I am trying to extract relation between two entities (entity1- relation- entity2) from news articles for stock prediction. I have used NER for entity extraction. It would be great if anyone could help me with relationship extraction.
Relation Extraction is a difficult task in NLP and most of the time there's not a one-size-fits-all solution to that. Depending on the task that you're trying to solve, I would suggest reading some literature about it on Google Scholar and see if there's something similar to what you're trying to do.
Sometimes, authors are kind enough to publish the code of their solution, which are mainly PyTorch/Tensorflow models (hopefully) trained on a specific dataset. One example is this paper.
If you want to stick with Spacy, there are some guides that might help you, but I'm not sure how well it could scale with the task that you need to solve.
Another more basic approach could be to just extract the shortest path between two entities in the semantic graph of a sentence. This might be quite limited, but can be fairly easy to implement.
One final idea that comes to mind is to use encoders and compute the similarity between sentences. If you're doing multi-class classification, this could help solving your problem.
Hope you find something useful among these.
I am starting to work with LLVM, clang, and passes. This is a developed field, and I quickly become overwhelmed with the wealth of information out there. I was hoping someone could shed some light on these topics, with just enough detail to understand the bigger picture and just enough information to get started. Some of the details which I think would be useful are:
Q1) What is the relationship between LLVM and clang?
Q2) I know LLVM has a large set of existing static analysis techniques are these called passes?
Q3) What code representation format do I need to use to run a pass?
Q4) How do I run passes?
I know this is a very broad question. I am just looking for someone to give me enough information so that I have a footing and can get started fiddling with these tools.
I am on my first job after completing my studies. My company does not follow any good coding or documentation practices, just a few basic rules. So there is no documentation of my project or any comments in the code. My only source of help are my teammates who clear my doubts if I dont understand something and they explained me the basic structure of the project.
So any advice on how do I understand such a project considering I am completely new to the corporate world and have no prior experience to understanding projects.
What steps should I take to understand the project better?
I read the below link
Advice on how to understand a project with long history quickly
but in that situation he actually has docs to read.
Thanks for your help.
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 3 years ago.
Improve this question
I write hobby code from time to time. The thing is these tools, classes or tiny libraries of code end up in a flash stick with hopeless future! I would love to develop my projects further, and let other programmers trust them. If you were going to use something you found on the Internet, what is the most important thing you look for in that programming tool or small library? e.g. would you consider separate documentation a must?
Thanks for all contributers. I'll try my best to summarized what have been said. Feel free to modify the list. Corrections and additions are more that welcome :)
Start a blog and let others know you
are here.
Choose the most
suitable license. Possibly Open
Source licenses are the best for
hobby projects.
Put your project where people can
reach it. Consider google-code,
github, sourceforge or
other sites.
Use public version-control and
bug-tracker, So others can acquire the
latest source code of your project to
compile and use.
Write a decent documentation, beside
commenting your code clearly of
course. The documentation should
explain the purpose of the library
and provide at least simple examples.
Write tests if you are willing to provide real-world code.
If you are building a library, put a
lot of effort into designing a stable
interface.
Get a blog, release code through it. Explain why you wrote it, what problem it solves. And encourage others to improve upon it, keep the code posted as current as possible. If your tools are useful you will very quickly develop a following that 'trusts' your code.
Separate documentation isn't a must for small tools, but anything creeping into the framework world should probably have ample documentation and examples if you want any serious adoption from the community at large.
The most important thing is that the library is that it be open source, so I can read the code myself. If that is not possible then I insist on documentation.
Also consider using a project-hosting site (like google code or github).
Have a clear license with your code if you don't have one already
(preferably one which encourages modifying / improving / sharing your
code ...)
Have public version control and/or a public bug/issue tracker and/or a mailing list. There are a lot of good sites which offer these services for free.
Seperate documentation is not a deciding factor to me (if the code is well documented and the code quality is high).
Documentation explaining why you wrote it, when you started it, and it's intended function. Understanding where you're coming from will allow me to see future ideas as well as short coming you may not have seen.
Technical documentation explaining the API and some examples on how to implement it. Ideally, keep your documentation in the format that follows the language. For example C# tends to use the XML syntax for defining items. This allows me to feel at home when I'm reading it.
Clean code -- I can't stress this enough because far too many people write exceptionally ugly code. If you're code is ugly and/or unreadable, it may be easier for me to write it from scratch on my own. At the very least, make your code consistent. If I can't understand the code, I won't feel comfortable with it.
Historical records explaining your changes. Seeing how the project has grown allows me to plan better. It also allows people to see how you learn from your mistakes and get a sense of your skill level. Compared to a forum, you can get a feel for how fast things get fixed and then placed in to a new release.
Think long and hard on what kind of license you want there. Public domain? BSD? GPL? More restrictive?
