Add user documentation for SLR tool #616

koppor · 2026-02-03T08:43:46Z

Example still missing

koppor · 2026-02-02T09:37:39Z

Clear sign of AI geneation.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add user documentation for SLR tool #616

Uh oh!

Diff view

Diff view

There are no files selected for viewing

koppor Feb 3, 2026

Uh oh!

koppor Feb 2, 2026

Uh oh!

-Original file line number
+Diff line change
@@ -0,0 +1,292 @@
+    # Systematic literature review
+    JabRef supports managing systematic literature reviews by structuring research questions, search queries, catalogs, and results within a dedicated study directory created by JabRef.
+    This feature helps organize search configuration and collected references, but it does not replace methodological decisions made by the researcher.
+    ---
+    ## Starting a systematic literature review
+    A new study definition can be created via:
+    `Tools → Start new systematic literature review`
+    <figure>
+    <img src="../.gitbook/assets/slr_start-menu.png" alt="Start new systematic literature review dialog">
+    <figcaption>Starting a new systematic literature review</figcaption>
+    </figure>
+    This opens the **Define study parameters** dialog, where JabRef stores the configuration used for search and data collection.
+    ---
+    ## Study parameters
+    The study definition dialog contains several sections that determine how JabRef performs searches and stores results.
+    <figure>
+    <img src="../.gitbook/assets/slr-define-parameters.png" alt="Define study parameters window">
+    <figcaption>Study definition dialog</figcaption>
+    </figure>
+    ### Authors and title
+    This section defines the study metadata:
+    * Study title
+    * At least one author
+    These elements identify the study and are required before the survey can begin.
+    ---
+    ### Research questions
+    Research questions define the scope of the study within JabRef.
+    They help relate queries and later screening decisions to the study purpose.
+    Each question should be:
+    * Clear
+    * Focused
+    * Defined by the research goal rather than tool settings
+    <figure>
+    <img src="../.gitbook/assets/slr-questions.png" alt="Research questions tab">
+    <figcaption>Defining research questions</figcaption>
+    </figure>
+    ---
+    ### Queries
+    Queries define how JabRef retrieves literature from the selected catalogs.
+    <figure>
+    <img src="../.gitbook/assets/slr-queries.png" alt="Queries tab of SLR setup">
+    <figcaption>Defining search queries</figcaption>
+    </figure>
+    Queries typically combine keywords using Boolean operators:
+    * `AND`
+    * `OR`
+    * `NOT`
+    Queries can be adjusted later if search results need refinement.
+    ---
+    ### Catalogs
+    Catalogs define the external services that JabRef queries for literature.
+    <figure>
+    <img src="../.gitbook/assets/slr-catalogs.png" alt="Catalog selection in SLR">
+    <figcaption>Selecting catalogs</figcaption>
+    </figure>
+    Examples include:
+    * ACM Portal
+    * arXiv
+    * DBLP
+    * Crossref
+    * IEEE Xplore
+    Using multiple catalogs increases coverage and reduces the risk of selection bias.
+    ---
+    ### Finalize
+    In this step, a directory for storing the study is selected.
+    ⚠ **The selected study directory must be empty before starting the survey.**
+    When the survey begins, JabRef initializes the directory by creating the internal study structure, configuration files, and result database.
+    The **Start survey** button becomes available only after the study definition is complete. The definition must include:
+    * At least one author
+    * Study title
+    * At least one research question
+    * At least one query
+    These requirements ensure that JabRef can execute searches and store results using the defined study structure.
+    ---
+    ## After starting the survey
+    Starting the survey triggers the search and data collection process using the defined study parameters.
+    During this process:
+    * Queries are executed on the selected catalogs
+    * Retrieved records are written into the study database
+    * The study directory structure is populated with configuration and result data
+    References can then be:
+    * Screened
+    * Filtered
+    * Tagged
+    * Exported
+    <figure>
+    <img src="../.gitbook/assets/slr-study-folder.png" alt="SLR study folder structure">
+    <figcaption>Generated study directory structure after starting the survey</figcaption>
+    </figure>
+    ---
+    ## Example Study
+    The example demonstrates how Jabref supports systematic-literature reviews.
+    ### Study Title
+    Machine Learning Techniques for Phishing Detection: A Systematic Study
+    ---
+    ### Research Questions
+    * RQ1: What ML techniques are used in research for phishing detection?
+    * RQ2: How is the performance of ML models measured for phishing detection?
+    * RQ3: What are the open challenges in implementing ML techniques for detecting phishing?
+    These research questions mainly focus on the technical side of phishing detection research. RQ1 aims to identify the approaches proposed for detecting phishing attacks. We can then categorize what methods actually exist and hence understand trends in those applied techniques. RQ2 focuses on how researchers analyse and evaluate their models, including the performance metrics. Finally, RQ3 highlights the challenges and the research gaps. This helps identify directions for future research.
