AAC CORE VOCABULARY ANALYSIS: TOOLS FOR CLINICAL USE
Katya J. Hill* and Barry A. Romich**
*Edinboro University of Pennsylvania
Edinboro, PA 16444
**Prentke Romich Company
Wooster, OH 44691
Vocabulary selection is an important component of the augmentative and alternative communication (AAC) service delivery process. Automated language activity monitoring (LAM) provides the quantitative data for identification and analysis of AAC language representation method(s) and vocabulary use patterns. LAM tools to facilitate the analysis of AAC core vocabulary for use with specific individuals or as a tool to develop databases for core vocabulary lists for specific populations are proposed. Implications are significant in the areas of clinical intervention, outcomes measurement, and research.
The goal of AAC intervention is to provide the supports and services that result in the most effective communication possible for the individual. Various AAC modalities such as natural speech, vocalizations, gestures, sign, and/or eye gaze may be employed for overall communicative competence. However, communicative effectiveness using assistive technology is accomplished through spontaneous novel utterance generation (SNUG). For people who rely on AAC systems, accomplishing spontaneous, interactive communication or SNUG is highly dependent on random access to core and extended vocabulary.
Vocabulary selection has been a major topic of AAC research. Utterance formulation is contingent upon the selection of vocabulary items to support conversational communication. Differences in age, gender, social roles, and environments are just a few of the factors that influence vocabulary needs. Current techniques used to identify vocabulary needs such as ecological inventories, communication diaries, and team brainstorming are time consuming, cumbersome, and ineffective.
Various investigations have documented the vocabulary-use patterns of individuals in daily environments. These investigations have supported the notion of vocabulary core lists and provide one source to identify core vocabularies for individuals who rely on AAC systems. Vocabulary selection questionnaires based on composite core lists are in development to facilitate this selection process (1). However, an even more efficient vocabulary source than composite lists are word lists compiled from the past performance of the specific individual for whom an AAC system is being developed (2). In addition, word lists based on language samples from like individuals, as from a language sample library, should be valid. Unfortunately with past procedures, obtaining and analyzing language samples from AAC users has been difficult, time consuming, and rarely done.
Recent developments in automated AAC language activity monitoring (LAM) provide the opportunity to obtain and analyze AAC language samples. The LAM makes quantitative data on vocabulary use readily available for clinical decision making. Since LAM data includes language event time stamps, the identification of language representation method (LRM) usage and the association of the LRM to specific vocabulary items is possible (3).
Quantitative measurement of vocabulary-use patterns of individuals who rely on AAC systems is now practical and efficient with the development of the LAM. Obtaining LAM data allows for analysis of vocabulary diversity including the identification of core and extended vocabulary items as part of the clinical intervention process. However, no automated tools for outcomes measurement have been available for clinical application and use. The objective of this work has been to develop core vocabulary analysis tools for clinical use based on LAM data.
This paper proposes procedures and tools for measuring vocabulary use-patterns from individuals who rely on AAC systems based on raw LAM data. Vocabulary diversity may be measured in terms of frequency counts, number of different words, types of words, and word list tables. However, this proposal is based on obtaining language samples from LAM for the specific purpose of identifying core vocabulary usage patterns and frequency counts.
Generally, the first process applied to raw LAM data is the editing of the data into utterances, e.g. “I love chocolate,” from the excerpt below. However, the presently proposed tools can be used without utterance segmentation. First, any pre-stored message is eliminated from the database, since these words would seem to be misleading in calculating spontaneous vocabulary use. Second, the LRM(s) used to select the vocabulary items can be identified. Spelled words are distinguished from predicted words. Once the vocabulary items have been tagged indicating the LRM(s), the clinician can perform various analysis procedures using the core vocabulary analysis tools.
These proposed procedures include: 1) alphabetical listing of the words in the sample with a frequency count; 2) frequency of occurrence listing of the words in the sample; 3) comparing the sample list against a reference list to identify word commonalties or matches; 4) calculating the percentage for sample list matching reference list; 5) alphabetical ordering of word matches and not matches; 6) frequency of occurrence of word matches and not matches. In addition, by tabulating the cumulative occurrence in the frequency of occurrence word listing, identification of core vocabulary is possible.
An excerpt from a narrative language sample using LAM is reported in the following model:
11:14:15 "went "
11:15:04 ". "
11:15:48 "I "
11:15:53 "love "
11:16:02 "chocolate "
11:16:57 "yes "
11:17:03 "but "
11:17:07 "we "
11:17:12 "had "
Use of the core vocabulary analysis procedures and tools allows the clinician to review the vocabulary-use patterns contained in this storytelling narrative about attending a concert. Identification of the LRM usage reveals that the individual used Semantic Compaction 83%, spelling 16%, and pre-stored messages 1% of the total utterances. This performance can be compared with pilot study data from high-end users that reveals LRM usage patterns of 93% use of Semantic Compaction, 6% use of spelling (including word prediction), and 1% use of pre-stored messages on AAC systems that provide access to all LRMs.
Use of the proposed tools then allows identification of the words that have been spelled and/or predicted (if word prediction is available). Examples of spelled words from the above language sample include: over, weekend, symphony, dessert, trouble, parking. By running the core vocabulary analysis features, these words can be compared against compiled reference core vocabulary lists or a previous vocabulary list of the individual. With this information the clinician will be able to make decisions regarding vocabulary management. Vocabulary management decisions include answers to these questions: 1) what is core vocabulary for the individual? 2) What is extended vocabulary for the individual? 3) How do both core and extended vocabulary relate to the reference or source list(s)? 4) Are the most effective LRM(s) being used for core?
AAC vocabulary selection and management is an ongoing process that must be individualized, dynamic, and periodically revised. Research on AAC vocabulary selection has relied on language sampling of able-bodied individuals to compile word frequency lists for specific populations. Few research studies have compiled word frequency lists from augmented communicators. No research has been based on the performance of augmented communicators who rely on the language representation method(s) supported by today’s technology. Consequently, clinicians and augmented communicators have little research to weigh the relative value of vocabulary selection and management decisions.
Fried-Oken and More (4) proposed use of a vocabulary source list and database to study vocabulary selection using formal semantic analysis. The core vocabulary analysis tools presented here will provide procedures to compare various source lists, establish a database on various populations, and allow for formal semantic analysis. The cumulative result of the use of these tools will be improved clinical intervention and substantive gains in improved spontaneous, interactive (SNUG) communication of people who rely on AAC.
1. Light, J., Fallon, K., & Paige, T.K. (1999). Vocabulary selection tool for preschoolers who require AAC. American Speech-Language-Hearing (ASHA) Convention. San Francisco, CA.
2. Yorkston, K., Smith, K., & Beukelman, D. (1990). Extended communication samples of augmented communicators: I. A comparison of individualized versus standard vocabularies. Journal of Speech and Hearing Disorders, 55, 217-224.
3. Hill, K. & Romich, B. (1999). AAC automated language activity monitoring. American Speech-Language-Hearing (ASHA) Convention. San Francisco, CA.
4. Fried-Oken, M., & More, L. (1992). An initial vocabulary for nonspeaking preschool children based on developmental and environmental language sources. AAC Augmentative and Alternative Communication, 8, 41-54.
The National Institute for Deafness and Other Communication Disorders of NIH has provided funding to Prentke Romich Company to support the work on LAM.
Katya J. Hill, M.A. CCC-slp
Edinboro University of PA
Assistive Technology Center
Edinboro, PA 16444