Search and Research
Thinking about Research
In the first part of this lesson, we will be looking at how to get started with research. Specifically, the content below will examine how to choose a research topic and tips on how to construct an effective search statement.
This short video takes you through some of the stages of selecting a research topic and illustrates that this process is usually iterative, meaning that you will have to revise and improve your topic after you start your research.
Whether you are using the Library’s search tools or the web, the right keywords are crucial to finding what you need. With any search you perform, what you put in impacts what you get back.
Tips for keyword generation:
- Consult your textbook or class readings. What types of professional, trade, or scholarly language are being used by the authors?
- Visualize it. Create a concept map from your ideas using a pen/pencil and paper.
- Read an encyclopedia article on the topic. Note specialized terminology, names of well-known experts, important dates, locations, and other potentially pertinent words.
- Use an AI search tool such as Bing CoPilot, Perplexity, or Claude to generate a list of terms, both broader and narrower. Use the prompt that gives the AI a role. Example: You are an expert in International Trade. What are broad and narrow search terms for . . . .
- Ongoing – your research should lead you to better and new terms, author names, and ideas to incorporate in your search.
Once you have the right words to start your search, you’ll be using one or more of these methods to find information, whether you know it or not:
- Keyword or Lexical – The search engine finds all the exact words you enter in your search.
- Natural Language & Semantic Search – You enter a natural language question, and the search engine breaks it into keywords, eliminates words like (a, the, of), and accounts for word order.
- Predictive Search – The search engine harnesses “narrow AI” to suggest searches based on common queries, location, your browsing history, and more.
- Field or Faceted Search – Retrieves information based on specific metadata and taxonomy systems created by people. Available on most e-commerce sites, library catalogs, and databases.
- Boolean Search / Search Operators – A mechanism behind most of the above searches. The” AND” operator is implicit in most web and library database searches.
- AND – Find only results containing all the keywords entered. Example: Paris AND Olympics
- OR – Find results containing any of the words entered. Example: Paris OR London
- NOT (in Google use the minus “-” sign) – Do not return results containing the word listed after the NOT or minus sign. Example: Paris -Olympics
Whether you choose to start your search using the library’s catalog, Google, or a Generative AI-assisted search engine, it is important to recognize that the systems behind search are complex and varied. They nearly always use more than one of the above information retrieval methods. Let’s explore more!
In the second part of this lesson, we examine how the technology of search works. This is an ever-changing field, but there are some basics that you should know. Knowing how search works can help you identify what type of search technology you are using (if unaware) and how to make the most of it.
Library Catalogs and Databases
Most library catalogs and databases perform a Keyword / Boolean Search, and like web search engines, they default to the AND operator. They also offer Field Searching. Field Searching allows you to search for particular types of things, like books or articles, and in particular parts of a record or document, like names in the author field or keywords in the title field. They use these tools for search because they are more appropriate for searching relatively finite collections of selected and vetted materials.
Example of a Boolean search using Search Fields:
Some library databases also incorporate search technologies that you would find in Web search engines or Generative AI enhanced search.
Web Search
Basic web search engines like Google, Bing, or DuckDuckGo use a combination of traditional search mechanisms and “narrow AI” tools – keyword search, Boolean operators, natural language processing, predictive analytics – to return results in a manner that attempts to put the most relevant and accurate results at the top of the list. Large Language Model (LLM) Generative AI is quickly changing the landscape, adding summaries to many search results pages. Let’s take a look at how basic web search works. In the next section, we will examine the impact of Generative AI.
How does traditional web search work?
LLM Generative AI and Search
What happens when you add LLM Generative AI and basic web search
If your search on Google or Bing is AI-assisted or if you are using a tool like Perplexity to search the web, you can expect to get a summary of your results with links to top-ranked sources. Here is what happens:
- If integrated directly with a search engine, the LLM doesn’t “read” the pages as humans do. Instead, it accesses their pre-indexed text data or raw content.
- Tokenization: The retrieved text is broken into small chunks, such as words or subwords. This allows the model to process and understand the structure and semantics of the content.
- The LLM Gen. AI identifies recurring themes, “high-signal” phrases, and contextually important details based on the user query.
- It reassembles the tokenized content into a summary that is coherent. It doesn’t quote or copy-paste from the results, but generates text based on tokens it identified. This can lead to responses that are original yet loosely paraphrased from multiple sources and can also make it challenging to directly link output to specific sources. It can also lead to so-called “hallucinations” if it reassembles the tokenized content coherently but inaccurately.
- Result Display: The generated summary is presented to the user, who can then check on the sources from which the summary was assembled.
Exercises
