Skip to main content icon/video/no-internet

Computerized databases for the study of communication disorders have grown over the past several decades. Many of these databases involve the collection of large amounts of spoken or written language accessed over the Internet. Conditions for access to these collections vary widely. The TalkBank System is currently the world’s largest repository of shared databases for spoken language and is freely open to researchers, educators, and clinicians. Its many component banks provide language corpora and resources for research and education in fields such as communication sciences and disorders, linguistics, psychology, education, and computer science. These databases include language samples of children and adults of all ages, in English and many other languages, and with and without clinical impairments.

Each of the TalkBank language banks has a separate web page with an index of its collection, providing corpus names, corpus contributors’ names, and short descriptions of the corpus contents. Each of the corpus names links to a page with more detailed information about the corpus as well as links for downloading the transcripts and media. TalkBank’s browsable database allows users to view video samples (or listen to audio samples) along with their corresponding transcripts. This entry explains how TalkBank works and then describes eight of the databanks available on the TalkBank website.

About TalkBank

Language samples in the TalkBank system are transcribed in CHAT (Codes for the Human Analysis of Transcripts) format and linked to the digitized audio and video files. CHAT is designed to operate closely with a set of programs called CLAN (Computerized Language Analysis), which permit the analysis of a wide range of linguistic and discourse structures. Automatic morphosyntactic analysis programs are built into CLAN for English and 10 other languages, producing a “%mor” line and “%gra” line below each speaker line in the transcript.

The %mor line provides complete morphosyntactic tagging for the utterance above, and the %gra line provides a grammatical dependency analysis. CLAN commands can be used to perform computations such as total words, total utterances, type-token ratios, word frequency analyses, mean length of utterance (in words or morphemes), words per minute, proposition density, lexical diversity, part-of-speech analyses, and grammatical complexity. Two omnibus utility programs, EVAL and KIDEVAL, allow users to compare an individual participant’s expressive language data with means and standard deviations for comparison databases from AphasiaBank and CHILDES (Child Language Data Exchange System), respectively.

EVAL produces over 30 morphosyntactic and fluency measures, and KIDEVAL produces over 40 measures, including DSS (decision support systems), IPSyn (Index of Productive Syntax), and Brown’s morphemes. These programs are also used to compare an individual’s performance changes over time. Manual coding can be added to the transcripts for word-level (e.g., paraphasias) and utterance-level (e.g., speech acts, main concepts, coherence) behaviors, which can then be automatically searched and analyzed.

Together, the TalkBank databases and TalkBank software tools provide resources for the study of language, language acquisition, and language disorders. The repositories and corpora that relate most directly to the study of communication disorders are described next in alphabetical order, beginning with the child language sites followed by the adult language sites. Several of the adult language repositories contain corpora that use similar discourse tasks, allowing for comparisons across diagnostic categories. Some databases are password protected and available on request and provision of the necessary professional credentials.

...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles

Sage Recommends

We found other relevant content for you on other Sage platforms.

Loading