Ideas

Got an idea about Haven OnDemand that you'd like to share with HP? Post it here! Before submitting your idea and for idea status definitions, read this page.
Search the Dev Community
Showing results for 
Search instead for 
Do you mean 
Post a Message
Status:
Under Review
Hi, I have been working with the ImageServer component which is behind the OCR API for a while now. For OCR it offers the capability to split the response by line, word or character. I have several ideas where this information would be really helpful - and I'm probably not the only one. So could you expose those options to the API? Almost as important: Please expose the confidence score ImageServers analyze function on the API! Cheers, Sebastian
Status:
Under Review

Sebastian,

Thank you for providing your feedback.

As part of our August and September releases we are planning a number of refinements to our OCR functionality and are certainly looking at how we can introduce the functionality you mention. We’ll be sure to update you further as we get closer to the August release.

Kind Regards
Chris

0 Likes
Status:
Under Review

Get Content API to return summary

Status:
Under Review
by arshad01 Level 4 on ‎10-12-2014 06:27

Hi

 

It would be great if the Get Content API can also return document summary like how Find Similar API does.

 

I have a workaround for this which I have given in one of my forum post (https://community.idolondemand.com/t5/Forum/Document-summary-with-GetContent-API/td-p/1074)

 

Thanks

Arshad 

Status:
Under Review

We're looking into this. Thank you for submitting your idea!

 

-Katie

0 Likes
Status:
Under Review

After much trial & error, I managed to get custom flavor of text index to accept some of my custom fields.

 

It sure would be nice if */*_PARAM and similar worked.   Yes wildcards are a little dangerous in the wrong hands, but it's also a hassle to input each singularly.   

 

To top it off, it seems that today you can only change fields upon text creation.  This is also pretty annoying, especially when you have to input each one individually.   A personalized field template - my custom would be useful in lieu of supporting add/subtracting fields.

 

Plus, there doesn't appear to be anyway to see field configuration until data with those fields is indexed.  I get this is normal IDOL behavior, but it's pretty annoying to not see what is intended to happen before it happens.   This is especially so since the input form for custom fields seems picky in unclear ways.

Status:
Under Review

@jb_hp

Thank you for your input. Developer feedback is important for the improvement of our platform. We collate and review feature requests at our regular planning sessions.

 

Jenn

Haven OnDemand team

0 Likes
Status:
Under Review

For the entity extraction there are many categories where non-English countries are supported (e.g. phone numbers or license plates). But why not non-English names??

Would be great if those could be included at least for major languages such as German, French or Spanish.

Status:
Under Review
0 Likes
Status:
Under Review

Detect face expressions

Status:
Under Review
by Monty New Member on ‎09-03-2014 08:50

An extension that could be interesting for the Face Detection API would be to detect face expressions, and return it along with the age when "additional" is true.

 

Something that could be defined like:

 

"expression": {
    "type": "string",
    "enum": [
        "happy",
        "sad",
        "angry",
... "neutral" ] }

 

and would look like the attached picture (I wanted to put it inline, but it seems that it is not allowed.)

  

In this line, there are a couple of things that could be added to the Face Detection API that could make it more interesting (at least for me):

 

  • Position/size of face features: eyes, nose, mouth,...
  • Some "basic" actions: smiling, winking, blinking,...
Status:
Under Review

We're investigating this. Thank you for sharing your idea!

 

-Katie

0 Likes
Status:
Under Review

Could you look at adding the following to the entity extraction API:

 

Stock ticker symbols for companies

More abbreviations or aliases for companies (e.g. "Coke" for Coca-Cola Company)

 

I can think of other categories along increasingly narrow use cases (brand names, medical terms, etc.).  Would it be possible to load custom grammars into this API?

 

Status:
Under Review
0 Likes
Status:
Under Review

I was trying out some photos that had the HSBC example logo.  It matched ok, but it also matched 2 other logs seemingly incorrectly.   I'm guessing that the scores are relatively low for seemingly bad matches.

 

http://www.panoramio.com/photo/56859617

 

To help tell which match is which, color coding the boxes to the responsedata would be nice.   The coordinates help some, but that requires some translation. 

 

This same issue applies to face detection - it's not that easy to tell which face's data goes with which face.

Status:
Under Review
0 Likes
Status:
Under Review

Clustering function

Status:
Under Review
by uday Level 3 on ‎05-05-2015 09:22

Hi,

 

Any idea when you provide clustering functionality of IDOL on premise would be available in IDOL on Demand.

I could see only related concepts. But we would like to make use of the clustering functionality which is available in IDOL on premise via IoD.

 

Status:
Under Review
0 Likes
Status:
Under Review

Just a thought coming out of untoward incidents in the airline industry.

