Generative AI is becoming an essential tool in research and learning today, and is being used for a variety of tasks that can greatly improve productivity. In early September, the Center for Teaching and Learning innovation published a series of Learning Tips for graduate students on generative AI tools useful for writing papers. We introduced Scispace, Connected Paper, and Consensus as generative AI tools useful for writing papers and discussed how graduate students can use these tools in their research. In this Tip, we'll highlight the key features of Consensus, Scispace, Perplexity, and DeepL, all of which are generative AI tools that can help professors increase productivity in graduate teaching and research. All four tools make it relatively easy to test out their functionality.

1. Consensus (https://consensus.app/search/), AI for Papers Search and Analysis

Consensus is an AI-based academic search engine that facilitates quick literature reviews. Like Google Scholar, it allows users to search for papers but integrates AI capabilities to summarize key points and analyze research findings easily. A notable feature is the Consensus Meter, which aggregates opinions across multiple papers to identify research gaps efficiently.

1) Search for papers and synthesize with Copilot

One of Consensus' greatest strengths is its ability to quickly review and synthesize research papers. Users can easily find the papers they're looking for by utilizing filtering options by year, method, journal, and research area. They can also refine their search by study design, methodology, sample size, open access status, and more. In response to your commands (chat input), Consensus analyzes a corpus of full-text research papers based on a language model. After natural language processing the corpus, it combines keyword search and vector search to present the 20 most relevant research results.

Consensus' Copilot feature is a generative AI tool that aids research and writing by providing key insights from search results. Users can automatically generate research questions, methodologies, key insights, and more with a single click, and can easily explain them to a wider audience by adjusting the style and audience. The feature supports a variety of formatting options, including drafting literature reviews, creating bulleted lists, grouping pros and cons, and more.

2) Consensus Meter

One notable feature of Consensus is the Consensus Meter. This feature helps you get a better understanding of the research at a glance by determining whether there is consensus among multiple studies on a particular topic. When a new paper is published, it is based on previous research, and there may be opposing viewpoints as well as supporting opinions regarding that paper. The Consensus Meter categorizes opinions about that paper as "Yes," "Possibly," or "No," and shows an aggregated result by reviewing the first 10 results and only adopting the answers that are deemed sufficiently relevant to the question.  This feature is useful for identifying agreement and disagreement with research questions to reveal research gaps, and allows you to analyze research results based on key information such as sample size per study or research methodology. Because research gaps are easily identifiable, researchers can quickly identify areas that need further exploration.

2. SciSapce (https://typeset.io/), an AI that specializes in summarizing and interpreting papers

SciSpace is a generative AI-based service specialized in searching approximately 282 million research papers and assisting users in understanding and analyzing them through interactive Q&A. Unlike other generative AI tools, SciSpace is specifically tailored for research papers, offering more professional and accurate information for summarizing and interpreting academic content.

1) Search for paper

Using SciSpace begins with entering a research question. Users start by inputting the problem they wish to address—such as “the severity of climate change.” In response, SciSpace analyzes and provides answers based on the top five relevant papers, enabling a comprehensive understanding of the research question and offering citations for each statement. This allows users to quickly locate and explore any topics they want to investigate further. Additionally, SciSpace supports multiple languages, including Korean, for searches.

2) Summarize and interpret the paper 

Below the answer to the question you entered, you will see a list of related papers, as well as papers referenced by the AI, some of which you can view in full text and download as pdf files directly from SciSpace. You can also filter this list by checking [PDF], [Open Access], or [Top-tier papers]. The list of papers is organized by columns such as [Insights], [TL;DR] (too long; didn't read), and [Limitations] to help researchers quickly understand the paper. You can add various new columns to this table using the [Create new column] button, such as conclusions, summarized abstract, results, summarized need, methods, and limitations of the study.

3) Chat with paper

If you want to learn more about a paper, you can click on the paper title to learn more about it. On the right side of the detailed paper screen, you'll see a dialog called Chat with Paper, which is where the generative AI comes into play. Researchers can ask Copilot AI specific questions about the paper and get answers to help them understand the paper. You can ask about the main conclusions of the paper, learn about the research methods applied in the paper, or ask about the implications of the research.

In addition to helping you identify and understand existing research papers, SciSpace includes other features to help researchers write their own papers. The menu on the left side of Scispace includes features like AI Writer, Paraphraser, and Citation Generator, with Paraphraser helping you rephrase paragraphs in your own words by adjusting the length and using variations.

3. Perplexity (https://www.perplexity.ai/), an AI tool that is strong in data search

Perplexity offers a unique "Focus" feature and "Search Tab" to provide a search experience similar to that of a search engine. The platform's Focus and Search tabs allow users to easily find the topic or information they are looking for. The Search tab provides summaries of related papers and the ability to ask additional questions about a topic, which is a key feature that has made Perplexity a useful tool for search and research tasks. The platform compresses information from multiple sources, provides references, and allows for communication through search. Next, let's take a closer look at Perplexity's key features and how to use them.

1) Search for Perplexity Utilization

The biggest advantage of Perplexity is its ability to combine multiple search results into a single summary document, providing quick answers. This not only saves time but also reduces unnecessary clicks during the search engine usage. The platform supports natural language searches, allowing users to ask questions in an intuitive way and receive answers. Additionally, it includes the sources of the referenced data in the responses, minimizing the need for further searches. Perplexity has access to YouTube and Reddit, offering a search menu based on popular topics, which also helps reduce the time spent selecting search terms.

2) Collections feature: customizable information organization

Perplexity provides a collections feature that allows users to organize and fine-tune their chat conversations. This feature allows users to structure information and easily access retrieved data by category as needed.

