Scientific　publications　and　patents　are　usually　viewed　as　respective　proxies　of　scientific　research　and　technical　development.　There　is　considerable　effort　spent　towards　establishing　topic　linkages　between　science　and　technology　with　the　lexical-　or　topic-based　approaches.　However,　due　to　the　heterogeneity　between　scholarly　articles　and　patents　in　terms　of　purpose,　statement,　and　quality,　the　performance　is　not　satisfactory.　To　understand　the　difficulties　of　topic　linkages　and　improve　the　performance,　a　framework　is　proposed　to　detect　the　commonality　and　specialty　between　scientific　publications　and　patents　from　the　two　perspectives:　linguistic　characteristics　and　thematic　structures.　Extensive　experimental　results　on　the　DrugBank　dataset　discover　five　commonness　and　five　significant　differences　in　terms　of　linguistic　characteristics.　For　example,　nouns　are　used　most　frequently　among　them,　and　scientific　publications　contain　more　word　tokens　than　patent　documents,　but　patents　have　usually　longer　sentences　and　use　more　clauses.　In　the　meanwhile,　common　and　special　thematic　structures　are　also　uncovered　between　scientific　publications　and　patents.　The　themes　about　general　description　in　the　pharmaceutical　field　are　shared　by　two　heterogeneous　resources.　The　scientific　publications　tend　to　explain　the　disease　mechanism　and　the　medication　content,　while　patents　bias　towards　the　preparation　and　practical　application　of　drugs.