Wikidata Explorer

Wikidata Explorer is a simple interface to the world of structured data and knowledge graphs.

This application was made for a UBC CPSC 210 project so don't expect it to be well coded. That's also why many things may seem odd because they were requirements for the class project which I could not change. Feel free to use it however you wish!

Video Demonstration

Click the photo to go to the video. The video does not exlain much about what is actually going on so you may want to read the introduction first.

Introduction

Wikidata Explorer allows the user to see the relationships between information about the world. The application is rather simple, You can select a concept, property, etc. to start with (or use the default concept). From there, basic information is shown along with important properties. You can also expand the properties list to show all of them or search for a specific one. If a property interests you, then you can pull its card out to the side and explore additional information.

This application is of interest to anyone interested in structured data, computational intelligence, or just finds the relationships between things fascinating. This application provides an overview of Wikidata that allows a better understanding of the relationships which are the core of Wikidata. It's helpful for those interested in using the Wikidata API as it allows you to take a glimpse into the structure which makes it much easier to determine how to best utilize the data.

I have been fascinated with Wikidata for a long time. It is incredibly interesting how concepts, properties, lexemes, and other data types are related to each other. I keep a list of ideas for coding projects on my phone so that I always have one when I need it. Many of the ideas don't work in Java or don't fit the requirements of this project (I mostly make CLI tools). It just happened that this one (item number 10 on my list) does work which is why I decided to work on it this time.

Wikidata Basics

Wikidata stores a collection of structured data, that means that every piece of data is linked to other pieces of data. The full format is rather complex, so I will just cover the basics. Any "concept" has a concept ID starting with a Q. For example, Q2 is Earth, and Q42 is Douglas Adams (the author) every concept has a label which may optionally be in multiple languages. Q2 has the label "Earth" in English, "Terre" in french, "Terra" in Italian, etc. Each language also has a description and a list of alternative names such as "Pale Blue Dot" in English and "pianeta azzurro" in Italian. There is then a list of statements. Each statement consists of a property (which has an ID starting with a P), a list of values for that property, an optional list of qualifiers for each value, and a list of references (each reference itself having properties and values) for each value. Every property also has a label, description, etc. describing it and often an associated concept.

It gets a lot more complex than this. There are many other data types, properties and relationships which can be encoded with this system. I won't get into all of it but one of the nice things about this system is that it allows you to take human language statements and turn them into values that you can look up. For example, the English statement:

• "Mount Everest is the highest mountain in the world"

can be transformed into the statement:

• "Earth (_^Q2) (_^item) → highest point (_^P610) (_^property) → Mount Everest (_^Q513) (_^value)"

or something even simpler for computers to understand like

• "Q2 → P610 → Q513".

While Wikidata itself may not be used for all of these applications, it is this kind of data and statement system that powers common assistants like Google Assistant, and Siri. That's why I was interested in making this program, many people may wish to learn about Wikidata's structured data system to see if they can incorporate it into their own projects and systems so this program is a good way to get an overview to see how data is organized.

User stories

• As a user, I want to be able to query information about an entity.
• As a user, I want to view statements and save those which I am interested in.
• As a user, I want to add multiple downstream entities to my view.
• As a user, I want to search for an entity by name.
• As a user, I want to ask simple questions and validate statements.

Usage

CLI

• Clone the latest version of the repository. There are a ton of bugs and other issues which I fix as I figure them out, so it's best to be on the latest version.
• Open the IntelliJ project and build the project.
• Navigate in a terminal emulator to the root of the project (if your terminal emulator does not support ANSI escape sequences, you will need to add IntelliJ as an argument to the program to use a fixed screen size). A listing of the directory should look as follows:

Project-Starter.iml       checkstyle.xml       project_x3v2b.plantuml
Project-Starter.plantuml   data            project_x3v2b.png
README.md        lib             src
UIDesign.png         out

• Figure out which version of Java to use. This is build with Corretto 8 bt should work with any version of Java 8. (This is most important if you have many versions of Java installed). On my computer, this is /Library/Java/JavaVirtualMachines/amazon-corretto-8.jdk/Contents/Home/bin/java
• Determine the required classpath, -classpath out/production/Project-Starter/:lib/gson-2.8.6.jar seems to work well for me but may need more added depending on what is already in the classpath. The lib folder is the first place to look for a missing class.
• Determine the required arguments. There are a few support. The most important one is the web which specifies that you want to include information from the Wikidata website rather than just the local cache. Using web information is much slower but avoids many possible null pointer issues which I have not added proper error handling for yet. The other important one which I mentioned before is IntelliJ which must be added if you are using a terminal without ANSI escape sequences. This includes the built-in terminal to IntelliJ but also the default Windows cmd.exe.
• Run the program! Your command will be built using the previous steps. Mine looks like this: /Library/Java/JavaVirtualMachines/amazon-corretto-8.jdk/Contents/Home/bin/java -classpath out/production/Project-Starter/:lib/gson-2.8.6.jar ui.CLInterface web
• Unless you had the IntelliJ option enabled, you will see some weird characters printed out. I see [[70;114R for example. That shows that my terminal is 70 columns wide by 114 lines tall. This is because of the secret way that the program determines the size of the screen. You never see this, but it prints out \u001b[s\u001b[5000;5000H\u001b[6n\u001b[u . This does the following:

"\u001b[s"             save cursor position
"\u001b[5000;5000H"    move to col 5000 row 5000
"\u001b[6n"            request cursor position
"\u001b[u"             restore cursor position

