will obtain additional users through referrals from the World Law pages, and if they wish they will also be able to place CGI search interfaces on their own pages which will search over the full texts indexed by our Web spider, but only the part of AustLII's index relevant to them.

We are only now starting this process of creating partnerships, so it is too early to report on its success. Our first international partnership is being established between Ralph Amissah's Lex Mercatoria site in Norway and World Law's International Trade pages. Other partnerships are being negotiated.

3.4 A Brief History of World Law and Project DIAL

AustLII had been developing a catalog of Australian law sites since July 1995, with some international indexing. The index was maintained as hand-created HTML pages until 1997, at which point it was moved into a mSQL database, and it became possible to search for individual catalog entries. At this point the appearance and functionality was similar to Yahoo!'s approach. In early 1997 Web spider software was customised, and the internet indexing software rewritten so that it could be used to 'target' the Web spider.

The first opportunity for extensive testing of the targeted Web spider was in Project DIAL, a project for the Asian Development Bank which aims to increase the accessibility of legislation on the internet. The Web-spider search facility, Greenleaf et al 1997 ). The more general version of the targeted Web spider was made available from AustLII's home page as 'World Law Search' in February 1998, when non-legislative material started to be added by the Web-spider. Two test 'Libraries' were added to allow searches to be restricted to legislative materials and to other internet law indexes - the precursor of the search limitation facilities which have now been added.

In June 1999 the Asian Development Bank entered into a three year agreement for AustLII to develop and host the full-scale Project DIAL facility, as a Regional Technical Assistance (RETA) of the Bank. The value of the RETA is approximately A$1 M, with nearly a half of those funds being used to assist and train users in the Bank's Developing Member Countries (DMCs) of the Bank.

During 1999 there has been a major redevelopment of the search facilities in World Law / DIAL, so that the scope of searches can be limited by the location in the catalog from where the user does the search. This enables users to limit their search scope (thus increasing precision) without having to understand any search commands in order to do so. It is the major single technical innovation in World Law / DIAL.

3.5 Other Limited Area Search Engines

AustLII's World Law Search is one of the first and one of the few examples of what we call 'targeted Web spiders' and others have called 'Limited Area Search Engines'. There are a couple of examples of this technology being used in the field of law.

JURIST launched in January 1998 what it calls a 'Limited Area Search Engine (LASE)' to make searchable all home pages of law Professors and course pages of law subjects in its index.

JURIST Search page

JURIST uses the Argos LASE which, according to its developers, was 'the first peer-reviewed, limited area search engine (LASE) on the World-Wide Web' at the time of its release in October 1996. Argos was developed to provide a more precise means of searching for scholarly literature on the ancient and medieval world. It's development was prompted by the same shortcomings with Internet-wide search engines as we have identified in the preceding discussion:

They identify advantages of the alternative, 'targeted' approach:

They also note that, because of its limited scope, it is possible for Argos to update weekly the indexes to all the sites it covers, rather than the couple of months that (in their experience) are taken by Internet-wide search engines to update sites. They estimate that this means that 98% of all links found by their search engine work at any given time.

Another example is that Knowledge Basket, the New Zealand Internet content provider, released Legal Search New Zealand in December 1997. It uses the Verity Web robot and Topic search engine, and makes searchable 25 New Zealand law sites at present. The advantages they claim for this approach are similar to those described above < http://www.knowledge-basket.co.nz/kete/nzsearch.html >.

4. Catalog Structure of World Law/DIAL

World Law Index, the intellectual index or catalog aspect of World Law has a conventional 'Yahoo-like' interface. The 'World' root page of the index is shown below, and all other sub-categories are located in a hierarchy under 'World'.

4.1 Content in the Catalog

Approximately 4,000 law sites are indexed in the catalog at present, with most sites indexed under a number of sub-categories.

