New AI tool in Virtual Library
Research Assistant. The University’s Virtual Library (accessed via University email log in) has implemented the “Research Assistant”, a tool based on generative artificial intelligence (AI) that uses large-scale language models (LLM). It allows you explore academic content by asking questions in natural language. The tool uses most of the content found in your library to identify five documents that can help answer your question. It then extracts the most relevant information from the description/abstracts of each source to write the answer. Below the answer, you’ll see the sources used to generate it along with in-line citations that let you clearly see which source was used to generate each part in the answer. Use these sources to delve deeper into the topic and to fact check the responses from the tool.
The Research Assistant is not a replacement for human expertise but uses artificial intelligence to automate otherwise time-consuming tasks. We’ve designed the Research Assistant make it easier to understand topics, their context, and resources published about it. Use the “view more results in your library search” button to find more documents relevant to your question. Click the AI-generated “related research questions” to explore topics similar to your question.
How are responses generated? Your question is converted into a query that the search engine understands with the help of a Large Language Model (currently GPT 3.5). The search engine then identifies the most relevant documents in the index. It ranks them according to how well they can answer the question and, again a with the help of the Large Language Model, creates an answer from the top 5 sources. Due to the nature of Large Language Models, answers to the same question are not always the same. There may be more than one possible answer and different resources that are relevant. If you are not satisfied with your answers, use the “Try again” button.
How to formulate a good question? To make the most of the Research Assistant, it’s essential to ask clear and detailed questions about academic or scientific topics. Be as specific as possible and phrase your query in the form of a question. Example queries can be found on the starting screen.
Supported questions/instructions. The Research Assistant supports local language searches. Most material in our index is in English. If you ask a question in another language than English, the Assistant will search in both, your local language and English, and write the answer in the language of your question. Note that there is a dependency on the Large Language Model and language support may vary. We are currently using GPT 3.5 for language processing and translation. You can add language-specific instructions to your question just as you can in ChatGPT. For example, you can add “answer in German” or “answer in French” to the end of your research question.
Unsupported questions/instructions. Some instructions are not currently supported by the Research Assistant, like requests for materials of a particular type (e.g. “give me peer reviewed articles about bird migration”) or from a certain time period (e.g. “give me the newest research on climate change”). You will still receive an answer when including these instructions and the ranking algorithm will take keywords like “peer-reviewed” into account, but the tool will not filter content by type or date. The Research Assistant does not yet support follow-up questions. Each question stands by itself. For example, if you ask “what topics did Simone de Beauvoir write about”, you cannot follow up by asking “and what is the content of that work” and expect the system to understand what you mean. At this time, you will have to include all relevant information in each question, e.g. “what is the content of Beauvoir’s The Second Sex”?