I am an Assessment Manager at Cambridge Assessment English, a non-teaching department of the University of Cambridge. I am responsbile for content production and process improvements using automation and machine learning techniques. I hold a Master's degree from Lancaster University in Language Testing and have been working in language assessment for over 10 years.
This research project takes an evidence-based approach to the identification of the criterial features of three different reading task types, namely, open cloze, multiple choice cloze and reading multi-item. Lexical content of these three task types varies depending on the proficiency level being tested with there being more complex advanced lexical chunks at higher proficiency levels and more simple at lower levels. Using corpus tools, the first part of this research project will analyse both part-of-speech (POS) and lexical 6-grams from the Cambridge Assessment English item bank for the above three reading tasks. This will enable us to place these n-grams on a continuum from A1 to C2 on the Common European Framework of Reference (CEFR) scale so that we can then identify out-of-range or problematic n-grams, either lexical or POS, for newly written reading tasks. With knowledge of what we are currently using in reading tests, I will then run analysis on L1 corpora to identify language which is not currently being used by Cambridge Assessment English.
The second part of this research is centred around the tested chunks of language around the gap in open cloze and multiple choice cloze task types. Items within these task types are tapping into lower level reading abilities, namely, word recognition, lexical access and syntactic parsing. Item Writer Guidelines (IWGs) specify that they should be placed into longer texts so that they appear authentic to the test taker, but is this extra cognitive load having a negative or positive effect on the candidate's ability? These items will be extracted with the minimal langauge required to answer the item correctly and then re-pretested and statistically compared for variance in difficulty and discrimination scores.
- MA Language Testing