Frequently Asked Questions

INDEXA Similarity uses a rigorous dual-layer evidence accounting framework to compute originality:

• Local Repository Comparison: Computes exact and near-match shingles across your institutional repository using cryptographic MinHash and Jaccard similarity.
• External Open-Access Retrieval: Fetches verified full-text publications from global indexes (Crossref, OpenAlex, HAL, ASJP, DOAJ) and matches overlapping text.
• Integrity Score: The mathematical complement of verified similarity (100% - Net Similarity).

University scientific councils typically establish institutional defense clearance thresholds based on degree and field:

• PhD Dissertations: Usually require ≤ 10% to 15% net similarity (excluding bibliography and canonical statutory citations).
• Master's Theses & Magister: Generally set at ≤ 15% to 20%.
• Undergraduate & Graduation Projects: Commonly permitted up to ≤ 20% to 25%.

Yes. INDEXA features smart section boundaries to prevent bibliographies from falsely inflating similarity scores:

• Tail-Confined Bibliography Detection: Automatically identifies reference markers (References, Bibliographie, المصادر والمراجع) located after the 65% length threshold, protecting literature reviews and tables of contents from accidental truncation.
• Quotation Marks Filter: Correctly ignores text formatted inside standard quotes ("...", «...», “...”).
• Statutory Canonical Exemption: Sacred verses, constitutional clauses, and canonical legal statutes are recognized and exempted from plagiarism penalties.

Many legacy Arabic PDF creation tools store characters in visual reverse order or use corrupt font CMaps (e.g. converting جامعة to ةعماج). INDEXA v2.4.14+ resolves this through intelligent multi-stage heuristics:

• Teh Marbuta Detection: In valid Arabic grammar, no word can begin with ة. The presence of words starting with ة is a 100% indicator of visual reversal.
• Automatic OCR Fallback: When reversed streams or non-standard glyphs (U+063B..U+063F) are detected, INDEXA triggers visual OCR to extract the clean, human-readable text layer.
• Boilerplate Suppression: Strips institutional cover page titles (university, faculty, supervisor names) so the genuine research topic is captured.

Absolutely NOT. Unlike third-party commercial cloud scanners that upload your faculty's intellectual property to global external repositories (reselling students' papers):

• All manuscript texts and hashes remain strictly within your private WordPress database.
• External comparison queries send only anonymized keyword hashes and title queries to scholarly APIs.
• Your university retains 100% intellectual property sovereignty and GDPR compliance.

INDEXA integrates with the world's most authoritative open-access scholarly indexes:

• Crossref: Over 150 million registered DOI publications, preprints, and conference proceedings.
• OpenAlex: Global knowledge graph indexing 250M+ scientific concepts and open-access works.
• HAL (Archives Ouvertes): The French National Science Archive featuring 3.5M+ peer-reviewed full-text papers.
• ASJP & Regional DSpace: Algerian Scientific Journal Platform (CERIST) and MENA university thesis repositories.
• Unpaywall: Direct, instant resolution of legal open-access PDFs for over 50,000 academic journals.
• Semantic Scholar & PubMed / arXiv: AI-indexed medical, physics, and computer science archives.

Rather than blindly blasting queries to 20 different sources simultaneously (which introduces unnecessary network overhead), INDEXA dynamically analyzes the manuscript's linguistic fingerprint:

• French Manuscripts: Automatically routes parallel queries to HAL, Crossref, and OpenAlex.
• Arabic Manuscripts: Prioritizes ASJP, regional DSpace institutional repositories, and Crossref.
• English Manuscripts: Routes to Semantic Scholar, Crossref, and OpenAlex.

This delivers the most relevant peer-reviewed papers in under 600ms while conserving server CPU and network bandwidth.

Most commercial publisher landing pages (e.g. Elsevier, Springer, Wiley) hide article texts behind paywalls or heavy JavaScript wrappers. In the past, this led to zero full-text retrieved.

Unpaywall solves this legal access challenge:

• When a DOI candidate is discovered, INDEXA queries the Unpaywall API (api.unpaywall.org/v2/{doi}) in less than 250ms.
• If a legal open-access version exists in any university repository worldwide, Unpaywall returns the exact, direct PDF URL.
• INDEXA immediately downloads the PDF directly, boosting full-text comparison coverage by over 300%.

International repositories such as Crossref and OECD iLibrary frequently index single charts, Excel spreadsheets, and statistical tables under individual DOIs (e.g. Tableau A.35. Prévalence du tabagisme).

These components contain zero prose narrative text. Attempting to download them wastes HTTP connections and exhausts retrieval budgets. INDEXA pre-filters out items marked as dataset, component, or titled with chart prefixes, ensuring your retrieval budget is focused 100% on genuine peer-reviewed articles.

