data_science

This post is part of a series of research for possible improvements in the follow-up of legislative proposals. Although our goal is to submit the project to the LABHacker (Civic Innovation Laboratory of the Chamber of Deputies), is currently an independent initiative. The information was collected from the website Open Data and the Chamber Portal and will be used here in accordance with its terms of use.

Following the demonstration of new ways of displaying the Disarmament Statute, we have prepared a second part of the essay. Here our initial goal is to explore the simultaneous search in multiple documents , what we'll call global search.

For this purpose we will use an application called Dynalist , but we could use others, for example: Workflowy , Checkvist or Moo.do . This option stems from the fact that we do not yet have this functionality incorporated into our platform, while it is already a standard behavior of free-to-use outliners.

Our differential is that we are able to generate, with just one click, the files with the properly structured standards. Thus, the files were generated in our environment and are only being displayed in a commercial use program. If our community confirms interest in this functionality, it will be developed for our platform, but today it is not yet available .

Well, with the caveat, let's go to the demonstration.

While searching a collection usually returns a list of documents, the behavior demonstrated in this post consists of bringing the relevant documents already open and filtered. Such a filter highlights the display of fragments in which the search returns a positive result . In addition, to give a little context, the hyearquically superior fragments are also displayed, which work as a kind of "crumbs" to know the path to the destination fragment.

In the first part of the demonstration, this behavior and its advantages have already been commented on. But only now have we demonstrated how it is possible to do this with several documents simultaneously. It's easier to see it at work than to explain:

In the left sidebar is the relevant collection. In this case, a PL, a law and the Penal Code were selected. When searching for the word "shot", the system reveals that the word is both present in the law and in the PL. The difference is that, in the current law, there is a prohibition of shooting by hunters for subsistence who give another use to their weapon . In the PL, the word occurs to register that The collector cannot fire shots .

In turn, it is possible to note that, both in the law and in the PL, the firing of a firearm continues to be provided for as a crime. In our view, this is the ideal way to perform quick queries in multiple documents, because The normative hierarchy itself already helps a lot in the comparison . After all, right from the start, you can see if the comparison is in the context of "crimes and penalties" or "possession" or "acquisition and registration".

This immediate situational awareness saves the interpreter a lot of mental burden.

A few more examples help illustrate the functionality. Let's see how the system behaves with the search for "omission" and then for "firearm":

Other information systems would require many more interactions to reach the same degree of evidence. Just the fact that there is no pagination (comings and goings) already helps a lot in the manipulation of the content.  In this sense, the experience is similar to what is expected from Google: the immediate solution of the issue. On the other hand, we are not averse to trying a new search argument, as this is the practice we already have in other tools. We want the answer to always be on the first page . At least that's how I personally do my searches.

This way we prove that it is possible to search and read several documents on the same screen, without changing pages.

The alternative so far, for an investigation like this, would be to set up an Excel spreadsheet, which has no search. People were used to copying and pasting text into programs designed to do mathematical operations , which is not a rational use. We have seen, more than once, huge normative texts pasted on a spreadsheet, so that the reader would have some notion about the parallel of the fragments in various norms.

And that brings us to the second goal , consisting not only of being guided by the normative structure embedded in the law, but also of being able to Create relationships between fragments of different documents . In other words, how do you create relationships without getting stuck in the metaphor of the Excel line?

Our proposal is to make it possible for one document to be consulted within the other, opening a window for creating the link manually. In this way, navigation becomes much more fluid:

We need to recognize, however, that the experience is still somewhat truncated, because it really is something very complex: navigating within a structured text, creating relationships with points from another structured text. It's not an easy task, especially in the case of creating "inline" references. Perhaps the server responsible for creating these bonds should be an expert.

For the reader, however, the experience seems pleasant, because what he notices is only a link within the text, which refers him to a certain point in another text, with which he has a relationship. For example, Article 46 of the PL is related to Arts. 12 and 14 of the Old Law:

In other words, creating text structured in this way has always been difficult. Now it has become possible, despite the difficulty. The advantage is that, once created, the reader will find it very easy to see relationships between points of normative documents that were previously hidden.

This can be useful, for example, in organizing and voting on amendments , because it is undoubtedly a moment in which two texts need to be confronted clearly. On the other hand, it can be useful for those who research and compare normative texts or, who knows, for those who draft a text proposal.

In any case, for the purpose of proof of concept, it seems to be demonstrated that it is possible to consult several texts (or several versions of the same text) with a very intuitive and fast search. Currently, as far as we know, there is no tool that facilitates the comparison of texts of this nature.