Some features of the catalog structure are:

• There is a page for every country in the world, which at the very least will contain an embedded search (see later) which will find at least 190 legal documents referring to even the smallest country (for example, 193 for the Faroe Islands; 239 for Niue; 385 for the Cayman Islands; 771 for Kyrgyzstan; 907 for Burkina Faso).
   

• The hierarchical structure of the catalog is organised to a large extent around indexing by country, with each 'country page' having a fairly standard structure of sub-categories ('Legislation', 'Courts', 'Education', 'Other Indexes' etc). Other cross-indexing from the 'World' page (for example, on the 'World >> Legislation' page) is essentially by cross-references to these sub-categories of the country pages. Such cross references are apparent by the '@' symbol following a sub-category.
   

• There is legislation from more than 60 countries accessible directly from the 'World >> Legislation' page, as illustrated in the extract below. Each link goes directly to the 'Legislation' sub-node of that country's page. A similar cross indexing structure is provided for Courts and Case Law, Education, Law Reform etc.

• Most sites and parts of sites indexed in the catalog are subject-indexed (where relevant) as well as indexed by type or source of content. The 'World >> Subject Index' page indexes sites by over 40 subject categories, and the 'Australia >> Subject Index' (which has been under development longer) has over 70 categories.

Extract of categories from the 'Australia >> Subject Index' page

4.2 Navigating to Other Places in the Catalog (The Hierarchy)

Every catalog page lists its hierarchical location in the catalog. Click on any point in the hierarchy to go back to that catalog page.

You can always get back to the start of the catalog by clicking on '>> World >>' (or on 'Australia' in the Australian part of the catalog). If you are in the 'World' part of the catalog and you want to get to a particular country page (e.g. Vietnam), click on '>> World >>' then '>> Countries >>' and then select 'Vietnam@'.

4.3 The 'New Additions' Page

Users can see at any time what content has been added recently to World Law Index (and to World Law Search) by checking the 'New Additions' page from the World Law home page or from the [New] button on the button bars in the system.

Extract from the 'New Additions' page

4.4 Targeting the Web Spider from the Catalog

Editing entries in the catalog also involves the editor deciding whether to send the Gromit Web spider to index every word on the site which has been 'targeted'. The harness program (Wallace) reads the list of instructions from the Web indexing software, and then sends off multiple instances of the Web spider program, each to download the content of a particular Web site. The harness program ensures that only one instance of the Web spider software is ever downloading from a particular site, to avoid saturating that site with spider requests and denying access to other users. The harness ensures that the Web spider is 'well behaved', causing minimum impact on the sites from which it downloads Web pages. We call this a targeted Web spider, as it is not designed to traverse the Web generally: its downloading is limited to the site specified in the original URL which 'targets' it.

Targeting the Web spider to start indexing at the correct page, so that it when it indexes all other pages to which its starting page is directly or indirectly linked, but are equal to or below the start page in the server's file hierarchy, it indexes all and only the desired pages, is a complex task. Some desired sets of data cannot be indexed because of the 'noise' they will bring with them. For others, it is impossible to find an appropriately located 'table of contents' page to use as the 'start page'. Other 'problems of targeting' have also been identified .

4.5 Maintenance of the Catalog and the Web-Spider Targeting

Catalog links (URLs) go out of date as pages are moved on remote sites, or the sites cease to exist. If the URL is also the 'target' for the Web spider to start its work, the spider will be unable to do so next time it returns to the site to update its download of pages. To assist the indexers maintain the catalog we have a program called Comet which checks all URLs in the catalog every day and sends a report to the indexers indicating which links appear to be broken.

Another maintenance problem is that the web spider is sometimes sent to a URL which does exist, but for some reason cannot download the intended pages from that starting point. Conversely, sometimes an editor does not notice that by starting the web spider from a particular page it will download far more pages than are intended. To assist in addressing these problems, the program that controls the sending of Web spiders ('Wallace') sends a report to the indexers if the Web spider downloads only a couple of pages from a URL, or downloads more pages than a pre-set limit (currently 1,000).

