Moving on From Google: Database Basics and Why You Should Still Learn Them

Everyone’s had those moments when a word or the name of a movie is just on the tip of our tongue in a conversation, but we just can’t seem to remember what it is. So, we go back and forth, explaining and describing it in thousands of different ways to our friends with hundreds of little random details we remember.

“It was about vampires, and there was this guy with long hair. I think Hugh Jackman played the main person” you say.

Our friends will either mention Van Helsing  or misremember The Greatest Showman. 

Moments like this sum up the process of conducting research: we have an idea of what we’re looking for and possibly have basic knowledge of the thing we want but just can’t put our finger on exactly what it is. So, we describe it to someone else and voila! They have the answer or at least understand parts of what we’re talking about. This is an interaction that’s been captured increasingly well over time by computers.

Natural Language Processing

Google, and other search engines, are a great tool because they are intuitive and accessible in how they work. Ignoring SEO, or Search Engine Optimization, a big reason for that is Google’s ability to process natural language and allow people to search things up using regular words and sentence structures without needing precise phrasing.

With adequate details and words similar enough to what we’re looking for, give or take a few typos, a modern internet search engine will translate what was put in and bring back resources that are a match. There isn’t a specific search logic someone has to learn or deep fundamental knowledge of a system that’s required to make it work. Even if a website or Reddit thread sits on the outer fringes of what you’re trying to find, it’ll pop up somewhere in the search results.

The ability to process natural language and allow people to use everyday speech to look for things online seems small, but it’s actually a huge development in search querying because it makes research more approachable and accessible. It’s also a feature that’s made its way into mainstream legal research databases so that effective research is achievable at almost any level of expertise. This makes the jump from doing searches on Google to doing searches within a database a lot less jarring, which is great because there usually comes a time when a researcher has to use one.

The Eventual Need for More Precision and Control

As a recap on my last blog about the pros and cons of Googling, my opinion as a librarian is that what’s found on the internet can be a good starting point to get familiar with a legal topic and the types of words that experts in that area generally use to talk about it. The main ways to find useful pieces of information, legal or otherwise, will entail using the right words. Browsing the internet allows you to pull key terms, phrases, and concepts that will help you find other resources.

There is, however, a need for high levels of scrutiny when consulting things like blogs or threads because of their potential unreliability and tendency to veer away from being informative. Internet resources can also lack the depth that a researcher needs to fully understand a legal concept. Both reasons are why almost every researcher should eventually turn to legal databases, which give access to in-depth and authoritative resources.

For the most part, if it’s offered, natural language searching is the default method I recommend starting with when using a legal database because of its ability to fetch so much information. So why is it even worthwhile to invest time in learning about things like Boolean operators or search query logic?

Like the limitations and problems introduced with internet search engines like Google, natural language searching as a computer process has its own drawbacks in terms of its precision and ability to filter out irrelevant sources.

It’s not so much an issue, but it is a fact that massive amounts of information can come up in general searches that use natural language processing, which can potentially bury what you’re looking for underneath thousands of irrelevant sources. Assuming that I put in the right key words, I want to demonstrate just how many resources a natural language search can pull.

Let’s say I’m interested in the effects of divorce on pet ownership. A search on LexisNexis, for instance, results in tens of thousands of results:

A common conundrum in modern research is finding way too many things to sift through. Again, it’s not bad, but it can be cumbersome. In the legal world, where specificity, ambiguity, and timeliness make or break cases, it’s critical to be able to filter out irrelevant search results fast and with precision.

Sometimes, you can just use different keywords so the system better understands what you’re trying to find, just like you might rephrase the same sentence to another person with different words or details. Other times, you need to change the actual type of search that’s being done and adjust its scope.

To do this on a computer, it’s good to spend time understanding the basics of database organization, facet filters, and search term connector logic.

Database Organization and Facet Filters

Let’s say you’re not ready to completely overhaul your search strategy to use key terms and Boolean connectors. Assuming you have a lot of results that need to be sifted through, a great way to begin reducing the amount you need to look through without changing your actual search query is by using facet filters.

Resources cataloged in a database are assigned additional descriptive characteristics and qualities, called facets, that act to organize the item within the database itself. In other words, items aren’t just aimlessly floating around inside the database. They’re intentionally organized in relation to each other based on shared qualities and distinguishing characteristics. 

For example, think of how movies and books are assigned genres based on their contents before being put on the shelves of a store. If you’re looking for a fantasy fiction book, you can ignore other shelves and go straight for the fantasy fiction aisle.  Even if a Lovecraftian horror novel also mentions things like fairies and mushroom people, it’s still perhaps not really what you’re looking for.  

Beginning to filter out resources based on some of their qualities will instantly cut down the amount of material you have to go through, saving precious time.

On LexisNexis, facets are found on the left side of the search results screen. Other databases may have this toolbar somewhere else, but they will all include a way to filter your results.

From here you can limit search results to only those that include the selected qualities. For example, taking the same search for pet ownership in divorce and filtering out the results for only published cases instantly removes around 4,000 cases from my results page. Further specifying that I only want cases that deal with the legal concept of community property that have been published in the last 10 years, I then narrow down my search results to 65 potential cases. Still a ton of cases, but a lot less than 10,000.

