BesluitBron 1.6.1-23 Status Sign in
Disruption Open Raadsinformatie has not been synchronised for some time. See the status page

Search precision

Last changed on .

An answer that cannot simply be taken at face value says so in the answer itself. That is the principle behind search precision. This principle applies to every tool in every connector, both now and in the future.

The reason is practical. The source platforms differ greatly in what they can deliver. Open Raadsinformatie always gives an exact number of results. OpenBesluitvorming only does so for narrow queries; otherwise, it provides an estimate. Tweede Kamer only searches titles and therefore misses up to 99 per cent of what is in the text. An answer that fails to mention these differences suggests a level of completeness that does not exist.

The four fields

A tool that returns a list of results has four fields. Listed below are four tools that do not yet do this. These are the fields:

  • total_is_exact: whether the figure quoted is the actual number or an estimate provided by the platform;
  • source_distribution: how the hits are distributed across the sources, together with the range (scope) over which that distribution was calculated. A distribution across the first hundred hits is, in fact, slightly different from a distribution across the entire set;
  • match_type per result: how closely this result matches the search query. exact_phrase means that the words appeared next to each other in that order, proximity that they were close together but not necessarily next to each other, loose_terms that they appeared somewhere in the text and may be several pages apart, and mixed_phrase_and_terms that the search query consisted partly of a phrase in quotation marks and partly of individual words. The latter value has been available since version 1.3.50. Prior to that, such a mixed search query was referred to as exact_phrase, which described the entire query in a way that was only half correct. From version 1.3.58 onwards, proximity is also returned by Rechtspraak for a search query such as "stikstof vergunning"~10: this requires two words within ten positions of each other. Previously, such a query would return exact_phrase or mixed_phrase_and_terms, which sounded more restrictive than what was actually being searched for;
  • more_to_find: whether there are any further results outside the displayed window.

These fields are not entered manually for each tool. A single shared facility constructs them, ensuring they have the same meaning everywhere. The construction fails if a tool’s description does not mention the fields. This applies to any tool whose name begins with search, and, from version 1.4.8 onwards, also to any name beginning with find_.

Four tools return a list without all four fields. For Open Raadsinformatie, these are find_document_pages, find_people and find_organisations. These return a total and a list, but none of the four fields. For Rechtspraak, find_rulings does return total_is_exact, source_distribution and more_to_find, but not match_type. This tool does not search for words. There is therefore no search query that matches a result.

Anyone who cites a search result in a decision, a statement of views or a defence is making a ruling as to its completeness. ‘There is no council decision on this subject’ is a different assertion from ‘this subject does not appear in the titles searched among these 331 sources’. The first assertion cannot be substantiated using the resources available here. The second can, provided it is stated where the search was carried out and what criteria were used.

The fields above provide precisely the data required for that second formulation. An answer based on BesluitBron should therefore specify the source, the scope and the nature of the match, and not just the number.

What the fields do not do

The fields provide information about the search query but say nothing about the accuracy of the document found. An incorrect council proposal remains incorrect, and a withdrawn motion remains in the register. Nor do the fields provide any information about what a platform has never recorded. For this, the scope of coverage for each platform is described at Bronnen.