Using text mining for study identification in systematic reviews: a systematic review of current approaches

摘要: The large and growing number of published studies, their increasing rate publication, makes the task identifying relevant studies in an unbiased way for inclusion systematic reviews both complex time consuming. Text mining has been offered as a potential solution: through automating some screening process, reviewer can be saved. evidence base around use text not yet pulled together systematically; this review fills that research gap. Focusing mainly on non-technical issues, aims to increase awareness these technologies promote further collaborative between computer science communities. Five questions led our review: what is state base; how workload reduction evaluated; are purposes semi-automation effective they; have key contextual problems applying field addressed; challenges implementation emerged? We answered using standard methods: exhaustive searching, quality-assured data extraction narrative synthesis synthesise findings. active diverse; there almost no replication or collaboration teams and, whilst it difficult establish any overall conclusions about best approaches, clear efficiencies reductions potentially achievable. On whole, most suggested saving 30% 70% might possible, though sometimes accompanied by loss 5% (i.e. 95% recall). Using prioritise order which items screened should considered safe ready ‘live’ reviews. ‘second screener’ may also used cautiously. eliminate automatically promising, but fully proven. In highly technical/clinical areas, with high degree confidence; more developmental evaluative work needed other disciplines.