5. Search Facilities of World Law/DIAL

The search facilities in World Law/DIAL search over two types of content: (i) the contents of the catalog; and (ii) the full text of all the remote pages indexed by the targeted Web spider.

There is now one interface to both the catalog and the search engine, with a search form located at the top of each catalog page (illustrated below). The very unusual feature of the search facilities in World Law / DIAL is that the scope of the search is (in default) limited by the location in the catalog from which the user carries out the search. In other words, if the search is from the 'World>> Legislation' page, it will search over all legislation from any country (but only legislation); if it is from the 'World >> Countries >> Germany' page it will search only over sites related to German law; and if it is from the 'World >> Countries >> Germany >> Legislation' page it will search only German legislation.

Search results are displayed with catalog pages ('categories') listed first, and then with the full texts of remote documents ('documents') listed. Both lists are sorted into likely order of relevance to the search query (relevance ranking).

5.1 Contents of the Search Facility

About 10 GB of text from targeted sites has been indexed to date, providing searches over about 1M pages of legal information. This is a search space about 50% larger than AustLII's Australian databases (approximately 6 GB). To date, the countries from which the largest components come are the .us domain of the USA (over 2 GB - mainly State legislation), non-AustLII Australian sites (over 1.5 GB), the .edu and .org domains and Canada (each about 1 GB), followed by another 3 GB or so drawn from 57 other countries. The emphasis is on legislation, law reform reports and law journals to date (because of Project DIAL), but major components of case law, law school sites and the like are being added.

5.2 Interface and Search Scope

The scope of a search is limited by where you search from (the context). As in the example below, if a user is at the 'World >> Law Reform' page, the default search scope will be 'Law Reform'. A search from this point will search over all Web sites listed on the Law Reform page, or those on any page which are sub-categories or cross-references from the Law Reform page.

To put it very simply, the way in which the search restriction mechanism works is that the search is first done over all pages retrieved by the Web spider, but before the results are displayed to the user they are 'filtered' by the URLs listed on the page indexed and its related pages, so that only relevant pages are displayed.

To broaden or narrow your search scope, you go to a more appropriate page in the catalog. Context determines search scope.

To search over everything available, the user could go back to the 'World' page (by clicking on 'World') and search from there. Alternatively, no matter where you are in the catalog, you can search everything ('World') by selecting 'All World Law' (from the 'in' option) instead of the default option limiting the search scope, as shown above. At present, the 'All World Law' option is by far the fastest way to search. But the more restricted search scopes may give greater precision. Test for what works best until the restricted scope searches become faster.

5.3 Search Options

World Law uses AustLII's SINO search engine, so all of the search facilities available for searching over AustLII's Australian databases can be used, but this is modified by new interface options. The following options are available:

The default option is 'any of these words'. If you want to do any other type of search, you must change this default.

• any of these words - equivalent to a Boolean OR between each separate word. This is the 'simple search' option, but often as effective as any other. Normally, users will simply enter a few words to indicate the main concepts for which they are looking (e.g. 'pollution river' or 'program patent').

• Users can also enter text in any form (e.g. 'I want laws on tax and bankruptcy' or 'tax taxation bankrupt bankruptcy') without knowing about search connectors, truncation etc. Results are ranked according to likely relevance, so the user will (usually) obtain a useful list of search results, although often not as complete as a Boolean search can provide. This is the search method recommended for inexperienced users, and is the default option. (This is AustLII's Freeform (Ranked Results) search method.)

• all of these words - equivalent to a Boolean AND between each separate word.
   

• this phrase - the words entered will be treated as a literal phrase even if they contain terms which would normally otherwise be treated as Boolean connectors (such as 'and', 'or' , 'near'). There is no need to put inverted commas around the phrase.
   

• this document title - only the titles of Web pages are searched, not the text.
   

