Generative AI Features in Major Databases & Platforms

When you are viewing a full-text article or dissertation in one of these databases, the Research Assistant displays in a sidebar on the right side of the screen, where there are prompts to display Key Takeaways, Important Concepts, Essential Details, Findings & Conclusions, Visualize this Document's Topics, and Related Research Topics:

Please note that all of these options may not be available for some content types, such as newspaper articles. Some options, such as Essential Details and Findings & Conclusions, cite the portions of the document used by the Research Assistant to generate these fields. It is important to check these citations to make sure the Research Assistant is interpreting the document accurately and in an appropriate context. If you run a search in a ProQuest database and then access a full-text document in your results, its Key Takeaways will be influenced by your search to make them more relevant to your query, and a "Relationship to your search terms" section will be included in the Key Takeaways. 

 

ProQuest has also introduced an "Ask a question" free text box that allows you to ask questions about the document. In the example below, the user has entered their own question, "What are some conservation challenges facing all 3 penguin species discussed in this article?" and has received an answer:

 

These answers can help you get a quick overview of a document, but they are not a substitute for reading the document itself. You should not cite a document or use it in your research based solely on the answers given by the Research Assistant; instead, you should read the document yourself to make sure you accurately understand it. 

 

The Research Assistant's answers, both to the standard prompts and your own questions, are generated from the full text of the document being viewed using Retrieval-Augmented Generation (RAG) techniques; no data from other documents or from public LLMs is used to generate them. This helps to minimize hallucinations, but it also means that the tool cannot answer questions on topics that are not addressed by that document, nor can it provide interpretation or critical analysis beyond what is presented in the document. The tool also has safeguards in place to try to prevent the generation of offensive or inappropriate content, so if a document contains some sensitive terms, it is possible that the Research Assistant will not return responses for that document. ProQuest is not using user inputs nor the full-text documents to train LLMs, and ProQuest is only using generative AI within a secure environment.

 

For more information, please see:

On the ProQuest Ebook Central platform, ProQuest has added a Beta feature for many ebooks, the Ebook Central Research Assistant. Within the online ebook reader, the Ebook Central Research Assistant displays traditional metadata that has always been available on the platform (publisher-provided description and subject terms) plus new AI-generated data points: chapter-level key takeaways and concepts. These are generated solely from the full text of each book chapter in order to minimize hallucinations and improve accuracy. This is a tool to aid in the research process, but it it not intended to replace thoughtful contextual reading. 

 

The Research Assistant is automatically enabled on eligible books when you open the full text in your web browser on the Ebook Central platform. You will see it on the left side of the ebook reader for PDFs or on the right side of the screen for EPUBs. It is represented by a circular purple icon with 3 stars in the middle:

When you are reading a chapter and the Research Assistant is open, it offers the options to generate key takeaways and concepts for the book chapter. After concepts are generated, a magnifying glass (search icon) appears next to each concept, and you may select it to run a keyword search in Ebook Central for that concept:

 

• Some publishers, authors, and titles have opted out of participation. For these books, the Research Assistant is not available.
• Even if the Research Assistant is available for a book:
  • The LLM model has built-in guardrails  to restrict generation of offensive or inappropriate content, so for chapters that contain sensitive terms, the Research Assistant may not generate key takeaways and concepts.
  • A minimum chapter length is required by the Research Assistant, so for very short chapters, it may not be able to generate key takeaways and concepts.