Arabic Speech Recognition System using CMU-Sphinx4

1 Système de Reconnaissance Automatique de l’arabe basé sur CMUSphinx H.

Entity Class

Research

Jurisdiction

Global

Characters Indexed

17K

Minted

2026-06-29

Source

https://arxiv.org/abs/0704.2201…

Pipeline

BST-PIPELINE-1.0.0

◈ Canonical Citation — Copy These URLs ROOT-LD · MANIFEST · INDEX

Canonical URL

https://boisestandard.org/ai/research/0704.2201/

Root-LD JSON

https://boisestandard.org/ai/research/0704.2201/root-ld.json

Manifest JSON

https://boisestandard.org/ai/research/0704.2201/manifest.json

Index JSON

https://boisestandard.org/ai/research/0704.2201/index.json

◈ Document Body — Full Text · 17K CharactersFTS5 INDEXED

H. Satori et al. 1 Système de Reconnaissance Automatique de l’arabe basé sur CMUSphinx H. Satori (1, 2), M. Harti (1, 2), and N. Chenfour (1, 2). (1) : UFR Informatique et Nouvelles Technologies d'Information et de Communication B.P. 1796, Dhar Mehraz Fès Morocco. (2) : Département de Mathématiques et Informatique, Faculté des Sciences, B.P. 1796, Dhar Mehraz Fès, Morocco E-mail: hsnsatori@yahoo.fr Abstract In this paper we present the creation of an Arabic version of Automated Speech Recognition System (ASR). This system is based on the open source Sphinx-4, from the Carnegie Mellon University. Which is a speech recognition system based on discrete hidden Markov models (HMMs). We investigate the changes that must be made to the model to adapt Arabic voice recognition. Keywords: Speech recognition, Acoustic model, Arabic language, HMMs, CMUSphinx-4, Artificial intelligence. Résume. Dans ce travail nous allons réaliser un système de Reconnaissance Automatique de la Parole (RAP) basé sur le CMU Sphinx4. Ce dernier est un projet Open Source de l’Université Carnegie Mellon. Nous allons démontrer l’adaptabilité de ce système pour la reconnaissance de la langue arabe. Mots-clés: Reconnaissance de la parole, Modèle acoustique, Langue arabe, HMMs, CMUSphinx-4, Intelligence artificielle. 1. Introduction La Reconnaissance Automatique de la Parole (RAP) est une technologie informatique permettant à un logiciel d’interpréter une langue naturelle humaine. Elle permet à une machine d’extraire le message oral contenu dans un signal de parole. Cette technologie utilise des méthodes informatiques des domaines du traitement du signal et de l’intelligence artificielle [1]. Les applications qu’ont peut imaginer sont nombreuse : aider les personnes handicapées, contrôle vocal des machines, réservation des vols, apprentissage d’autres langues, etc. [2]. Vue l’importance de la RAP, plusieurs systèmes ont été développés pour la reconnaissance vocale, parmi les plus connus: Dragon Naturally Speaking, IBM Via voice, Microsoft SAPI et d’autre. Aussi, il y a des Open Sources comme HTK [3], ISIP [4], AVCSR [5] et CMU Sphinx [6-8]. Nous sommes intéressés à ce dernier qui est un système basé sur les Modèles de Markov Cachés (MMC), en anglais Hiden Markov Models (HMM) [9]. Nous avons constaté que le système de reconnaissance de la parole CMU Sphinx 4 est librement disponible (Open Source) et il est actuellement l’un des systèmes de reconnaissance de parole les plus puissants. Le CMU Sphinx permet à des groupes de recherche avec des budgets modestes de développer et de conduire des applications de recherches dans la reconnaissance de la parole. Pour ces raisons et d’autres, nous avons choisi ce système pour développer notre application pour la reconnaissance de la langue arabe [10-11]. Notre travail s’inscrit dans le cadre général de la RAP il est parmi les premiers travaux traitant la langue arabe utilisant l’Open Source CMU Sphinx. Nous présentons dans ce trava