• this Boolean query - any Boolean search may be entered, using AustLII's logical and proximity operators This is the most powerful form of searching, where searches can be made using Boolean and proximity connectors, and the results are then also ranked in likely order of relevance. This allows reasonably broad searches (to aid completeness) with the relevance ranking then providing more precision.

5.4 'Search this Site'

Where it has been possible to send the World Law / DIAL Web spider to a site, the icon appears next to the listing in the catalog. If you click on the words 'search site' or the icon, then you are taken to a 'Search Site' page which automatically limits the scope of the search to the one site selected. This function allows World Law Search to be used to search specific sites which have no search engine of their own, or have a search engine which does not have the same features as the SINO search engine used for DIAL Search. It also overcomes the need for users to learn to use the features of multiple search engines.

In the following, two of the sites may be searched, but the other two may not:

When a site is selected, a 'Search selected site' page is presented and searches from that page are limited to the site selected:

5.5 Display of Search Results

Results are displayed as shown below. Items are ranked in order of likely relevance. Catalog categories are listed first ('World Law Categories'), with the search over the text of the catalog pages, not just the title of the page.

The following example is the results of a search for the phrase 'financial intelligence' over the whole of World Law.

The percentages given against each item found are based on the most relevant item being given a score of 100% (provided it contains all search terms - otherwise less), and all subsequent items being given a percentage ranking proportional to that, according to their likely relevance.

The search form containing the executed search is displayed at the top of each search results page, so any search can be modified easily in light of the results obtained.

The present method of displaying results relies upon the remote site attaching informative titles to its HTML pages, as it is these titles that are displayed. While most sites do achieve this to a reasonable degree (as can be seen from the above example), some sites fail to provide any title at all, so that only the URL of the site can be displayed in default. This is not informative enough. A number of alternatives are under consideration. One is to display the first 50 words or so of the document, similarly to what is done in Alta Vista. A second is to include the name of the site from which each the document comes (as described in the catalog), or perhaps the name of the catalog page, after the display of the title. This would at least make it easier to recognise which countries and systems particular items are from. These choices have significant processing overheads, and we are concerned not to unduly slow the delivery of search results.

6. Storing Searches to Create a Self-Maintaining Index

One of our main tactics in creating a sustainable World Law catalog is the use of 'stored search links' (or 'stored searches') in the catalog.

For example, on the Intellectual Property Subject Index page (below) there are various stored searches for different aspects of intellectual property. The hypertext links that appear on the page are each a search of World Law that a legal indexer has created in order to find a general set of documents which relate to the topics of the searches. For example, the link entitled 'World Law search: trade marks and related laws' is in fact a stored Boolean search for 'trade mark or trademark or unfair competition or passing off'. These examples are relatively simple searches, but searches of any complexity can be stored.

Intellectual Property Subject Index page

The significance of these 'stored searches' of DIAL Search in the DIAL Index is twofold:

• Stored searches are by experts. The legal indexers responsible for developing and maintaining the catalog have expertise in search techniques, and know what types of searches are most effective over World Law. By creating stored searches at sensible locations in the index they make this expertise available to users of the index who are unlikely to have the same level of search expertise. For example, the search for 'patent law' above is actually a search for 'patent* or brevet* or octrooi*', utilizing both the French and Dutch terms, and truncation, to search effectively for the relevant terms in Spanish Portuguese, German and it.
   

• Another example is the above link for 'World Law Search: integrated circuits protection (semiconductor chip or circuit layout)', which is in fact the stored search listed below. Most users would not be aware that 'integrated circuit', 'semiconductor chip' and 'circuit layout' are all expressions used in legislation to refer to the same type of legal protection, and yet all three terms are used in the six most relevant documents found. An unassisted user would be unlikely to find the 620 documents relating to this topic. A user who only searches for 'integrated circuit' will miss over 200 documents.

