diff --git a/ucsinfer/__main__.py b/ucsinfer/__main__.py
index b4cb286..4ca2550 100644
--- a/ucsinfer/__main__.py
+++ b/ucsinfer/__main__.py
@@ -103,6 +103,11 @@ def recommend(ctx, text, paths, interactive, skip_ucs):
     catlist = [x.catid for x in inference_ctx.catlist]
 
     for path in paths:
+        _, ext = os.path.splitext(path) 
+        
+        if ext not in (".wav", ".flac"):
+            continue
+
         basename = os.path.basename(path)
         if skip_ucs and parse_ucs(basename, catlist):
             continue 
diff --git a/ucsinfer/recommend.py b/ucsinfer/recommend.py
index 6c92abd..5378df9 100644
--- a/ucsinfer/recommend.py
+++ b/ucsinfer/recommend.py
@@ -1,10 +1,13 @@
 # recommend.py
 
 from re import match
+
 from .inference import InferenceContext
 
+from tabulate import tabulate
+
 def print_recommendation(path: str | None, text: str, ctx: InferenceContext, 
-                         interactive_rename: bool):
+                         interactive_rename: bool, recommend_limit=10):
     """
     Print recommendations interactively.
 
@@ -16,7 +19,7 @@ def print_recommendation(path: str | None, text: str, ctx: InferenceContext,
               `print_recommendation` should be called again with this argument.
             - if retval[2] is a str, this is the catid the user has selected.
     """
-    recs = ctx.classify_text_ranked(text)
+    recs = ctx.classify_text_ranked(text, limit=recommend_limit)
     print("----------")
     if path:
         print(f"Path: {path}")
@@ -28,7 +31,7 @@ def print_recommendation(path: str | None, text: str, ctx: InferenceContext,
         print(f"- {i}: {r} ({cat}-{subcat})")
 
     if interactive_rename and path is not None:
-        response = input("#, t [text], ?, q > ")
+        response = input("(n#), t [text], c [cat], ?, q > ")
         
         if m := match(r'^([0-9]+)', response):
             selection = int(m.group(1))
@@ -43,16 +46,27 @@ def print_recommendation(path: str | None, text: str, ctx: InferenceContext,
             text = m.group(1)
             return True, text, None
         
-        elif m := match(r'^c (.*)', response):
+        elif m := match(r'^c (.+)', response):
             return True, None, m.group(1)
 
+        elif m := match(r'^b (.+)', response):
+            expt = []
+            for cat in ctx.catlist:
+                if cat.catid.startswith(m.group(1)):
+                    expt.append([f"{cat.catid}: ({cat.category}-{cat.subcategory})", 
+                              cat.explanations])
+            
+            print(tabulate(expt, maxcolwidths=80))
+            return True, text, None
+            
         elif response.startswith("?"):
             print("""
 Choices:
 - Enter recommendation number to rename file,
-- "t [text]" to search for new recommendations based on [text]
+- "t <text>" to search for new recommendations based on <text>
 - "p" re-use the last selected cat-id
-- "c [cat]" to type in a category by hand
+- "c <cat>" to type in a category by hand 
+- "b <cat>" browse category list for categories starting with <cat>
 - "?" for this message
 - "q" to quit
 - or any other key to skip this file and continue to next file
@@ -60,6 +74,8 @@ Choices:
             return True, text, None
         elif response.startswith('q'):
             return (False, None, None)
+        else:
+            print()
     else:
         return None