Language Understanding (LUIS)
Welcome to the LUIS API Forum
(Language Understanding Intelligent Service)
Categories
Documentation – Any ideas or suggestions for the API Reference or Documentation.
Language Support – Submit a request to have a particular language supported.
luis.ai WebApp – Ideas and suggestions for luis.ai and functionality used through the WebApp.
Programmatic API – Any ideas or feedback pertaining to features or enhancements to LUIS programmatic API.
Samples & SDK Request – Let us know if you would like to see a tutorial/sample or a SDK provided.
Attention!
We have moved our Customer Feedback & Ideas for Azure Cognitive Services portal to the Azure Feedback Forum.
-
Sentiment analysis
I am a Chinese developer in sentiment analysis. When i see your text English-Chineses translation, i find a mistake. Sentiment analysis(in English)-情感分析(in Chinese). You translate sentiment analysis as "情绪分析“, which is another term in natural language processing study field. Thanks.
By the way, I am a product manager of dialogue platform.
1 vote -
Can't create composite enities with V3 API
I'm not able to create composite entities. I use the V3 API.
1 vote -
Can't create composite enities with V3 API
I'm not able to create composite entities. I use the V3 API.
0 votes -
Including both a positive and negative sentence in an utterance while testing breaks sentiment analysis
When including both a positive and negative sentence in a single utterance like "i love you. i hate you." while testing, the sentiment section says "Enable sentiment analysis to get sentiment score" (see screenshot). Sentiment analysis is clearly enabled since it works for most utterances, and there's no option to enable or disable sentiment analysis in the test panel anyway. When querying a published endpoint, the opposing sentences seemed to be combined into a single "document" score correctly. More info about sentiment analysis v3: https://docs.microsoft.com/en-us/azure/cognitive-services/text-analytics/how-tos/text-analytics-how-to-sentiment-analysis?tabs=version-3#view-the-results
1 vote -
Cannot switch tenants
Cannot switch tenants
Subscription list only shows my tenant subscriptions2 votes -
Testing Panel Inspect function
In the Inspect function in the Testing Panel, previously you could see the top 10 intents returned plus their scores. This information has now been removed but was useful to check the clarity of the intent to ensure there are no potential overlapping intents.
4 votes -
Testing panel history
When using the testing panel, previously if you trained your model then used the test panel again, you could see your previous questions asked. In the new version released this week, the test panel is cleared once you train, so you need to type question again, very frustrating!
4 votes -
When editing an utterance / delete button deletes whole utterance
When editing an utterance, if you use the delete button to delete a letter/word from an utterance, LUIS thinks you want the delete the whole utterance
5 votes -
Newer Version of LUIS (V3), utterence search box is not working properly.
Newer Version of LUIS (V3), utterance search box is not working properly. Can it be fixed?
5 votes -
Order versions in LUIS authoring by "last-modified" descending
we are uploading automically via API most of our LUIS content.
since the last version, the order in "versions" manage is always "by name asceding".could you do it automatically by "last-modified" because one is always interested about the latest changes not about the oldeest
1 vote -
bok gibi site
aq sitesi
1 vote -
LUIS dashboard "Overall predictions"-chart wrong?
In the dashboard there's some data about prediction.
We move a model from different models through DTAP.
These models can go from 99% to 0% and from 0% to 99%.When you click on one of the intents under "Incorrect predictions" you get to the intent's utterances.
These all have pretty good scores and the "Nearest Intent" column is filled with -1.00 values.
Enclosed are two screenshots edited together to show my steps and confusion.I was wondering if there's something funky going on with maybe absolute values being compared to formulate the data in the dashboard.
2 votes -
geographyV2 should recognize US states
geographyV2 should recognize US states.
1 vote -
Line Breaks in Example Utterances
I'm trying to add support for subjects. Subjects almost always look like this:
"##Subject Text
Not subject text"However, because there's no support for line breaks in example utterances, I can't easily teach the bot using Machine Learning. I was able to work around this by using regex, but that's not a solution that always works.
I would like it if Utterances supported the ability to have line breaks during training.
1 vote -
Some intents not displayed in preview portal https://preview.eu.luis.ai/
Viewed 18 times
1
Some intents not displayed in preview portal https://preview.eu.luis.ai/
When accessing a model via the preview portal I can see only a subset of the intents compared to the production portal. I can see the missing intents in the old portal, or if I use the search function in the new portal. There is a bug in the display of the intent list.
1 vote -
Romanian Language Support
Hello, can you tell us when we should expect that Romanian language will be supported by Luis?
Thanks
3 votes -
LUIS API Error "Missing phraselist features field." does not provide enough information to fix the problem
When using the LUIS API, I am getting "Missing phraselist features field." - however, there is nothing in the API documentation or on the internets that explains what this is and how to fix it...
1 vote -
Prompt before deleting Entities from the Entity palette.
I miss-clicked on the X to delete an Entity from the Entity Palette, which immediately deleted it and all records of it. This could potentially lose someone a lot of time if they haven't saved their version in a while.
This was done on the preview site.
5 votes -
I can build a batch testing file creation tool
I can build a batch testing file creation tool and open source it on Github. Since it is recommended not to have duplicate intent and training data, It could help you make sure you aren't duplicating data when you enter utterance examples. It could also allow you to more easily build training data by knowing about your entities and allowing you to select them. It would prepare and allow you to download your training file
1 vote -
[Bug] Dispatch App Training evaluation shows 0% correctness
After refreshing a Dispatch-App, I often run into the problem, that the numbers in the Training evaluation are completely off and just show 0% correctness in regards to matching intents.
When I'm testing the App/Chatbot at the endpoint, everything works as intended. I also can't recognize a pattern yet, it just happens from time to time.
2 votes
- Don't see your idea?