 

There was a crowd-sourced initiative recently to figure out every inch of the global map on recognizing images to come to a location at last. I beleive this was done primarily out of the google maps satellites.

 

If we do come up with a way to access these image pixels(of course would need some forerunner program that can feed the image recognition api's - which I tried to access, but was not successful with), it might help narrow down a location or a single pixel in the map, and save time - it will speak a lot for the community at large.

Status:
Under Review

Hi Essen,

 

We have plans to open up our image technology for users to be able to training their own datasets, while the exactly scenario you outline may not be possible, it will allow your users to come up with new and interesting ways to use our image recognition technology.

 

We'll be sure to keep you updated when it is available.


Regards

 

Chris

Haven OnDemand team

0 Likes
Status:
Under Review

It would be useful if the field names for things like DRETITLE and DREDATE and such were configured appropriately for title, date and maybe even reference fields with maybe a special text index flavor template for this use case.  

 

It would make trying out HOD so much easier when one is coming from IDOL on-premise.   On-premise CFS is doing lots of data manipulation where there appears to be ZERO equivalent of CFS in HOD right now.

 

 

Status:
Under Review

@jb_hp

 

Thank you for your input. Developer feedback is important for the improvement of our platform. We collate and review feature requests at our regular planning sessions.

 

Jenn

Haven OnDemand Team

0 Likes
Status:
Under Review

Summarization of Content

Status:
Under Review
by Quantubly New Member on ‎07-01-2015 08:42

Whilst ConceptExtraction gathers he occurance of words within a text, it doesn't highlight the more valuable sentances which is the first step to natural summarization of a blob of wordy data.

 

Would a feature be added that splits word data into sentances. Then for each sentance which has the occurance of words that would appear in the ConceptExtraction API a weighting could be given. It could be as simple as a score equivalent to the number of times a word appears in the document, and that number applied to each word in the sentance. For example;

 

These are the number of times each word occurs from the ConceptExtraction API;

Joe = 10

Cristoph = 12

Meg = 15

 

And a sentance in the doucment was;

Joe and Cristoph think Meg is tip top

 

Then that sentance would get a score of 37

 

This would allow me to pull, potentially, the more important sentances from a chunck of data.

Status:
Under Review
0 Likes
Status:
Under Review

Object Store API

Status:
Under Review
by uday Level 3 on ‎04-08-2015 05:29

Hi,

 

Today we are using Object Store API to store the documents and uses the reference in IoD api's.

Once i store file in object store there is no way to delete the file.

 

Please let me know what other operations i can do on the file stored in object store.

It would be great if you provide us the methods to manage the file in object store like update and delete of the file

 

Status:
Under Review

Hi uday,

 

Additional functionality around the store object API are planned for the long term future.

 

We'll be sure to update you as these plans progress.


Regards

 

Chris

Haven OnDemand team

Status:
Under Review

Sentiment analysis - fine tuning.

Status:
Under Review
by Essen Level 2 on ‎09-12-2014 08:32

The sentiment analysis works as best as the dictionary supports. But this will never be the case in real time application, particularly with the social media. It might also differ from one geography to the other. How about opening up keyword mapping to the specific geographic community, controlled by a moderator. This would end up with a very rich and accurate analysis of sentiments, and can be targeted at varying geographies as well.

Status:
Under Review

Essen,

 

Thank you for taking the time to provide your suggestion. 

 

We are looking at option to introduce customisation to the Sentiment Analysis APis, but have not finalised this plans at present.

 

We'll be sure to update you when we have more details.


Regards

 

Chris

0 Likes
Status:
Under Review

Image Count

Status:
Under Review
by Murali_V Level 1 on ‎02-08-2015 08:48

This is a use case that has practical use.  In manufacturing plants they get a truck loads of raw material that need to be counted manually. The current process is predominantly manual, time consuming and also inaacurate. As as example let me say a truck load of pipes. As simple use case if I take that all pipes of the same diameter, then given a sample image, it will be great take a photo and have the number of pipes counted. In fact with pipes of different diameter, it will be still great to have the number of pipes of different diameter counted. May be when trucks enter a manufacturing facility / plant if photos can be a automatically clicked and the count displayed on a small screen it will be all the more useful.

As a use case I would like to try this.  Would any API allow adding sample images to the database?

 

Thanks,

Murali

Status:
Under Review

Hi Murali,

 

We are looking to add training API for our image recognition.   We will be sure to update you as our plans progress.

 

Regards

 

Chris

Haven OnDemand

Social Media
† The opinions expressed above are the personal opinions of the authors, not of HPE. By using this site, you accept the Terms of Use and Rules of Participation