When the custom dictionary is not applied

You registered a word in the custom dictionary but the analysis result stays the same? Most of the time the cause is a missing dictionary name, the wrong dictionary type, or a conflict between dictionaries. Let's go through it step by step with this page.

Symptoms

Even though you registered a word with UpdateCustomDictionary, in the morphological analysis result:

• the registered word still comes back as NA/UNK,
• out_of_vocab is still OUT_OF_VOCAB rather than IN_CUSTOM_DICT, or
• custom_dict_name is empty.

Causes

Suspect these three things.

Fix

1. Specify custom_dict_names in the analysis request. Creating a dictionary does not apply it automatically. You have to state the dictionary names to use in every request. You can give several, and the dictionary specified first takes priority.

   Copy

   {
     "document": { "content": "바른은 한국어 형태소 분석기입니다.", "language": "ko_KR" },
     "custom_dict_names": ["my-domain", "company-names"]
   }
   

2. Check that the dictionary type is right. The word must go into the DictSet that fits its nature.

   DictSet	Purpose
   np_set	Proper noun (NP)
   cp_set	Compound noun (CP)
   cp_caret_set	Compound noun splitting (mark the split point with ^)
   vv_set	Verb (VV)
   va_set	Adjective (VA)
   mm_set / mag_set / ic_set	Determiner / adverb / interjection

3. Check for conflicts. Use CheckConflict to check for duplicates within the same dictionary (duplicated) and conflicts between dictionaries (conflict). If the same headword is in NP/CP/CP^/VV/VA at the same time, it is reported as a conflict, and it is best to leave only the one type you intended.

4. Check that the update was applied. The custom dictionary is managed through the CustomDictionaryService API, and the files placed in the server's dictionary directory are binary (a format managed by the server). When you change it with UpdateCustomDictionary, Bareun detects file changes (additions, edits, deletions) in the dictionary directory in real time and also checks every 60 seconds, performing a zero-downtime reload. If you do not see the change applied right after, wait a moment and analyze again.

Prevention and tips

• You can tell that registration worked when the out_of_vocab of the morpheme in the analysis result changes to IN_CUSTOM_DICT and custom_dict_name is filled with the dictionary name.
• index.dup.txt is excluded from watching, so do not edit it directly.

Frequently asked questions

Q. Is a dictionary applied automatically once I register it?

No. You must specify the dictionary names to use in custom_dict_names in every analysis request for them to apply. You can specify several, and the one specified first takes priority.

Q. I changed the dictionary. Do I need to restart the server?

No restart is needed. Bareun detects file changes in the dictionary directory in real time and rereads the changes without downtime. A 60-second periodic check also runs alongside.

Q. Can I put the same word in several dictionaries?

If the same headword is in several types such as NP/CP/CP^/VV/VA at the same time, it is reported as a conflict. We recommend checking with CheckConflict and leaving only the one type you intended.

Q. Can I edit the dictionary files directly in a text editor?

We do not recommend it. The files placed in the server's dictionary directory are binary (a format managed by the server), so you do not edit them directly as .txt. Manage the custom dictionary through the CustomDictionaryService API (UpdateCustomDictionary). We recommend version-controlling the word list source in Git and applying that list through the API.