Electronic Digital Speech Corpus (EDSC)

The Electronic Digital Speech Corpus is a collection of recorded speech samples that have been digitized, annotated, and made publicly available for research and development in speech technology. Its primary goal is to provide a highquality, reproducible dataset that can be used for tasks such as acoustic modelling, phonetic analysis, speaker identification, and language learning tools.

Key Features

• Multispeaker coverage: Recordings from a balanced set of speakers differing in age, gender, regional accent, and speaking style.
• Phonetic richness: The corpus contains a wide variety of phonemes, diphthongs, and prosodic patterns, allowing for detailed phonetic research.
• Highresolution audio: Sample rates of 44.1kHz or higher, with 16bit or 24bit depth, ensuring minimal loss of acoustic information.
• Timealigned transcription: Wordlevel and phonemelevel alignments are provided in standard formats (e.g., TextGrid, ELAN).
• Open licensing: Most versions are released under Creative Commons or similar licences, encouraging wide reuse.

Applications

Researchers and developers use the EDSC for:

• Training automatic speech recognition (ASR) systems.
• Testing texttospeech (TTS) synthesis pipelines.
• Studying linguistic variation across dialects.
• Building language learning applications that require accurate pronunciation feedback.

Access and Tools

The corpus is typically hosted on university repositories or on platforms such as OpenSLR and Linguistic Society. Users can download the raw audio files, annotation files, and metadata in bulk or via a searchable web interface. Many projects also provide readymade scripts for loading the data into popular machinelearning frameworks like Kaldi, TensorFlow, or PyTorch.

Searchable Dictionary Database of the Tamil Verb

Tamil, a classical Dravidian language spoken by over 70 million people, has a rich verbal system that encodes tense, aspect, mood, polarity, and politeness. A searchable dictionary database specifically for Tamil verbs aims to capture this morphological complexity while offering an intuitive interface for scholars, language learners, and developers.

Why a VerbCentric Database?

Verbs are the grammatical engine of any language. In Tamil, a single root can generate hundreds of inflected forms through agglutination. Traditional printed dictionaries present only a limited set of paradigms, making it difficult to:

• Identify all possible suffix combinations for a given root.
• Understand subtle semantic shifts caused by auxiliary verbs.
• Retrieve examples of usage in modern contexts.

A searchable, digital repository solves these problems by providing dynamic generation of forms, crossreferencing, and corpusbased examples.

Core Components

1. Lexical Entries

Each verb entry contains:

• Root form (e.g., to study).
• Partofspeech tag and semantic field.
• Verb class (, , etc.) and conjugation type.
• Transliteration and IPA pronunciation.

2. Morphological Engine

A rulebased engine generates the full paradigm ondemand. It handles:

• Finite tenses (present, past, future) and nonfinite forms (infinitive, participle).
• Aspect markers (perfective, progressive).
• Mood and modality (imperative, conditional, desiderative).
• Politeness levels and honorifics.

The engine draws on a finitestate transducer model that can be extended as new morphological patterns are discovered.

3. Corpus Integration

Realworld usage examples are drawn from a curated Tamil corpus containing news articles, literature, and spoken transcripts. Users can search for any generated form and instantly see frequency, collocations, and example sentences.

4. Search Interface

Features include:

• Rootbased search (enter the base verb).
• Formbased search (type any inflected form).
• Advanced filters for tense, mood, or register.
• Export options (CSV, JSON) for linguistic analysis.

Impact on Research and Education

The database supports a wide range of activities:

• Computational linguistics: Training morphological analyzers and neural sequencetosequence models.
• Secondlanguage teaching: Interactive exercises that let learners generate and test verb forms.
• Lexicography: Providing uptodate entries for printed dictionaries and digital encyclopedias.
• Dialects studies: Comparative browsing of regional variations in verb morphology.

Getting Started

To explore the database, visit the public portal at tamilverbdb.example.com. No registration is required for basic queries; a free account unlocks batch download and API access.

Developers can integrate the API into applications using endpoint examples such as:

GET /api/verb?root=&tense=present&polite=true        

This request returns a JSON object with all presenttense polite forms of , together with usage examples.