I'm looking for a condensed form of the NEO-PI test. Current short forms come in at 40-50 questions, I'm wondering if it's possible to achieve a very basic degree of accuracy with 20 items.
There is a very short 10 item questionaire. The BFI-10 (see Rammstedt & John, 2007). I have never used it on my own, but i just read through it quickly. Although it seems fairly acceptable. So, if you want to use it, you should definitely read the article carefully.
Regarding your question: I would answer a careful 'yes'. In some cases it is possible to achieve a very basic degree of accuracy with just using a few items. However, this highly depends on the way the test is constructed and the kind of construct you want to investigate. The broader your construct, the more items you will need in order to account for every facet appropriately (e.g. a very broad personality aspect like 'extraversion' vs. a specific facet like 'assertiveness'). Furthermore, the longer your test, the better the reliability will tend to be (see Spearman-Brown Formula).
To sum up: It depends on the situation whether or not a short questionnaire can be basically accurate. You should always check for it´s reliability and validity.
Rammstedt, B., & John, O. P. (2007). Measuring personality in one minute or less: A 10-item short version of the Big Five Inventory in English and German. Journal of research in Personality, 41(1), 203-212.
If your aim is to measure the Big 5 only, then perhaps 4 or 5 items per factor will give you a rough idea. If you're planning on publishing the research, then I'd strongly encourage you get at least 8 items per factor. The IPIP website has a range of options for measuring the Big 5 with different length inventories: http://ipip.ori.org/newMultipleconstructs.htm There are many other Big 5 instruments available including the BFI and BFI 2.
If you need full facet-level measurement (i.e., the 30 facets), then I'd encourage you to use the full length instrument.