Since　the　criteria　for　teaching　evaluation　may　change　periodically　and　different　subjects　may　need　distinct　quality　indexes,　a　fuzzy　linguistic-based　system　for　teaching　assessment　is　designed　to　meet　these　requirements.　One　of　the　key　modules　of　the　system　is　its　model　for　assessment.　Traditional　assessment　model　is　based　on　interval　estimates　that　calculate　means　and　standard　deviations,　but　it　cannot　work　well　under　incomplete　information.　A　new　assessment　model　is　thus　proposed　to　concentrate　on　the　uncertain　information　of　different　attributes.　Numeric　examples　show　that　the　proposed　new　model　can　produce　more　reasonable　evaluation　results.　©　2011　IEEE.