AI for literature research
Generative AI tools (like ChatGPT) that answer many questions in everyday life or assist with the planning of tasks are generally not helpful for literature research or writing (scientific) papers.
There are specialized tools for this purpose, an (incomplete) overview of which is provided here.
In the "Portal Digitale Lehre" there are additional recommendations for other specialized tools.
Only by using suitable tools can you easily achieve good results!
Overview of AI tools
There are different types of AI tools.
Assistants, finders and connectors are mainly important for literature research.
- Assistants can make it easier to get started with a topic; you can have search terms or research questions suggested to you.
- Finders are similar to classic databases: you enter specific search terms, phrases or questions and receive matching results. Unlike general AI tools, this approach primarily uses scientific databases as its source.
- Connectors work differently: Here you start with existing literature (e.g. by entering a DOI). The tool then shows you similar or cited works, ideal for delving deeper into your topic or finding research gaps.
All approaches can make your research more efficient - depending on where you are in the research process.
Some database providers now also offer advanced search options with AI, e.g. Scopus AI or Statista AI.
AI assistants such as ChatGPT help you to formulate your research question and generate relevant search terms - including synonyms, related terms and translations.
Possible applications
- Developing targeted search strategies for your research
- Set up a search matrix (generate synonyms, related terms and technical terms)
- Quick access to a topic
- Automatic summaries and structured results
- Starting with rough entries, you can approach the topic by asking questions ("Explain the topic in more detail" or "Add methods of the studies")
Limitations:
- AI summaries may be inaccurate or omit important details, please review the results critically.
- You should not search literature sources with AI assistants, they are often made up ("hallucinations"), even if they look correct.
- Assistants are not a substitute for academic databases, but a useful companion for the preparation and structure phase.
Finders help you to find relevant articles quickly. They analyze your search terms or questions and provide suitable results - including helpful functions that make it easier for you to assess whether a hit is really useful.
Frequent functions
- Summaries of the first search results: You get a quick overview of the content without having to read every article in full.
- Relevance filter by AI: The tools highlight important aspects so that you can decide what you want to look at more closely.
- Direct links to full texts: If available, you will be forwarded directly to the original source.
- User accounts: Many tools offer the option of saving searches or hits - practical for further work at a later date.
Partially available functions
Some tools can also analyze PDF files, for example to automatically create summaries or overviews of results.
But:
- This function can often only be used for certain PDFs.
- With many tools, it is only included in the paid version.
- You should always check yourself whether the content has been reproduced correctly and completely.
Do not upload any content that is protected by copyright or relevant to data protection - many tools store data on servers outside the EU!
Restrictions you should be aware of
- Language: Most content is in English.
- Subject areas: Focus is mostly on scientific and technical topics - humanities or social science literature is often underrepresented.
- Source types: Mainly journal articles; monographs and printed works are hardly covered.
- No access to printed literature: The tools only access digital content.
- No access to literature behind a paywall: Relevant and up-to-date references from databases, e-books, and electronic journals that require a fee or license will not be included in the analysis.
Examples (selection):
- Semantic Scholar: Semantic Scholar is a free AI-supported search tool with a focus on scientific papers. It is particularly suitable for initial research or for in-depth research.
- Elicit: Offers research results based on a question and presents the results in tabular form. Scientific literature is delivered as a result.
- SciSpace is a research platform that offers many other functions in addition to literature research. Natural language search entries lead to answers that also contain references. SciSpace also finds monographs, as it accesses data from OpenAlex, Semantic Scholar, Google Scholar and various repositories.
Connector tools find literature using a snowball system
These tools help you to find relevant specialist literature based on an existing publication. This is particularly useful if you already have a suitable article or paper on your topic and are looking for further sources based on this. The procedure is reminiscent of the snowball system: starting from a text, related or cited works are suggested.
It is also possible to start without a specific publication - for example via keywords or subject headings.
Frequent functions
- Find thematically similar publications - based on abstracts, titles and citations
- Citation analysis - shows you which works cite the source paper, for example
- Visualizations - e.g. networks that make connections between publications visible
- Direct link to the full text (if available)
Restrictions
- Few filter options: Targeted narrowing down of hits is often only possible to a limited extent
- References must be checked: Information is sometimes incomplete or incorrect and must be revised manually
- No automatic summaries: Although the abstract is usually displayed, no additional content summaries are generated
Examples (selection):
- Research Rabbit: Research Rabbit is a free application for organizing and managing research projects and literature sources.
- Litmaps: Creates a network of thematically similar publications based on a publication. In the free version with registration, only a limited number of topic networks can be created.
- Open Knowledge Maps: Open Knowledge Maps is a free application. It creates a clustering by subject area based on a search query.
Some database providers now also offer advanced search options with AI.
Here is a brief overview, which is regularly updated:
Scopus AI
Scopus AI is a search tool developed by Elsevier that is based on the multidisciplinary abstract and source database Scopus; search queries can be made in natural language, e.g. as a question. The results include a brief thematic summary, a visual representation of key terms as a concept map and suggestions for more in-depth questions.
