Unknown　word　recognition　technology　is　of　great　significance　to　improve　the　precision　of　text　segmentation　and　syntax　analysis.　Social　network　has　become　an　important　platform　for　sharing,　disseminating,　and　acquiring　information.　Unknown　word　recognition　based　on　micro-blog　short　text　has　become　a　research　hot　spot,　while　the　micro-blog　text　contains　a　large　number　of　nonstandard　terms　and　network　buzzwords,　which　has　increased　the　difficulty　of　unknown　word　recognition.　This　paper　proposes　a　Chinese　unknown　word　recognition　method　for　micro-blog　short　text　based　on　improved　FP-growth　(POS-FP).　Firstly,　the　POS-FP　algorithm　is　used　to　get　frequent　itemsets　from　micro-blog,　and　the　N-grams　model　is　used　to　filter　out　unknown　words　from　frequent　itemsets.　Secondly,　the　improved　mutual　information　and　left-right　information　entropy　are　used　to　verify　the　internal　features　of　candidate　unknown　words.　Then,　context-dependent　and　open-source　methods　are　used　for　external　verification　of　candidate　unknown　words.　Compared　with　traditional　methods,　this　algorithm　improves　the　recognition　rate　of　unknown　words　in　micro-blog　short　texts.