• A stored search rarely needs to be updated . An expert who creates a stored search only has to do so once. When more data is added to the World Law by the Web spider, the expert does not have to change the stored search, because it will now find relevant new legislation and other materials as well as the old materials (assuming the search was well-constructed in the first place). The Intellectual Property page will to some extent update itself without editors adding new links to it. In contrast, sets of ordinary hypertext links to legislation must be updated constantly when legislation changes, or other new material is available. Stored searches can to some extent create a 'self-maintaining catalog', partly overcoming the impossibility of detailed subject indexing of world-wide legal information.
   

• Users can leverage more precise searches off general stored searches. The results display of a stored search allows the search to be modified by the user. For example, the following search form appears on the search results page for the integrated circuits stored search above. A user who wishes to search for the relationship between copyright law and integrated circuits law need only add 'copyright near ...' to the previous (stored) search, and press 'Refine Search', to carry out this more precise and expert search. In this way, World Law users can 'leverage off' the expertise of the legal indexers.

• Poor searches can find better searches. Inclusion of stored search links means that, since the catalog is searchable, a search can find other searches. For example, if a user does a World Law Search for 'breach of confidence', only one category is found, the Intellectual Property Subject Index page above, which contains the stored search 'DIAL Search for trade secrets or confidential information'. However, if that stored search is then selected, over 150 items are found and ranked, because the stored search was for 'trade secret or segredo commercial or breach of confidence or confidential information'. The user may not have been aware that in most commercial contexts, 'trade secret' is a conceptually the same as 'breach of confidence', but the invitation to repeat a search, and the stored search links, will assist the user to 'find' this information.

6.1 Finding Law about a Country

One particularly effective use of stored searches in World Law Search is to use them to enable users to find starting points for research concerning the laws of a named country. In the example below, the stored search is for 'fiji* or fidji* or fiyi* or fidschi* or figi*' , so as to find entries in most common European languages. The titles of the first few results show the effectiveness of the multi-lingual search. The total of 1366 items show how much is available even for a small country like Fiji.

Because the relevance ranking tends to give short documents and documents that use a search term in a title, many of the internet law indexes that have a separate page for that country will appear near the top of the list, so the user can they quickly review existing intellectual law indexes for that country. Here, the first 15 items found are a mixture of the 'Fiji' pages in other internet law indexes (JurWeb, ICL and ILRG), documents about human rights compliance, and Asian Development Bank law and development project notices.

6.2 Future Development of Stored Searches

The Intellectual Property stored searches are an example of how the current 'World >> Subject Index' pages will be developed so that each subject category has a basic set of stored searches that will keep that part of the subject index reasonably current. Resources will be available for more intensive intellectual indexing of some subject index pages, but this will not always be possible, so the use of stored searches will allow a moderate cost 'across the board' extension of the subject index.

In the longer term work will be done on the use of legal thesauri to create large scale sets of stored searches and their distribution through the catalog. Good thesauri are difficult to find on the web.

7. Multi-Lingual Law Indexing and Searching of the Web

The resources available on the Web for legal research are biased in favour of English, but there is a very large quantity of non-English language materials if only they can be located. Apart from the inherent value, the availability on the Web of the laws of a person's own country is more likely than most other factors to encourage that person to undertake legal research using other countries laws on the Web.

The key to the development of a genuinely world-wide free access catalog and search facility for law is definitely the formation of 'indexing partnerships' with legal institutions with expertise in other key languages for materials on the Web (Chinese, French, Spanish, German, Russian). Development of the technical capacity to handle different character sets is also required.

As the range of non-English materials searchable in World Laws increases, it is likely to become valuable to be able to limit searches to materials in a particular language. This will probably be implemented by the indexer indicating the language of non-English materials at the time of adding them to World Law Search, with an option for users to exclude or include materials in particular languages.

7.1 Automated Translation of Pages - Useful but Limited

All pages in World Law have a [Translate] button that takes the user to Alta Vista's automated translation service, provided by Systran translation software and ensures that the Systran page has inserted in it the correct URL for the DIAL page that the user was just viewing (in the example below, the World page). The user then only has to select to which language the DIAL page is to be translated, press the 'Translate' button, and then be returned to the DIAL page translated into the language of choice.