Scopus AI draws exclusively on the metadata and abstracts of documents in Scopus published since 2003 and identifies the most relevant content from the more than 7,000 publishers in the database. The knowledge base is updated daily to ensure that answers are always based on the latest information.
Scopus AI is particularly suitable for gaining an initial overview of a research topic and finding literature for scientific papers. Further information can be found on help page of Elsevier.
Statista Research AI
Statista Research AI is an AI tool from the well-known online platform Statista, which provides data and statistics from business, society, politics and research. The tool combines Statista's extensive database with modern AI technology (Claude 3 Sonnet).
After entering a prompt (e.g. a question or a topic), an answer is generated based on up to 10 relevant sources that the AI considers appropriate for the query. Relevant statistics and infographics are directly integrated into the answers and the sources used are highlighted. The tool suggests further questions to delve deeper into the topic.
Ebsco databases
The databases of the provider "Ebsco" offer a search function for searches in natural language (Natural Language Search - NLS), in which queries can be made in everyday language instead of search terms. The search mode must be activated for the search (Advanced Search: Search Options / Search Modes / Natural Language Search, Simple Search: Set the slider to "Natural Language Search". Further information can be found on the Ebsco help pages: the AI search mode in natural language and Frequently asked questions about AI search mode.
- Literature search with AI - tips and tools
This presentation is dedicated to the following questions: What AI tools are there that are specifically designed for literature research? What functions do they have, for which research areas are they (not) suitable and which tools currently already deliver quite good results? What aspects (e.g. data protection, good scientific practice, copyright) need to be considered?
- Information page on scientific work with AI at the University of Tübingen
There are now a large number of AI tools designed for scientific literature research. Here you will find an overview of these tools as well as information on which subjects and search queries they are already or not yet suitable for. There is also an overview table of many other AI tools that can be useful for academic work.
- „Wissenschaftliches Arbeiten und Schreiben mit KI-Tools“ von Ina Kießling und Eva Schiegg
The University Library of Mainz has produced a comprehensive handout on the use of AI tools in the research and writing process, which provides an excellent introduction to the topic and is updated regularly. It is available in German language only.
- Overview of AI tools of the TDL of the University of Siegen (Team Digitale Lehre)
Digital tools with integrated AI functions are currently being developed in a wide variety of contexts and in a wide variety of designs. under the respective headings, the TDL has listed exemplary and currently common AI tools that can be used in studies and teaching. They have looked at some AI tools and created their own profiles with further information (costs, terms of use, etc.). Although many tools require registration, at least the basic version is free of charge. The overview is available in German language only.
Prompting
Prompting refers to the method of giving an AI specific instructions in order to obtain concrete answers or results. Good results when using AI depend heavily on how the query - i.e. the prompt - is formulated.
The clearer and more targeted the instruction, the more precise and helpful the answers will be. With these simple but effective tips, you can optimize your prompts and improve the quality of AI responses.
- Prompts should be specific, contextual and as targeted as possible. The more precise and detailed, the more effective.
- Assign a role to the AI so that it responds specifically from a certain perspective.
- Describe the scenario, the target group and the context of the application/task. (Background information, relevant details, purpose, ...). If GDPR and copyright allow, provide files to provide specific context or templates/examples.
- Specify the form and length of the output (bullet points, text: Essay (500 characters), table, email...).
- Specify the style of the output (scientific, suitable for children, ...).
- Ask questions to refine the request ("Is something still unclear?", "Do you need more information?").
-
If necessary, prompt in English to get better results.
Example prompt:
Generate keywords for 'classic' literature searches
"I am a [discipline] scientist and am writing a term paper on [topic]. I am looking for scientific literature on the topic. I need suitable search terms to search for literature in catalogs and databases."...
... "What (other) terms can I use in specialist databases for my research? Name at least [X] terms."
... "What are good synonyms, generic and subordinate terms, related terms for 'Y'"?
... "How can I truncate these terms for a database search?"
Search matrix
... "Please create an overview of these search terms for the above topic in tabular form and include synonyms, generic terms, subordinate terms and terms (English, French)."
- ChatGPT Prompts
Effective prompts are the key to fully utilizing the impressive power of ChatGPT. As an advanced AI model, ChatGPT processes input and generates responses that are customized to each prompt. Therefore, the quality and clarity of a prompt determines how relevant and helpful the AI's response is. If you understand how to formulate precise prompts, you can use ChatGPT in a targeted manner and achieve optimal results. - Open prompt catalog
The use of generative AI or large language models (LLMs) already plays an important role in the university system. With the Prompt Catalog, we would like to offer the opportunity for an exchange of experience: successfully used prompts can be documented in the Prompt Catalog and made available for reuse. In order to be able to understand and use the results yourself, a short description and the AI model used are requested in the form. - Prompting guide
Interaction with artificial intelligence (AI) has developed considerably in recent times. But one crucial question remains: How can AI models be controlled efficiently in order to provide more precise, helpful and contextually appropriate answers? This is where our prompting guide specifically for teachers comes in - a practical guide that introduces various techniques for formulating effective prompts. This guide was developed by a team of four teachers and trainers from the education sector.
