❔ Best, most reliable NER (Named Entity Recognition) Library for .NET?
I'm looking for a NER library that is supported with .net framework & .net core.
I do not want to use services like google or azure to do my NER as for my use case it is very mild and I don't want to pay for something so small.
Upon looking at NER libraries, there isn't really a clear direction on the most straightforward to use, reliability and well documented. I'm just looking for a NER library like SpaCy. I know there's SpaCy.NET but don't really know how well it performs as it isn't a full implementation and seems to be dead. Looking for something more maintained.
Overall I am looking for
*something easy to use
*well maintained
*fairly efficient and uses some sort of prediction algorithm to help predict a few things based on context (optional but would prefer)
*one that works in .net core and framework, or just one or the other but in this case framework will be what I am using
The closest so far I have found was Stanford NER but it has so many dependencies so was looking to see if anything better exists. It can be a wrapper to anything as well as long as it's fairly straight forward to use
9 Replies
I don't suppose you could explain what Named Entity Recognition is
never heard of such a thing, but there may be things out there that just aren't described under that terminology
Yea I didn't either until someone else told me a while back and it fit my needs.
NER is an NLP, it's like a way of analysing text and catgeorizing certain things based on your text. For example, if I had a bunch of text talking about stock markets. Each sentence will have some things in common, you can tag each word/sentence as an entity and train the model. Eventually, the model will start learning based on what you fed it and then any other sentence/essay/page that it reads it will be able to classify the text and label them thouroughly. You can then use these labels to get data depending on your needs.
Some do it to use it to determine sentiment, some use it to detect what kind of text it is (like say you had a bot scanning marketplaces, with NER you'd be able to know the condition of the product, if it's damaged, what kind of damage and so on) at ease. So it's basically just a way of getting data from text in an easy manner without having to hardcode a bunch of words and structures yourself. I guess it's a form of AI?
I myself am not too deep into the subject but was trying to give it a go to see if it is actually useful. I had a perfect mini project that would prove if it was or wasn't. But yea, most NER is with python and trying to find a well maintened library that does NER on C# is kind of hard. Barely any info on it and documentation is scattered all over the place.
Here is a visual of NER tagging a post with a pre trained model (not mine of course)
Just to give you an idea of what it does. This person is using it to categorize crypto posts
ahhhhh
yeah, I got nothing, sorry
well it depends really on how much you would need to pay a cloud service for that, how much is your traffic, would that really be that much? over the complication of searching a package for that?
Yea no worries
A lot. Not worth it really.
They give you 5000 text records a month and apparently 1000 characters counts as one. Overall for a month it won't fit my needs.
Payment side of things it can range from 700 per month or 1 dollar per half million. But I'm not sure what the catch is with that.
Meanwhile you can do it free on python. Was hoping there was one with c# that wrapped around one with a full implementation. But yea I think I'm better off just writing a mini python script and then redirecting stdout if need be. But would prefer full control with c# itself though.
eh also it depends on the complexity of the problem
if you take a library that doesn't update, and you have to deal with ml stuff, and you lose hours to integrate it it's not great either
Fairly small complexity. Just takes a look at text that gets posted from a feed, The text gets categorized based on a few simple keywords (that can be in random orders/pronunciation) and then the software creates the appropriate class and inserts into a db with the appropriate object.
NER is perfect for this, but the issue is the lack of well maintained and fully functional libraries (with full feature support) in C#
NER is perfect for this, but the issue is the lack of well maintained and fully functional libraries (with full feature support) in C#
Was this issue resolved? If so, run
/close - otherwise I will mark this as stale and this post will be archived until there is new activity.