3) Ability to search and utilize academic papers

Perplexity provides the ability to search academic papers. If your professors use perplexity for their classes or research, this feature will be most useful to you. Perplexity preserves the original wording, summarizes it into paragraphs, links to sources, and makes fewer errors than ChatGPT. It may not always find the paper you're looking for, but it will provide the source and include a related video in the answer. You can change the interface language from English to Korean in Settings, and you can adjust it to not learn your data.

4. DeepL ( https://www.deepl.com/ko/translator ), AI for Translation

DeepL started in 2008 as Linguee, an online dictionary in Germany. Since its establishment, Linguee has collected millions of texts in dozens of languages. DeepL's engineers applied machine learning, neural networks, and artificial intelligence to this linguistic repository, creating one of the most powerful and accurate machine translation tools on the market. DeepL Translator was launched in 2017 and currently supports over 30 languages and more than 800 language pairs.

1) Translate text 

DeepL's text translation feature has a user interface similar to Google Translate and Papago, making it easy to understand intuitively. DeepL automatically detects the language when you input text and translates it into the language the user wants. Additionally, its neural machine translation (NMT) system understands the overall context, allowing for more accurate and fluent translations. One notable advantage of the translation feature that distinguishes it from Google Translate is that users can click on any word they want to edit and make changes directly. When you click on a specific word you want to change, the AI automatically suggests alternatives that fit the context better. This feature allows for smoother translations, which is a significant advantage. For example, in the image below, clicking on the word "이정표" shows options like "지표," "기준," and "전환점" as possible translations.

2) Translate the file

DeepL also supports file translation. By uploading files such as PDFs, docx, pptx, or txt, users can translate the entire file into Korean (or another language) and download it in the same format. Images embedded in the file will not be translated, but will remain in place, which is advantageous for translating English files. For the free version, the file size is limited to 10 MB.

3) Download and install the app version

Finally, the most powerful feature of DeepL is that once you download it as an app, you can connect it to other English websites and programs. First, go to DeepL and go to [Apps] > [Desktop App] to download the app and install it on your PC. Instead of copying & pasting the part you want to translate from the Deepl site, you can drag the part you want to translate from various documents such as Google Docs, websites, etc. and press the Ctrl+c+c shortcut to open the app and translate it immediately. Being able to translate with shortcuts without having to open a new window makes your work more efficient and convenient. If the existing post is an editable file, you can insert your translation by pressing [Replace].

In this Teaching Tip, we've covered generative AI that professors can use in their research: Consensus for searching and analyzing papers, SciSpace for accessing and searching paper databases, Perplexity for strong data searches, and DeepL as a translation expert—all of these tools will aid efficient research and exploration, making them highly useful in graduate courses as well. Moving forward, the Center for Teaching and Learning innovation will provide information on various tools and methods that can be helpful for professors, so we appreciate your continued interest.

Reference

Consensus

- Consensus. (n.d.).  Consensus 홈페이지. https://consensus.app/

- Toolify.ai. (n.d.). Consensus AI: 실제 연구 논문을 검색하고 요약하는 인공지능 도구. https://www.toolify.ai/ko/ai-news-kr/consensus-ai-2625226

- Academic English Now. (2023, 4, 10). Automate your literature review search with Consensus AI [비디오]. YouTube. https://www.youtube.com/watch?v=ntDF70L_Yug

- Walker, L. (2023, 3, 20). Consensus AI Research Assistant [비디오]. YouTube. https://www.youtube.com/watch?v=GpNogGPf3S8

- Stapleton, A. (2023, 1, 15). How To Use Consensus AI - Don’t Get Left Behind! [비디오]. YouTube. https://www.youtube.com/watch?v=I8VC6R7-J6M

SciSpace

- Ibrahimov, V. (2023, 2, 15). How to accelerate literature review with SciSpace AI? EASY TUTORIAL.  Medium. https://medium.com/@Vugar_Ibrahimov/how-to-accelerate-literature-review-with-scispace-ai-easy-tutorial-8d3de617b33a

- Burns, J. (2022, 11, 1). SciSpace: An all-in-one AI tool for literature reviews.  The Effortless Academic. https://effortlessacademic.com/scispace-an-all-in-one-ai-tool-for-literature-reviews/

- SciSpace. (n.d.).  SciSpace 홈페이지. https://typeset.io/t/about/

Perplexity

- Perplexity. (n.d.).  Perplexity 홈페이지. https://www.perplexity.ai/

- Brown, C. (2022, 10, 10). How To Use Perplexity AI For Beginners [비디오]. YouTube. https://www.youtube.com/watch?v=qpN-pjev-vM

- Stapleton, A. (2023, 5, 20). How To Use Perplexity AI For Research - Terrifyingly SMART! [비디오]. YouTube. https://www.youtube.com/watch?v=Ne-j_U2Vkfw

- 감자나라ai. (2023, 7, 5). [Full] 퍼플렉시티 사용법, Perplexity Tutorial [비디오]. YouTube. https://www.youtube.com/watch?v=7ImcGugl2NE

DeepL

- Techforword. (n.d.).  DeepL: More than just a translation tool. https://www.techforword.com/blog/deepl-more-than-just-a-translation-tool

     - Phrase. (2022, 9, 15).  DeepL: AI translation tool and its capabilities. https://phrase.com/blog/posts/deepl/

Sang-eun Lee, Min-young Ku, Ye-jin Kim (2024), Useful Generative AI Tools for Graduate Courses and Research (CTL Teaching Tips #47). Seoul: Sungkyunkwan University, Center for Teaching and Learning Innovation.