◈ Root-LD — Three-Layer Provenance Record HELIX POSITION 1

LAYER 1 — ANCHOR · IMMUTABLE · PROVENANCE CORE · FEDERATION ID · 2026-06-29

{
  "@type": "rld:Anchor",
  "rld:uuid": "b1f6f8a3-359f-4cb3-a629-c6d70c604487",
  "rld:federationId": "bs-9225ec04",
  "rld:contentHash": "51797eca292971d09005fdcc38f45e315782cdcdb669661e9e6579f1c84d0594",
  "rld:primarySource": "https://arxiv.org/abs/0704.2201",
  "rld:sourceDomain": "",
  "rld:sourceVerified": false,
  "rld:collectionMethod": "fetch",
  "rld:collectionDate": "2026-06-29T22:01:19.837422+00:00",
  "rld:generationMethod": "automated",
  "rld:humanVerified": false,
  "rld:specVersion": "1.0",
  "rld:mintedAt": "2026-06-29T22:03:26.515355+00:00",
  "rld:pipeline": "BST-PIPELINE-1.0.0",
  "rld:pipelineRunId": "BST-PIPELINE-1.0.0-2026-06-29T22:01:19Z",
  "rld:entityClass": "research",
  "rld:mintTier": "research",
  "rld:schemaType": "ScholarlyArticle",
  "rld:immutable": true,
  "rld:immutableNote": "This anchor is immutable after mint. Amendment or re-mint produces a new helix position. The prior record is preserved in rld:crawlRecord. Constitutional Law II: the timestamp is the record.",
  "rld:manifest": {
    "bs:entityClass": "research",
    "bs:mintTier": "research",
    "bs:schemaType": "ScholarlyArticle",
    "bs:hasBody": true,
    "bs:hasRegulatory": false,
    "bs:hasResearch": true,
    "bs:hasDomain": false,
    "bs:hasSupplyChain": false,
    "bs:hasEnforcementDate": false,
    "bs:hasPenaltyProvision": false,
    "bs:hasAuthors": false,
    "bs:hasDoi": false,
    "bs:jurisdiction": "global",
    "bs:corpusSlug": "research",
    "bs:charCount": 17325,
    "bs:pageCount": 5,
    "rld:toc": {
      "rootLd": "https://boisestandard.org/ai/research/0704.2201/root-ld.json",
      "manifest": "https://boisestandard.org/ai/research/0704.2201/manifest.json",
      "indexRecord": "https://boisestandard.org/ai/research/0704.2201/index.json",
      "body": "https://boisestandard.org/ai/research/0704.2201/#root-ld-body",
      "anchor": "https://boisestandard.org/ai/research/0704.2201/#root-ld-anchor",
      "recursive": "https://boisestandard.org/ai/research/0704.2201/#root-ld-recursive",
      "regulatory": null,
      "research": "https://boisestandard.org/ai/research/0704.2201/#root-ld-body/research",
      "domain": null,
      "supplyChain": null
    }
  },
  "rld:linkPod": {
    "bsCanonicalUrl": "https://boisestandard.org/ai/research/0704.2201/",
    "rootLdUrl": "https://boisestandard.org/ai/research/0704.2201/root-ld.json",
    "manifestUrl": "https://boisestandard.org/ai/research/0704.2201/manifest.json",
    "indexJsonUrl": "https://boisestandard.org/ai/research/0704.2201/index.json",
    "sourceUrl": "https://arxiv.org/abs/0704.2201",
    "corpusUrl": "https://boisestandard.org/corpus/research/",
    "vocabUrl": "https://boisestandard.org/vocab#"
  }
}