• Academic Mode (Recommended): Full multi-provider scholarly scan querying Crossref, OpenAlex, HAL, Semantic Scholar, and Unpaywall.
• Turbo Mode: Ultra-fast parallel scan querying 2-3 focused databases with zero secondary query delay (perfect for rapid pre-submission draft screening).
• Regional Mode: Specializes in North African & Arab repositories (ASJP, AJOL, DSpace Algiers, KAUST, AUC).
• Deep Web Mode: Incorporates Google Scholar SerpAPI and scholarly web search engines.
• Local Only: Instant private scan strictly against your university's internal thesis and manuscript repository.

INDEXA deploys a robust multi-tier mathematical comparison pipeline:

• N-gram Shingling: Deconstructs text into rolling 5-word to 8-word shingles to detect exact and paraphrased passages.
• MinHash & Locality-Sensitive Hashing (LSH): Generates compact mathematical fingerprints for sub-millisecond local corpus matching.
• Jaccard Similarity Coefficient: Calculates exact statistical overlap ratios between document shingle sets.
• Levenshtein Dynamic Edit Distance: Pinpoints near-duplicate sentences even when small synonym substitutions or word order variations occur.

Students and ghostwriters often use evasion techniques to fool basic text scanners. INDEXA's Digital Document Forensics neutralizes these maneuvers automatically:

• Zero-Width Characters: Detects and strips hidden unicode characters (U+200B, U+FEFF, U+200C) inserted inside words to break shingle tokens.
• Cyrillic & Latin Homoglyphs: Identifies visual twin letters (e.g. Cyrillic 'а', 'е', 'о' mixed into French or English words) and normalizes them back to the host alphabet.
• Invisible / White Text: Flags 1pt font sizes and color-masked text layers designed to manipulate word counts.
• Caesar Character Shifts: Automatically unrolls font-subset shifting before comparison.

AI generation (ChatGPT, Claude, Gemini, DeepSeek) exhibits distinct statistical language regularities compared to authentic human academic prose. INDEXA assesses AI signal probability through four stylometric indicators:

• Perplexity Uniformity: Human writing displays high variance in perplexity (mix of simple and complex sentences), whereas LLMs produce unnaturally smooth, predictable token sequences.
• Burstiness Scoring: Measures rhythm and sentence length diversity.
• Syntactic Template Repetition: Detects overused AI transitional phrases and formulaic paragraph closures.

Large dissertations (60,000 to 120,000 words) exceed PHP memory and execution time limits if scanned in a single monolithic pass. INDEXA's Chunked Scanner processes them cooperatively:

• The manuscript is cleanly divided into logical chapter sections or bounded word windows.
• Each section is analyzed with its own fingerprint and transient state.
• If a server request finishes, progress is stored safely in the database and automatically resumes in the next AJAX request (status: processing).
• The final report reconstructs cumulative chapter-by-chapter similarity heatmaps.

INDEXA supports all major academic file formats with native UTF-8 sanitization:

• PDF (.pdf): Text-based PDFs via native parser with automatic OCR fallback for scanned pages.
• Word (.docx & legacy .doc): Deep XML text and footnote extraction via PHPWord.
• OpenDocument (.odt) & Plain Text (.txt, .rtf): Full text and structural headings extraction.
• Encodings: Native UTF-8, Windows-1256 (Arabic), ISO-8859-1 (French/Western European), and Windows-1252.

Every certificate generated by INDEXA includes triple-layer institutional verification:

• Unique Certificate Identifier: E.g. IDX-2026-1RFBJESB5XJO permanently recorded in your university database.
• Dynamic QR Code: Scanning the QR code with any smartphone camera immediately opens the live verification page on your institution's domain.
• Cryptographic SHA-256 Hash: Confirms the exact manuscript text content evaluated at the time of clearance.

Yes. INDEXA includes dedicated A4 print engineering:

• Strict 1-Page A4 Guarantee: Features an optimized @media print stylesheet that removes browser URLs, headers, and UI buttons.
• Direct PDF Download: A dedicated "Download PDF Certificate" button client-renders the document into a high-resolution 300DPI vector PDF ready for official signing and stamping.
• Multilingual Switching: Instantly switch the certificate language between English, French, and Arabic before printing.

The Certificate of Integrity is an official administrative credential presented to scientific councils and defense juries. Low-level diagnostic counters (e.g. "Full-text sources retrieved: 5; retrieval failures: 2") are technical debugging data intended for journal editors and system administrators, not defense juries.

In v2.4.14+, this diagnostic ledger is preserved inside the administrative report and database, while the Certificate displays clean, authoritative compliance indices.

