AI tools for finding scientific sources

Table of contents

AI assistants that are integrated into various scientific databases or AI applications linked to citation datasets can help you to find scientific sources: peer-reviewed journal articles, books from scientific publishers, conference papers, and preprints.

Such applications understand natural language, so you can search for sources by asking a question, describing your research topic, or copying text you have written into the search. In some applications, the search is based on a key research article (“seed paper”) that you have found previously. Applications utilize, for example, semantic or vector search and natural language processing (NLP) to enhance search functions, form search queries on behalf of the user, determine the relevance of sources, and find semantically similar sources. In addition to searching for scientific sources, these assistants and applications often summarize and classify information found in sources using generative AI.

Typically, such assistants and applications are user-friendly and have intuitive interfaces. They work best in English and for searching English-language sources, but some also support searches made in Finnish.

Applications provide a link to the source if it is openly available or if the application is linked with JYKDOK. Otherwise the full texts of published journal articles can be searched separately from the “International Articles search” tab in JYKDOK.

Benefits of AI assistants:

  • Interdisciplinarity – AI searches can bridge gaps between different scientific fields by finding connections and integrating information from various disciplines, thus providing search results that are not immediately apparent.
  • Natural language searching – The quality of search results does not depend on the user's ability to define the correct keywords and master the search syntax of different databases before conducting a search.
  • Creativity – Search results may include sources that might be surprising at first glance but offer new insights for research.

Weaknesses of AI assistants:

  • Relevance of search results – Key search results relevant to the topic may be missing from the results, unlike if the search were conducted in a specialized scientific database or Google Scholar.
  • Coverage of search results – AI assistants and applications may have access to a very limited amount of citation data and bibliographic references. They may only search openly available sources, sources indexed by their partners, or they may lack access to the most recent publication references. The same applies to more traditional database searches: to obtain comprehensive search results, searches often need to be conducted in multiple databases.
  • Opacity of search results – The user has less control over the search results compared to a Boolean search conducted in a scientific database. It is not always clear why the assistant or application offers certain search results and on what basis it assesses the relevance of the source.

For these reasons, if you are aiming for a systematic, comprehensive (sensitive), precise, and replicable search for your literature review, for example, your primary search method should be database searches using keywords / Boolean search. However, as a complementary method of information retrieval, AI assistants and applications are worth trying.

Keenious

  • Licensed and recommended by the University of Jyväskylä!
  • Keenious is a tool utilising LLMs and vector search for finding scientific sources.
  • Keenious is best for exploring new topics, multidisciplinary information seeking, and supporting scientific writing.
  • If Keenious asks you to sign in, use either your university email address or JYU Microsoft two-step authentication.
  • Keenious works in the browser and as an add-in in Microsoft 365 Word.
  • Keenious does not save your inputs permanently, nor does it use them for AI training.
  • Keenious retrieves its bibliographic data from OpenAlex.
  • Comprehensive user guides can be found on Keenious's own website.
  • Read how Keenious finds and recommends sources.

How to use in a browser

When you open Keenious, you will see a text field on the front page, similar to those in AI chatbots (e.g., Copilot).

Type into the text field what kind of sources you want to find. The more context and details you provide, the better.

From the icon on the left-hand side of the text field, you can upload a "seed paper" to Keenious in PDF format. A seed paper is, for example, an article that discusses a topic for which you want to find more sources.

Search results

  • On the right-hand side of the page, you will get a list of publications related to, and recommended, based on your text.
  • You can change the language of the recommendations. For example: search for English-language publications using a Finnish-language seed paper.
  • Click on an interesting publication > You’ll see the article’s abstract, authors, and citation count.
  • You can check whether the article is available in JYKDOK databases by clicking Check JYU access.

Filters menu

  • Filter the results by, e.g.
    • Year published
    • Cited by count
    • Topics
  • Narrow down your search with keywords / search phrase
    • Use the search phrase suggested by Keenious
    • Or enter keywords or a search phrase constructed with Boolean operators into the search field.
    • The search field supports the operators AND, OR, and NOT; phrase search " "; and grouping of search phrases with parentheses ().