The resulting translation seems adequate to convey the meaning of most of the items on the page.

The World page and search options translated automatically to French by Systran

The Alta Vista/Systran translation facility is at present limited to translations from English to French, Spanish, Portuguese, Italian, or German, and vice-versa. This translation facility is also only a prototype, and sometimes has inadequate processing power to translate very long pages. It is also not recommended to use it to translate documents with complex grammar, or where accuracy is vital (such as legislation). However, for pages such as menus, or lists of search results, it is usually extremely helpful.

The translation facility has uses beyond translations of catalog pages:

• A World Law user can use the translation facility to translate their proposed search terms into another language before conducting a search, so as to obtain multi-language synonyms.
   

• Search results pages may be translated, so that if a search finds documents in another language, their titles can be translated into a language understood by the searcher.
   

• By initiating the translation facility before leaving the World Law pages (either on a catalog page or the search results page), and then selecting a remote document in a foreign language to browse, the Alta Vista translation facility will 'follow' the user as he or she browses from page to page, asking each time if this page is also to be translated between the two languages.

The combined result of these translation features for a world index is revolutionary: instead of being an 'English only' facility, it is now effectively available in six of the most pervasive European languages.

7.2 Embedding Multi-Lingual Searches

Where embedded searches are included on the catalog pages, if it is possible they are being constructed as multi-lingual searches. For example, on the Mozambique page, the embedded search for 'Search World Law: Mozambique' is in fact a search for 'mozambi* or mocambiq* or mosambi*'. Similar multi-lingual searches by country names have been placed as stored searches on all country pages (more than 200) in World Law.

At present the languages used cover most common European languages, with translations constructed using the Eurodicautom service.

Translations of country names in a number of Asian languages will be added next to the embedded searches. As these searches are permanent, broad searches of all the documents indexed from time to time, an investment of time in the construction of such searches is probably more effective than equivalent time spent indexing sites in detail.

8. Future Directions

8.1 Building a Perverse Portal

Internet 'portal sites' have been described as designed to perform two functions ( Hinton 1999 , Chapter 3): (i) providing users with the range of tools they need to find the content they want; and (ii) obtaining large audiences so as to generate revenue, typically through advertising (and the user surveillance it usually involves). World Law aims to be a portal for legal information, but it is a perverse one, because it is based around the idea of free access to Web resources and does not depend on advertising or individual subscriptions to sustain its development and maintenance.

If World Law succeeds it will be because the technical infrastructure that we are creating, and the collaborative environment within which it is used, is capable of attracting the interest and cooperation of others around the world with a commitment to the provision of free access to legal information via the Internet, and an expertise in some part of the Internet's rapidly expanding legal content.

8.2 Building Internet Legal Research in Asia - DIAL Training

One aim of World Law is to build an audience for legal research resources on the Internet which goes beyond the normal users in law firms and law schools of the developed world, and provides a facility which is also valued and used in the developing counties of the world as an affordable and genuinely international resource for legal development.

First DIAL training session for Mongolian teachers from the Legal Retraining Centre, Ulaan Baatar, held at the University of Melbourne on 9 July 1999.

The Asian Development Bank's funding of the training component of Project DIAL is a significant experiment in achieving this broader goal. DIAL involves in-country training of government lawyers (and others as resources permit) in seven Developing Member Countries (DMCs) of the Bank: Vietnam, Philippines, China, India, Pakistan, Indonesia and Mongolia. The DIAL training team will involve creation of a team from eight countries: local trainers in each of these DMCs, a Regional Training Coordinator from the Philippines[ 22 ], and the DIAL lead consultants at AustLII. 'Train the trainer' courses are planned to begin in Mongolia, the Philippines and Vietnam during 1999, and in the other countries in 2000. In addition there will be on-line training materials that anyone can access, and DIAL-user email list where anyone trained in DIAL use will be able to receive online support from the DIAL training team in any aspect of Internet legal research.

