Work

Transcribing, searching and data sharing: The CLAN software and the TalkBank data repository.

Year: 2010
Type: article
Abstract: To get a flavor of CLAN’s potential, we advise the reader to test the output of the program by looking at the TalkBank database on the net.1 We suggest that you go to www.talkbank.org, open the Browsable Database folder, proceed to CABank, and then open the folder Examples in your browser. When you open one of the files in the folder and the transcription has come up on your screen, click on a transcript line and the browser will play sound or video while highlighting the transcript line by line. Note that it is not necessary to install and run CLAN to do play transcriptions and to work with the data. However, to produce transcriptions, the CLAN program has to be installed. It currently runs under Windows and Macintosh, and files are movable across the platforms. In the remainder of this paper we will describe 7 different issues related to the transcription and analysis process based on the CLAN tools developed by Leonid Spektor at Carnegie Mellon2 University: Recording and Data formats: This process involves the use of camcorders and audio recorders to record interactions. Because the interactional contexts involved vary widely, a variety of different recording methods are required. Transcription: When transcribing, the researcher listens to the recording and produces a transcript. The preferred practices of transcribing vary in the research community and, consequently, the CLAN editor supports several styles of transcribing. The transcription is directly linked to the data file. Analysis: Once a basic transcription is complete, a researcher will work on the data to identify interactional phenomena to be studied in detail. The researcher may then wish to add additional features, codes, and annotations to the transcript. Searching: In this operation, a researcher attempts to locate specific interactional phenomena across multiple transcripts. Searching may focus on activities, keywords, phrases, or codes. It is also possible to track the co-occurrences between words and conversational features such as overlap, interruption, loudness or whispering. Collections: To further support searching and detailed analysis, researchers build up collections of transcripts or transcript fragments that illustrate particular interactional phenomena. Publication: Researchers typically use digital editors like Microsoft Word to compose articles for publication. These articles will often include segments from transcripts. Sharing: Once an interaction has been transcribed, the researcher may wish to share both the transcript and the media in a research project, with students and colleagues. 2. CLAN Overview CLAN has two parts: the programs and the editor. The programs include 24 command-line analysis and search programs, 7 programs for morpho-syntactic analysis, and 35 utility programs for data reformatting (MacWhinney 2000). We will discuss the programs further in the sections 5 and 6 on Ancillary Analysis and Searching. The file format in CLAN is called CHAT and all files are saved as xxx.cha. CLAN has existed for more than two decades and has a large user community in several fields. The program is developed according to needs suggested by its users. CLAN uses a sweeping update system. New versions with new, expanded and de-bugged features are continuously uploaded on the CHILDES website. All changes are documented in the CLAN manual which can be downloaded from http://childes.psy.cmu.edu/clan. For transcription and playback, the relevant part of CLAN is the editor. The editor uses many of the same conventions as Microsoft Word. However, unlike Word, it allows the researcher to link individual segments of the transcript directly to the audio or video media. This form of linkage to the media is crucial in terms of allowing users to playback transcriptions to verify their accuracy. The idea of linking transcripts directly to media was introduced to systems like CLAN, COALA, and SyncWriter in the early 1990s. Since then, new editors such as ELAN, Transcriber, ANVIL, EXMARaLDA, Praat, Phon, and all incorporated this same feature. In this regard, these digital transcription editors differ markedly from Microsoft Word, since Word does not allow the user to link the transcript to the media. CLAN has four major features that recommend it to research on interaction. These strengths include: Flexible playback and transcription: CLAN provides several methods for transcribing data and for linking transcripts to audio and video files, which we will discuss in detail below. Full structuring: A major strength of CLAN is that it provides a complete system for converting the Jeffersonian transcription conventions to an unambiguous and semantically accurate format. This format is based on the XML coding scheme that is slowly replacing HTML as the basic method for computerized structuring of documents and data on the web. The important thing for our current purposes is that, using additional facilities, XML can be extremely well structured and computationally explicit. Using these methods, CLAN provides the only method currently available for full digital structuring of CA-style transcriptions. Interoperability: CLAN includes utility programs to convert CHAT into the formats required for ELAN, EXMARaLDA, and Praat. Moreover, it can convert from each of these formats back to CHAT and we can show that data will go through this forward and backward conversion without any change or error. Because it is not possible to convert to and from TransAna, we do not recommend use of TransAna. Database linkage: CLAN relies on the CHAT format that is used throughout the TalkBank (talkbank.org) and CHILDES databases (childes.psy.cmu.edu) – the two largest databases for spoken language. Because all of these data are in CHAT, users of CLAN have good access to these databases for playback and further analysis. The major challenges of CLAN for its users derive from its strengths. The XML schema provides an explicit and unambiguous characterization of how talk is composed of words, turns, pauses, overlaps, prosodies, and other features of a detailed transcription. These explicit definitions and formats must all be used accurately and unambiguously. New symbols and characters can be added, but this must be done in a systematic fashion throughout the database and programs. To use the system correctly, users must learn how to format each transcription symbol in precise accordance with the CHAT manual. A second challenge for users of CLAN is that the search programs currently have no user-friendly graphic user interface (GUI). Because of these two weaknesses, initial learning of CLAN can require some days. By comparison, users can learn a relatively unstructured system as TransAna in a day or less. However, the transcripts produced by such unstructured systems are far more limited in terms of the support they provide for analysis, interoperability and data maintenance. We will now examine in greater detail the use of CLAN for interaction analysis.
Source: PubMed
Language: German
Open Access status: green
OpenAlex ID: W23095260
FWCI: 10.03
Cites: 50
Cited by: 156
Related to: 9