Chat

  • On the left-hand side, you will see an AI-generated summary of the currently selected text; an explanation of how the text relates to the seed paper; and possibly a list of related topics.
  • You can continue exploring the topic freely using the chat. For example, you can ask for sources related to a specific concept or context.
  • Source recommendations will always appear on the right-hand side of the page.

How to use in Microsoft 365 Word

Install the Keenious add-in in your JYU Microsoft 365 Word

  • Open a new Word document file.
  • Click Add-ins.
  • Search Add-ins > Keenious Research Explorer > Add

Open the text document you are working on.

  • Open Keenious in the sidebar.
  • The chat, search, and filtering functions are similar to those in the browser version.

AI assistants in databases

ProQuest Research Assistant

  • AI Research Assistant is integrated into some databases on the ProQuest platform, supporting you in searching for scientific sources and summarizing articles. With the assistant's support, you can also familiarize yourself with the main concepts of articles and brainstorm related research topics.
  • The assistant retrieves bibliographic data from the ProQuest database in which you are conducting your search.

How to use

  • Use JYKDOK to access the database ProQuest Central.
  • Perform a search using a keyword/Boolean search phrase.
  • Filter search results to include only those with full text.
  • At the top of the search results page, you will find AI-recommended keywords to expand or refine your search. Keyword suggestions are not necessarily generated for all searches.
  • Select a publication of interest from the search results and click to open the reference details. If the publication includes full text, the Research Assistant will open on the right side of the page.

Searching for sources

  • Use the magnifying glass icon to search for related sources.
  • Search using indexing terms
    • Perform a search using generated subject keywords.
  • Explore suggested sources
    • Find other similar publications.

Familiarizing yourself with the publication and its main concepts

  • Here is the key takeaway
    • A brief summary of the publication.

Ready-made prompts:

What are the findings or conclusions?

Brainstorm related research topics.

Describe the important concepts.

  • Research Assistant generates a summary of the publication and answers to ready-made prompts based on the full text. Always verify the generated answer with the full text to ensure accuracy.
  • Remember that the generated answer is not a scientific source. If you want to cite the publication, you must read the publication yourself.
  • With the help of Research Assistant, you can also familiarize yourself with, for example, Finnish-language publications. Note that answers to ready-made prompts are generated in English, and key concepts from the publication may be incorrectly translated into English.

EBSCO Natural Language Query

  • Licensed and recommended by the University of Jyväskylä!
  • Natural Language Query is an AI-assistant integrated into databases on the EBSCO platform, supporting you in your search for scientific sources by creating a search phrase for you.
  • The search then retrieves bibliographic data from the EBSCO database you have selected.

How to use

  • Access EBSCO platform databases - such as PsycINFO, Business Source Elite, and eBook Collection - through JYKDOK.
  • Basic search: switch on Natural language search.
  • Advanced search: open the tab Search options > Choose Natural language search.
  • This feature only works if you are using the interface in English! You can switch languages from MyEbsco > My preferences > Select your language.

Write a research question, research topic, or request in English on the search bar. For example: "What are the effects of austerity measures on consumers?"

  • EBSCO AI-assistant turns your natural language query into a Boolean search phrase. To review the search phrase, click Show refined query on the top of the search results page.

JSTOR AI Research Assistant & semantic search

  • Licensed by the University of Jyväskylä, but requires creating a personal account!
  • JSTOR is a multidisciplinary database that includes scholarly articles, books, and a wealth of (historical) primary source material.
  • The reference information for recommended publications comes from JSTOR’s own database.
  • Please note that JSTOR’s AI assistants cannot be used if:
    • The publication listed in the search results is not available in JSTOR.
    • The assistant cannot machine-read the text of the publication.