References

Greenleaf G (1998) Developing the Internet for Asian Law - Project DIAL (A feasibility study and prototype (Asian Development Bank, January 1998, 156 pgs) - at < http://www2.austlii.edu.au/~graham/DIAL_Report/ >.

Greenleaf G, King G, Mowbray A, Austin D and Matthews J (1997) 'Future-proofing a global internet index by a targeted Web spider and embedded searches' Australian Society of Indexers Annual Conference 'The Futureproof Indexer' 27-28 September 1997, Katoomba, Australia - at < http://www2.austlii.edu.au/~graham/Futureproof/indexers.html >.

Greenleaf G, Mowbray A and van Dijk P (1995) 'Representing and using legal knowledge in integrated decision support systems - DataLex WorkStations' Artificial Intelligence and Law , Kluwer, Vol 3, Nos 1-2, 1995, 97-124.

Hinton S (1999) Portal Sites: Emerging structures for Internet control Research Report No 6 La Trobe University Online Media Program, January 1999.

Lawrence S and Giles C L (1999) 'Accessibility of information on the Web' Nature , Vol 400, 8 July 1999, pgs 107-9 (Macmillan, UK).

Footnotes

1 . For more details see part ' Greenleaf 1998 .

2 . See < http://beta.austlii.edu.au/links/World/Countries/Mongolia/Legislation/index.html > for examples.

3 . 'Universal Resource Locator' or internet address of a Web page.

4 . For the detailed arguments see ' Greenleaf 1998 .

5 . This section summarises part 2.3. ' Intellectual indexes (directories) for law ' - in Greenleaf 1998.

6 . See < http://beta.austlii.edu.au/links/World/Other_Indexes_and_Search_Engines/> for many examples.

7 . This part draws on part 2.4. 'Automated indexes and Internet-wide search engines' - at < Greenleaf 1998 .

8 . See Lawrence and Giles 1999 , and ' How big are the search engines?' (Search Engine Watch) and references linked there from, for detailed discussion of all these matters.

9 . In 1996 it was claimed that Alta Vista only indexed about 10% of the pages of moderately large Web sites (600 to 6,000 pages in the example cited), and not denied by Alta Vista. Alta Vista now claims to index sites without any limit on pages. The claim by John Pike, webmaster of the American Federation of Scientists, and the reply by Alta Vista are available at < http://www5.zdnet.com/anchordesk/talkback/talkback_11638.html >and discussed in 'The Alta Vista Size Controversy' on Search Engine Watch.

10 . See Martin Koster ' The Web Robots Pages ' for details of the operation of Web robots.

11 . See the ' Robots Exclusion ' page, dealing with both the standard and the Meta Tag for robot exclusion.

12 . Lawrence and Giles 1999 . Earlier studies in 1996 had estimated that the best search engine covered about 20% of the then 150 M Web pages.

13 . For example, on Alta Vista, a search for Vietnamese legal materials requires a search which is limited to materials which are located on a server in Vietnam (the 'domain:vn' delimiter) or contain 'Vietnam or Viet Nam' - and this is still somewhat hit or miss.

14 . Examples are Google and Direct Hit.

15 . As at January 1998 - see Greenleaf 1998 .

16 . Aspects of such an approach, in a pre-Internet context, are explored in Greenleaf, Mowbray and van Dijk 1995 .

17 . Geoffrey King (to 1998), Daniel Austin, Philip Chung and Andrew Mowbray have all worked on the software and systems side of this project.

18 . Developed by Daniel Austin. Geoffrey King developed a previous version.

19 . The names 'Wendolyn' and 'Wensleydale' have been reserved for future software developed for this project. For further details see < http://www.wallaceandgromit.com/>.

20 . Developed by Daniel Austin.

21 . Developed by Andrew Mowbray.

22 . Rachel Romana of CD Asia and her colleagues.