LAYER 2 — BODY · FROZEN AT MINT · COMPLETE MEASUREMENT SNAPSHOT · 17K CHARS

{
  "@type": "rld:Body",
  "@id": "https://boisestandard.org/ai/research/0704.2201/#root-ld-body",
  "rld:frozenAt": "2026-06-29T22:03:26.515355+00:00",
  "rld:bodyType": "research_corpus",
  "bs:identity": {
    "@id": "https://boisestandard.org/ai/research/0704.2201/#root-ld-body/identity",
    "title": "Arabic Speech Recognition System using CMU-Sphinx4",
    "slug": "",
    "entityClass": "research",
    "corpusSlug": "research",
    "bsCanonicalUrl": "https://boisestandard.org/ai/research/0704.2201/",
    "sourceUrl": "https://arxiv.org/abs/0704.2201"
  },
  "bs:research": {
    "@id": "https://boisestandard.org/ai/research/0704.2201/#root-ld-body/research",
    "doi": "",
    "arxivId": "0704.2201",
    "authors": [],
    "venue": "",
    "publicationDate": "",
    "researchCategories": [],
    "abstract": "",
    "citationCount": 0,
    "openAccess": false,
    "license": "",
    "note": "Research corpus record. Every field traces to the source. arXiv API or Semantic Scholar as primary source. Constitutional Law I."
  },
  "bs:provenance": {
    "@id": "https://boisestandard.org/ai/research/0704.2201/#root-ld-body/provenance",
    "sourceUrl": "https://arxiv.org/abs/0704.2201",
    "collectionMethod": "api",
    "collectionDate": "2026-06-29T22:01:19.837422+00:00",
    "contentHash": "51797eca292971d09005fdcc38f45e315782cdcdb669661e9e6579f1c84d0594",
    "pipelineVersion": "BST-PIPELINE-1.0.0",
    "mintedAt": "2026-06-29T22:03:26.515355+00:00"
  },
  "bs:textSummary": {
    "charCount": 17325,
    "fullBodyUrl": null,
    "excerpt": "##PAGE:1##\nH. Satori et al.\n\n1\n\nSystème de Reconnaissance Automatique de l’arabe basé sur CMUSphinx\n\nH. Satori (1, 2), M. Harti  (1, 2), and N. Chenfour (1, 2).\n(1) : UFR Informatique et Nouvelles Technologies d'Information et de Communication B.P.\n1796, Dhar Mehraz Fès Morocco.\n(2) : Département de Mathématiques et Informatique, Faculté des Sciences, B.P. 1796, Dhar\nMehraz Fès, Morocco\nE-mail: hsnsatori@yahoo.fr\n\n\nAbstract\n\n In this paper we present the creation of an Arabic\nversion of Automate…"
  }
}

LAYER 3 — RECURSIVE · EMPTY AT MINT · GROWS THROUGH CORPUS PASSES · LAW VII TORUS

{
  "@type": "rld:Recursive",
  "@id": "https://boisestandard.org/ai/research/0704.2201/#root-ld-recursive",
  "rld:edgeCount": 0,
  "rld:edges": [],
  "rld:appendedAt": [],
  "rld:pendingSlots": [
    {
      "rld:slot": "related_entities",
      "rld:status": "pending",
      "rld:pass": "Bidirectional Corpus Pass",
      "rld:note": "BS entities with semantic connections to this record"
    },
    {
      "rld:slot": "law_corpus",
      "rld:status": "pending",
      "rld:pass": "Bidirectional Corpus Pass",
      "rld:note": "USC titles, CFR sections, EU AI Act, Idaho law governing this entity"
    },
    {
      "rld:slot": "research_corpus",
      "rld:status": "pending",
      "rld:pass": "Research Corpus Pass",
      "rld:note": "arXiv papers, Semantic Scholar records referencing this entity"
    },
    {
      "rld:slot": "supply_chain",
      "rld:status": "pending",
      "rld:pass": "Supply Chain Corpus Pass",
      "rld:note": "Supplier, manufacturer, distributor, retailer edges — the full chain"
    },
    {
      "rld:slot": "domain_entities",
      "rld:status": "pending",
      "rld:pass": "Domain Corpus Pass",
      "rld:note": "Web entities semantically connected to this record"
    }
  ],
  "rld:mintNote": "Recursive layer initialized at mint. Empty by design. Edges form through accumulated corpus passes. Law corpus, research corpus, domain graph, supply chain — every connection this entity has in the world accumulates here. The graph builds itself. Constitutional Law VII — Torus.",
  "rld:recursiveSpecUrl": "https://root-ld.org/spec/1.0/recursive",
  "rld:initializedAt": "2026-06-29T22:03:26.515355+00:00"
}