How to Use

  • Go to the JSTOR database via JYKDOK.
  • Create a personal JSTOR account and log in.

Semantic search

  • Search for articles and books using natural language.

Type your question or research topic into the search bar. For example: How are Japanese religious festivals linked to politics?

You’ll first get a list of search results based on keyword search > select Semantic Results.

  • You’ll receive 25 publications related to your question or research topic.

AI Research Tool

  • With the assistant, you can “read” articles and books: summarize them and ask questions about the content. You can also search for related publications and explore key concepts of the topic.
  • When you open an article, the AI Research Tool appears on the right-hand side of the page.

Ask any question related to the content of the text (Ask a Question). For example: How are nationalism and Shinto linked?

  • The tool will answer your question based on the publication and provide page numbers to help verify the accuracy of the answer.
  • Please note that sometimes the tool may fail to find page numbers. Also, to truly understand the context and meaning of the answers, you still need to read the article yourself.

You can also use ready-made prompts:

  • What is this about? > Summary of the content.
  • Show related content. > Lists other publications related to the topic.
  • Recommend topics. > Related themes and concepts.

Dimensions

  • Licensed and recommended by the University of Jyväskylä!
  • Dimensions is a multidisciplinary and large database containing references to scientific publications, policy documents, and research data, as well as information on received research funding, patents, and ongoing/completed clinical trials.
  • Access Dimensions through JYKDOK.
  • The AI assistants retrieve bibliographic data from the Dimensions database.
  • The bibliographic data in Dimensions is collected directly from publishers and databases such as Crossref and PubMed Central, as well as other open sources. The references are not specifically curated, so it is always advisable to assess the quality and reliability of the sources yourself.
  • Dimensions also contains some full texts, and scientific publications found in Dimensions can also be searched in JYKDOK using the SFX link.
  • Dimensions utilizes artificial intelligence extensively. For example, reference data is enriched with AI-generated classifications and information. Additionally, several AI features have been developed to facilitate the use of the database

How to use

Natural language query

  • Look for a star symbol on the top navigation bar and click it.
  • You will receive a Boolean search phrase and filters for your query, which you can then use to complete your search in Dimensions.

Type your request, for example: "Find academic papers published in the past 5 years in the field of International Relations focusing on the Gramscian hegemony theory."

  • Note: this is a beta version, and at the time of testing, does not generate as complex and extensive Boolean search phrases as many generative AI chatbots do.

Summarize

  • The "Summarize" button is available both within the search results list and when accessing an individual reference.
  • This feature generates a TL;DR summary, Key highlights of the paper, and lists Top keywords you can use to continue your search.

Chat with PDF

  • Chat with PDF differs from the other two AI features, as it is not directly integrated into the Dimensions database. Instead, it requires the use of a free version of an application called Papers, for which you need to create a user account.
  • The Chat with PDF button is available both within the search results list and when accessing an individual reference. You will be redirected to the Papers application in your browser, where you need to Register (Try Papers for FREE) or Sign in (Sign into an existing Papers account) if you already have an account.
  • (Step 1: Download from publisher > Go to publisher) Download the PDF file of the publication from the publisher's website to your computer > (Step 2: Upload downloaded PDF) Upload the PDF to the application.

You can ask the Papers AI assistant any questions regarding the article's content. Inquiries may include the research methodology, main findings, or definitions of concepts. For example: What does energy justice mean in the context of this paper?

  • You will receive a generated response to your question, along with links to the relevant sections of the publication used by the AI assistant to generate its response.

Semantic Scholar

  • Semantic Scholar is a multidisciplinary database indexing scientific publications.
  • Free.
  • Creating a user account is not mandatory.
  • The AI assistants retrieve bibliographic data from the Semantic Scholar database.

How to Use

Search with your research topic or keywords, for example: text data mining methods. 

  • You can use natural language and also do your search in Finnish if you are looking for Finnish sources.