+    <figure>
+    <img src="../.gitbook/assets/example-research-slr.png" alt="Research Questions on Jabref">
+    <figcaption>Research Questions on Jabref</figcaption>
+    </figure>
+    ---
+    ### Search Strategy
+    The goal of the search is to gather relevant research studies on machine learning techniques applied to phishing detection.
+    #### Queries
+    ("phishing" OR "phishing website" OR "phishing email") AND ("machine learning" OR "ML" OR "artificial intelligence" OR "deep learning")
+    <figure>
+    <img src="../.gitbook/assets/example-queries-slr.png" alt="Queries on Jabref">
+    <figcaption>Queries on Jabref</figcaption>
+    </figure>
+    #### Catalogs
+    The search was conducted using major academic bibliographic databases to ensure wide coverage of research in computer science and security. The databases include IEEE Xplore, CiteSeerX and Crossref.
+    #### Inclusion-Exclusion criteria
+    To ensure that only relevant and high-quality studies are selected, inclusion and exclusion criteria are defined.
+    * Inclusion: Studies must focus on phishing detection as the primary problem and apply machine learning techniques. Studies must report experimental evaluation or performance analysis of the proposed methods.
+    * Exclusion: Studies are excluded if they do not use machine learning techniques or are not related to phishing detection. Studies are also excluded if they are duplicates, opinion papers, tutorials, or if the full text is not accessible.
+    ---
+    ### Study Selection Process
+    Initial search results: 120 studies
+    After title screening: 77
+    After abstract screening: 66
+    After relevance filtering: 59
+    Further quality filtering: 34
+    Finally selected primary studies: 24
+    In JabRef, these stages can be represented using groups for screening levels, keywords to mark inclusion or exclusion decisions, and notes to record screening justifications.
+    <figure>
+    <img src="../.gitbook/assets/initial_research.png" alt="Initial search results">
+    <figcaption>Initial Search Results</figcaption>
+    </figure>
+    <figure>
+    <img src="../.gitbook/assets/title_screening.png" alt="After Title Screening">
+    <figcaption>After Title Screening</figcaption>
+    </figure>
+    <figure>
+    <img src="../.gitbook/assets/abstract_screening.png" alt="After Abstract Screening">
+    <figcaption>After Abstract Screening</figcaption>
+    </figure>
+    <figure>
+    <img src="../.gitbook/assets/relevance-filtering.png" alt="After Relevance filtering">
+    <figcaption>After Relevance filtering</figcaption>
+    </figure>
+    <figure>
+    <img src="../.gitbook/assets/quality-filtering.png" alt="After Quality filtering">
+    <figcaption>After Quality filtering</figcaption>
+    </figure>
+    <figure>
+    <img src="../.gitbook/assets/selected_studies.png" alt="Selected Primary studies">
+    <figcaption>Selected Primary Studies</figcaption>
+    </figure>
+    ---
+    ### Data Extraction
+    For the 24 selected papers, some basic info was written in the Notes section of each entry in JabRef. Things noted included:
+    * Which ML method the paper used
+    * What it tries to detect (email, URL, website, etc.)
+    * What type of input or features were used
+    * Which performance metrics were reported
+    * Any strengths or limitations mentioned in the paper
+    This shows how JabRef can be used to store small notes about each study while doing the review.
+    #### Notes related to RQ1- Techniques observed
+    * Many papers use traditional ML like Random Forest, SVM, XGBoost
+    * These depend on manual features like URL, email headers, text patterns
+    * Deep learning is also common — CNN, LSTM for text/email/web content
+    * Transformer models are used for text classification in some studies
+    * Recent work includes Large Language Models
+    * Some papers fine-tune transformers
+    * Some use in-context learning or multi-agent setups
+    * A few try self-supervised or contrastive learning
+    #### Notes related to RQ2- Performance evaluation
+    * Most studies report accuracy, precision, recall, F1-score
+    * Traditional ML often performs well and is lighter
+    * Deep learning and transformers show higher results with more data
+    * LLM-based systems tested for understanding realistic phishing content
+    * Some papers test against adversarial or modified inputs
+    #### Notes related to RQ3- Challenges observed
+    * Phishing emails and websites look more real now
+    * Normal URL/text features are not always enough
+    * In some studies, datasets are old
+    * New types of phishing are not much covered
+    ---
+    ### Outcome of the example
+    This example shows how Jabref supports structuring an SLR workflow, organising screening stages, and storing extracted notes for selected studies.
+    ---
+    ## Best practices
+    * Define research questions before constructing queries
+    * Start with broader queries and refine if needed
+    * Use multiple catalogs to retrieve records from different sources
+    * Document queries used in the study definition
+    * Use tags and groups to organize screening decisions
+    This supports consistent and reproducible handling of search results in JabRef.

Add user documentation for SLR tool #616

Uh oh!

Add user documentation for SLR tool #616

Uh oh!

Uh oh!

Diff view

Diff view

There are no files selected for viewing

koppor Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

koppor Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!