If you plug in my search to see the results yourself, though, you may notice a few things about how LexisNexis is processing my query:

1. It’s running “pet ownership,” “in,” and “divorce” as separate terms

2. It’s bringing up results that include either “pet ownership” or “divorce” but not always both

This brings up a fundamental issue in my search strategy, which is that the way I worded my query and the way LexisNexis understood it may not be relevant to me or my research because it is not how I wanted it to be interpreted. Even if I filter my heart out and apply every facet I can, the bulk of my results may still be irrelevant. So how do we fix it?

Term and Connector Searches

At some point, a researcher might have to change their initial query because a system isn’t properly understanding a search and fetching more clutter than it is useful information. This can either mean rephrasing a natural language search with new words or taking a new approach all together and applying terms and connectors search logic. 

Term and connector searches use simple Boolean logic and special modifiers to explain to a database how a set of words should relate to each other and how to read a string of keywords. Searching using this type of logic allows researchers to add important context that directly tells a search engine how to search. In other words, instead of relying on the machine to make guesses about what you want, you’re telling it exactly what you want. This generally must be toggled on, either by clicking on the advanced search option, or using Boolean operators right off the bat.

In most databases, you also have the option to perform a search within the results list using terms and connectors. In Lexis, it’ll look like this:

Taking my results for pet ownership in divorce, let’s refine the list of resources that came up with another search and show off some common Boolean connectors and word modifiers. We’ll still use the same initial search terms: pet, ownership, and divorce.

Boolean Operators

In my experience, there are four important and universal types of Boolean logic operators that alter how terms will connect with each other in a search. Each has a specific function and will affect what a researcher finds. The order in which these connectors are used in a search also matters, just like the order of operations in mathematics. Since the specific syntax can vary between databases, most of them have full support guides on how to use them:

1. Lexis Nexis Support Center – Search Connectors
2. Thomson Reuters Help Article – Search with Terms and Connectors

Despite the difference in which characters are used, Boolean operators serve the same function across every database engine. Below is a list of the four basic types:

OR

Indicates to an engine to find resources with any one of the search terms, but not all of them. Good for connecting synonyms and expanding search results.

Example: “Pet OR Ownership OR Divorce” means that results will contain any of the three terms, but not necessarily all three.

Proximity (/#, /s, /p)

Tells an engine to find resources with the search terms within a specific distance of each other. This distance can be specified by word count (/#), sentence (/s), or paragraph (/p). 

Proximity connectors are great to use because they can help ensure that keywords appear within context of each other in ways that matter, rather than being randomly used in a document. They also help to account for instances where filler words might separate key terms and phrases just enough to where quoted searching doesn’t yield results. 

Examples:

• “Pet /15 ownership” tells an engine to find documents with the word “pet” within 15 words of “ownership”
• “Ownership /s divorce” tells an engine to find documents with the word “ownership” in the same sentence as “divorce”
• “Pet /p divorce” tells an engine to find documents with the word “pet” in the same paragraph as “divorce”

AND

Tells an engine to find resources that strictly contain all the terms listed. This is used to narrow search results and ensure that each word or phrase appears at least once in a document.

Example: “Pet AND Ownership AND Divorce” means that results will contain only resources that have all three terms

NOT

Tells an engine that relevant resources contain a particular word instead of another connected term. Good for differentiating search terms that have nuanced meaning or have similar, but specific uses and meanings.

Example: “Pet NOT Ownership NOT Divorce” states that all results will contain results with the word “pet” and exclude those mentioning ownership and divorce.

Special Word Modifiers

Aside from Boolean operators, there are special characters and symbols that also directly affect how a search engine reads terms. Here are three of the most common modifiers:

Truncations (!)

Tells an engine to search for the root of the word in documents, expanding the potential search results that come up by accounting for a variety of word endings.

Example: “Pet AND Own!” will ask the engine to look for resources with the words pet and any word variation with the root own (own, owning, owner, owns, ownership, etc.)

Quotations

Allows you to search for the exact spelling and order of a phrase or word, with no allowed variation. This can potentially narrow your results and filter out relevant documents, so stick to using these for specific terms of art or legal phrases.

Example: “Pet ownership” asks an engine to search for resources that contain instances where the exact phrase “pet ownership” appears as a single term instead of allowing each word to appear separately from each other.

Wildcards (*)

This modifier allows a researcher to account for varied spellings of words, such as the American English version of “grey” versus the British English spelling of “gray.” It creates a universal character as a placeholder that can be used for any possible letter that has a chance of appearing.

Examples: “Colo*r” accounts for both “color” and “colour,” which are variations of the same word. Alternatively, “b*rd” can represent both “bird” and “bard.” 

With natural language searching becoming more developed and commonplace, one thing becomes clear: it’s not necessary to use connectors and operators anymore for most research. Sticking to natural language searching works well enough, and more often than not a person will find what they need. Especially as AI technologies that make use of natural language processing continue to improve, terms and connectors might become even less relevant in research practices. Still, understanding the basic functions and uses for Boolean operators, word modifiers, and facet filters is a time-tested way to save time and refine queries when natural language searches fall short.

By James Phaphone | Temecula Branch Librarian and edited in collaboration with Yanis Azzou and Matthew Palacios | Riverside Library Assistants