• The Interactive Report: An in-depth analytical tool for authors and reviewers showing side-by-side matching sentence snippets, source URLs, shingle highlights, digital forensics, and full text diffs.
• The Certificate of Integrity: A single-page official attestation bearing the student's name, thesis title, faculty threshold clearance, QR verification code, and formal institutional signatures.

Certificates are saved permanently in the wp_indexa_certificates database table. Verification links and QR codes do not expire, ensuring that five or ten years later, university committees can still authenticate historical thesis clearance records.

Yes, absolutely. INDEXA v2.4.16 is engineered with a high-throughput, non-blocking cloud architecture:

• Zero Background Daemons: Requires no Node.js, Python, or Redis servers; runs entirely within standard PHP/WordPress execution.
• Smart Open Access Prioritization: Open Access repositories (ArXiv, HAL, ASJP, Wikipedia) are prioritized before commercial publisher DOIs, guaranteeing higher full-text hit rates.
• Adaptive Micro-Batch Guard: Dynamically regulates external repository fetches within optimal execution windows, preventing server thread saturation.
• Low Memory Footprint: Strict 8MB file response caps keep execution well within 256MB RAM allocations.

On shared hosting, slow external academic servers hanging for 30+ seconds would trigger server gateway timeouts. INDEXA v2.4.16 prevents this through:

• Dynamic Stopwatch Check: The full-text retrieval loop checks microtime() before each candidate.
• Graceful 22s Time Guard: If total elapsed time reaches 22 seconds, further external attempts are halted gracefully, and the system transitions to finalize the score and report with zero 504 timeouts.
• Concurrent Multi-cURL: External queries are dispatched in parallel batches using Requests::request_multiple in ~600ms.

On shared hosting servers, multiple websites share the same outbound IP address. If multiple rapid requests hit OpenAlex from that IP, OpenAlex temporarily returns HTTP 429 (Too Many Requests).

INDEXA v2.4.16 handles this with zero downtime:

• Immediate Failover: If OpenAlex returns 429, INDEXA immediately skips sequential retries to avoid wasting time.
• Polite Pool Headers: Automatically passes a validated mailto: parameter to access the polite pool.
• Dynamic Fallback: The search seamlessly relies on Crossref, HAL, or Semantic Scholar to ensure the scan never returns empty.

INDEXA caches external search queries, DOI metadata lookups, and Unpaywall responses in WordPress Transients for 7 to 30 days:

• If another student or researcher uploads a paper on a similar topic, external search responses return in 0 milliseconds directly from local cache.
• Zero external HTTP requests are made for previously verified DOI publications.
• Significantly reduces server CPU utilization and bandwidth consumption.

Yes. INDEXA includes a fully secured REST API (/wp-json/indexa/v1/) supporting:

• Automated Document Submission: Programmatically submit manuscripts and receive webhook scan completions.
• HMAC SHA-256 Token Handshake: Single Sign-On (SSO) links allowing students and reviewers to view reports without separate WordPress user logins.
• OJS (Open Journal Systems) Plugin Integration: Automatically trigger similarity audits when articles are submitted for peer review.

Yes, seamlessly. INDEXA v2.4.16 features an advanced Institutional Team & Seats Architecture designed specifically for university faculties, research laboratories, and scientific journal editorial boards:

• Master Institutional Wallet: A central university or publisher account holds the master quota of scan credits, eliminating the need to issue individual payment cards to faculty members or editors.
• Child Seat Provisioning: The institution can register sub-accounts for Chief Editors, Section Reviewers, or Department Professors with distinct roles and independent login credentials.
• Atomic Credit Allocation & Reclaim: Supervisors can grant any number of credits to a team member in real-time, or instantly reclaim unused credits back into the institutional pool at any time.
• Unified Audit Oversight: Central administrators can monitor scan activity, certificate issuance, and compliance logs across all affiliated sub-accounts.

INDEXA v2.4.16 is engineered to handle large-scale academic dissertations, doctoral theses, and multi-volume journal articles with strict memory safety:

• Maximum Upload File Size: Accepts files up to 50 MB across supported academic formats (PDF, DOCX, DOC, ODT, RTF, TXT).
• Document Length & Word Capacity: The intelligent Chunked Scanner processes comprehensive dissertations up to 120,000 words (250+ pages) seamlessly by splitting long manuscripts into cooperative batches to prevent server exhaustion.
• Network Payload Safety Cap: Outbound external full-text retrievals are strictly capped at 8 MB per candidate, preventing slow academic servers from exhausting PHP RAM.
• Long Token Filter: URLs, DOIs, and citation hashes are automatically excluded from Document Forensics manipulation scoring, guaranteeing zero false-positive warnings for lengthy scholarly references.