• This moves the cursor far away which the console can't do, so when the position is read, it finds the actual bounds of the console. The cursor is then put back as to not mess up the display. This is the best way that I have found to do this to work on most platforms. There are other methods to find the console size but none work well on multiple platforms . Usually, the program would just see the cursor position [[70;114R as input to standard in. Java is weird though and prints it out like you had typed it as user input. So you then just have to press enter to input that to the program. It's weird, but you must press enter and do not type anything else.
• Now you should be able to get to using the program! I recommend typing load to load the cache, so you don't have to wait to gather all the web data. You can type exit at any time to exit the program. The starting entry is for Douglas Adams. It shows ten of the statements about him. To view a statement, type his id (Q42), then the statement you are interested in. So to look at his mother, type Q42 P25. If you are not interested in an entry, add an R after. So if you don't care about his spouse, type Q42 P26 R to remove it from the list. If you know the property number for something, you can add it to the list using the search bar. To add notable works, type Q42 S P800.
• To learn more about a property, type its ID when it is open, so if you had P25 opened, type P25 to have the page about that specific property open . If you are interested in the subject of a statement, type the property then the subject ID. So if you had P25 open, you can type P25 Q14623678 . This will show you the specific statement (since one property can have multiple statements associated, see P800). This page will then show you the references and qualifiers for that statement. These cannot be opened further. Now that this page is open, you can see the general page by typing in the ID. So to see the general page about Janet Adams, type Q14623678. Note that anything you type which opens something can be typed again to close it. But be careful not to close things in the wrong order, or it may become impossible to close something further down the chain.
• If you are not interested in Douglas Adams, you can use the general search overall to look for something else. For example, type S Q2 to learn about Earth, or S Q76 to learn about Barack Obama. Or even S Q14527788 to learn about the birth certificate of Barack Obama.
• You can also type save to save anything which needed to be loaded from the web to the local cache file. This way, it can be used without being loaded from the web again next type you open the program and type load.
• The question is, how do I find these IDs? I've memorized a few to use as examples, but I don't have a way to search by words (yet . . .). All you can do is go to Wikidata itself and search for what you want and look at the URL or title to find the ID. You can also use the Wikidata pages to find properties that might be interested. The point of this program is not to help you find the information, but to help you see the connections between data.

GUI

• To use the GUI, you can easily just run GUInterface from IntelliJ.
• Drag a panel to move it around.
• Drag outside of a panel to move everything.
• Click save or load to save or load the local cache.
• Type something in the search bar at the top and press enter to load up that ID.
• Don't ask why the colours don't quite work (It is not easy to get buttons and JMenuBars to change their colour).
• Press X on a panel to close it.
• Press the right arrow on a panel to open another panel with that item's ID . (Similar to just typing an ID on its own in the CLI version).
• Use the search bar on an item to search for a property by ID.
• Click the arrows to view more information about that property or statement.
• Be sure to click on one of the arrows next to the image to see a photo of that item. I recommend searching for Q76 (Barack Obama) and clicking on his arrow for the image. There are quite a few other good ones too, I also like the image for Earth (Q2). Currently, this only shows one image, but I am considering finding a way to view all images as some items have multiple.
• It's fun just to play around and look for connections! I recommend that if you run out of space, start going upwards, there are some weird issues with jumping around when you go down then add a new panel. I have spent hours debugging it but I have absolutely no idea. It may be an issue with the DragLayout which I had to find on StackOverflow. Swing is annoying to use, so I hope it is not too big a deal that I had to borrow this one bit from the internet. I added a comment with a link to the original location. I could have maybe figured that part out on my own, but it would have taken many more hours than I was willing to put in.
• That's about all there is to it!

Phase 4

Task 2

• Test and design a robust class: I have an Exception called NotFoundException. This is mainly thrown in the WebCollector and LocalCollector classes if data for a specific ID could not be found. This exception is then caught in several different classes. In some cases , there is nothing that can be done, so an unchecked exception is thrown because there is no easy way for the program to continue. Having this exception does allow for may of the NullPointerExceptions which happened in the past to no longer happen however, so it's overall quite good.
• Include a type hierarchy: I have a very complex hierarchy in multiple locations. The main one is the Value hierarchy. A Value is any datatype on Wikidata. A Datum extends it and is some Value which has a Wikidata URI . There are further subclasses of Datum for the different types of pages . There is also the additional package that contains many datatypes that do not have their own URI. There are even more subclasses for specific uses. It's quite a complex hierarchy with many classes. Maybe if I document my code someday, it will be possible to fully understand what is going on. At this point, even I couldn't fully explain what is going on.
• Make appropriate use of the Map interface: Besides the specific JSON cases where I need a Map, I use one for my cache of data collected from the web in this session in my WebCollector class. I also use it for the lists of statements, qualifiers, and references, so I can look them up by ID.
• Make appropriate use of a bi-directional association somewhere in your code : There are several places where I have bi-directional associations. I can't possibly list all of them, but if you want to see one of them, you can look at the LayoutManager and ItemViewController classes.

Task 3

At this point, there is not much I can do to improve the structure and design without spending a huge number of hours. There are a lot of classes that both handle data and display of that data. There are also a lot of places where things not existing cause problems which I cannot easily fix . Overall, there just needs to be a much better design to the whole thing which more clearly separates what each thing does. Statements and DatumLink and several other classes shouldn't be items but are because when I wrote them, making them not items would require a lot of changes elsewhere. There could be a much better use of the composite pattern than the not-so-great usage that I have now which only somewhat works. Have a look at model.uml and Dependencies.uml (for dependencies.uml, select all and click show dependencies to see the craziness). Then you will understand how hard changing anything is. My entire project has extremely high coupling which is quite bad if I ever want to expand it . That being said, I will probably never do anything else with this project. If I ever wanted to do it again, I would write it as a web app. Several nice JavaScript libraries let you use Python rather than JS for client-side implementations. I've also updated the project_x3v2b UML diagram which is created by SketchIT and plantUML. It seems to not work properly but is another nice visualization.