I have used starter-pack-rasa-stack bot , when i gave input message as my name , it is treated as None. Here it is the files GitHub - RasaHQ/starter-pack-rasa-stack: 🎒Everything you need to build your first contextual chatbot with the open source Rasa Stack - includes files and training data
Your name is probably not being captured by the bot. Add a sample utterance with your name in the nlu file. It should work then.
It is fine. But if we change the name some times it is working , some times not.
Names are difficult to capture as there can be large variations. Try using the spacy’s PERSON entity, it will improve. Also, there need to be sufficient examples for it to work properly.
Alternatively, you can use a lookup table for identifying names.
Do you have any documentation for that and some other features
Here you can find about lookup tables: dataformat You can store all the names in this format and provide a few examples in your nlu file.
Spacy has a few inbuilt entity identifiers like PERSON, DATE, TIME. You can you these to improve entity extraction. To view all the inbuilt ones, take a look at this: spacy entities
Another link from RASA docs: nlu
Thanks @srikar_1996 , Is there any example bot in git hub
Here are a few sample bots by RASA: examples
Hi @srikar_1996 , By using spacy’s PERSON also it is not working
How many training examples do you have? And for what all cases is it failing to capture? I used to use the PERSON entity initially and it worked fine but since I had to use a variety of names I started using the lookup table now.
These are my files nlu_data.md (1.8 KB) <a class=“attachment” domain.yml (435 Bytes) nlu_config.yml (246 Bytes) stories.md (779 Bytes)
By using lookup tables it recognizes the names starting with capital letters only.
I looked at your files, I’m guessing you want to capture Indian names? I suggest using a lookup table. Add this to your domain file lookup:PERSON and add all the names you want in the .txt file or under the above line. After that you need to include a few examples from your lookup table in the nlu file. Atleast 3~4 names which are in your lookup table need to be there under the name intent.
Also, this will work properly if you have a similar sentence structure. One word replies are difficult to capture, usually for all the one word replies you can have a general intent like inform which captures these things.