Search results page:

  • List of relevant sources.
  • TLDR (Too Long; Didn’t Read): a brief summary of the publication’s main goals and results.
  • You can filter results by discipline and publication time.
  • Click open a reference that interests you.

Reference page:

  • TLDR summary
  • Figures and tables
  • Citations: publications citing the paper.
  • References: the paper’s bibliography.
  • Cite: save citation information in different citation styles.
  • Highly influential citations: publications significantly influenced by the selected paper.
  • Related Papers: similar publications.
  • Topics: related topics > Go to a separate Topic Page Beta for more information on the topic, frequently cited publications, and the latest articles on the topic (Definition, Related Topics, Papers often cited for this topic; Recent papers on this topic).
  • Note that not all references contain the above information!

Additionally, Semantic Scholar offers the following features for some papers:

  • Ask a question about > ask questions related to the paper, the language model provides answers based on the paper, justifies the answers with references, and indicates the page number.
  • Semantic Reader > Find the paper’s essential points - Goals, Methods, Results - faster with automatic highlighting.

When logged in, you can:

  • Save and sort search results into folders.
  • Find relevant sources using references saved in folders.
  • Create public folders and share collected references with colleagues.
  • Enable AI-generated publication recommendations (Research Feed).
  • Enable notifications (Paper Alerts / Author Alerts) for new citations or new papers by selected authors.

Visual tools

LitMaps

  • LitMaps is a so-called freemium application, offering a stripped-down free version in addition to a paid version.
  • Searches can be performed without logging in, but all features of the free version are only available by creating a user account. Please do not use the university email address.
  • LitMaps utilizes citation networks to find publications linked in various ways and presents search results interactively and visually.
  • LitMaps retrieves bibliographic data primarily from the following indexes and databases: Semantic Scholar, OpenAlex, and Crossref.

How to Use

Enter a seed paper into the search, for example: the author, DOI, Pubmed ID, or arXiv ID of a relevant publication you have previously found. You can also search for a seed paper using keywords.

  • Explore Related Articles.

Search results page:

  • Recommendations for relevant sources and a visual representation (literature map, Litmap) showing how the recommended sources are connected.
  • Adjust your search recommendations with the Explore function > More like this / Ignore for now / Never recommend this.
  • Adjust search recommendations using the search algorithm, selecting either Top Shared Citations & References (get publications connected to your literature map through citations); Common Authorship Patterns (get publications from the same/similar authors); Similar Abstract & Title Content (get publications with similar content to those in your literature map).
  • Export articles: save search results in RIS, BibTeX (.bib), or CSV format.

When logged in:

  • Monitor: enable email notifications when new relevant articles are published.
  • Design: create your own mind-map from the sources you find.

Research Rabbit

  • ResearchRabbit is a free application that requires creating a user account. Please do not use the university email address.
  • ResearchRabbit recommends publications to users based on bibliographies, citation chains, and text similarity, presenting search results interactively and visually.
  • The application retrieves bibliographic data from the Semantic Scholar and PubMed databases.

How to Use

Create a collection and add one or more seed papers to it by entering the article’s title, DOI, PMID, or keywords into the search. You can also upload references as a .RIS or BibTeX file.

Search Results Page:

  • You can proceed by choosing whether the application shows you recommended similar articles (“Similar Work”), previously published articles (“Earlier Work”), or articles published after the seed paper (“Later Work”).
  • You can also explore the seed paper’s bibliography (All References), publications citing the seed paper (All Citations), or other publications by the same authors (These Authors).
  • The application shows you networks between articles and authors, allowing you to navigate through the network to discover new, interesting articles.
  • ResearchRabbit works best with slightly older publications, as citation and other networks have had time to develop.
  • You can add relevant sources to your collection for safekeeping and continue refining your search based on them.
  • You can save search results in formats such as .RIS, which is compatible with Zotero.

How to cite:

Open Science Centre (2024), AI in information seeking. Jyväskylä: University of Jyväskylä. Online resource. (Retrieved: dd.mm.yyyy)