A note on whether or not you mind being contacted and if there are any restrictions in this. For example, would you mind updates? Me explaining security holes? Or perhaps you might use a forum or wiki?
The ability for me to get your latest work and/or nightly builds. SVN or something. This is useful so I know if a bug I found is already fixed.
I think that documentation is a key point for your project.
The document must indicate:
what is the purpose of your library
what are the main features
a really short tutorial, to make it run in 5 minutes.
Many examples
I let people trust my code in a number of projects, but I urge people to make and maintain their own tests, and I make sure that I'm content with the unit tests.
Documentation is always good, but I'm very guilty of finding time to do as much as I would like. But having the author fairly contactable is helpful.
Posting it in an open source repository such as code.google.com or sourceforge.net is probably where to start...
Next to attract attention, document clearly and succintly the purpose of the library / application as outline in one of the answer above.
Finally, blogging and direct mail exchanges happen...
One reason documentation helps people trust your code, is that they know whether a given feature is something which you intended the code to do (and which you will, all else being equal, preserve in future versions of the code), or something that the current code just so happens to do, but which might change at any time as a side-effect of a bugfix or just a refactor.
Some people prefer find out what code really does by looking at it, and that's fine, but documentation tells you (a) what the code is supposed to do, and with any luck (b) what the next version of the code will do. If I want to use your code long-term, and take bugfix updates as you provide them, then I need to know that you've designed an interface that I can rely on and that you're willing to stick to. Documenting it is a strong hint that you're at least trying to do that.
I'm not really asking about how programmers learn how to program. More about specific technologies. If I wanted to learn Hibernate, how do I know what I should know prior to Hibernate? Should I learn JPA before, during or after Hibernate? Is there a better solution to Hibernate? (And I'm not really looking for information on Hibernate specifically)
Maybe stackoverflow is the place to find these answers, but it seems like with the shear vastness of frameworks, apis, libraries, programming languages, platforms, and whatever other techie word you want to use, it takes an extremely long time to come up to speed on what technology to use, when and what you need to know prior to using it.
Sometimes the best way to learn is to just dig in to a framework. Sure, you could use someones wrapper API around something, but if there is something wrong w/ hibernate, then you wouldn't know what's happening.
And to answer "how do i know what i should know prior to hibernate", you don't, that's why you are learning. When learning about c++, started out with simple data types, but i didn't know about pointers yet, didn't need to, but i learned about them when i got there. Just gotta jump in and start playing around.
I use Wikipedia to compare various technologies to copmlete a task, although it can be incomplete with regards to commercial closed-source frameworks (probably because fewer people have access to them).
For specific technologies such as Hibernate, Java, JPA, LDAP (OpenLDAP in particular), Log4J, anything Apache: they all have wikis and/or forums associated with the product that are usually more helpful than a Google search for learning. Many even come with tutorials and you should try them.
Find a book on the subject and read it. Then email the author with additional questions. Most of these authors are more than happy to help especially if you've bought and read the materials they worked so hard to produce.
If that's still not enough for you, go to a conference covering the subject, if you can make it. Again you can meet many of the people responsible for maintaining and/or creating these technologies and I've found they are always willing to answer questions.
go to sites like Coding Horror, Slashdot, Techcrunch etc and find out what people are talking about. Usually if something is popular it's probably something you might want to talk a look at.
There are these things called "books" that are filled with all kinds of knowledge.
A lot of the time the documentation and/or tutorial for any technology or project will mention what prior knowledge is assumed or useful.
So for example hibernate: http://www.hibernate.org/hib_docs/v3/reference/en/html_single/#tutorial-intro
"This tutorial is intended for new users of Hibernate but requires Java and SQL knowledge"
For me, the things that have helped my career and taught me what questions to ask are:
Podcasts -- .NET Rocks, etc., which introduce and discuss new technologies and put them in context
Join your local users group, and stick around after the presentation to talk shop with the folks there; you can learn a lot just by hearing what other people are doing and what they are working on learning next
Just look around online and start trying to use whatever tool/technology your trying to learn. As you try to learn one thing, you'll realize your lacking knowledge in other needed areas. at which point you can repeat the process of looking around for this new item you need to learn.
for example, maybe you want to learn Rails, so you start following rails tutorials, but you realize you suck at Ruby. so then you start to focus a bit more on the details of Ruby, then come back to Rails with a little more knowledge and continue on till the next roadblock. this isn't really totally correct, but you get the idea.
you won't always find a full guide of how to use everything. just give it a shot and work it out on your own if you have the time
There is an infinite number of things one could learn. Maybe a better approach would be to think of a project that interests you, or join an open source one, and then learn what you need to know to accomplish what is needed in that project. When you're done, pick a new project that might include new things not learned in the last project.
As far as free sources are concerned, as a .NET programmer I like www.asp.net, and there are many others, such as the ASP.NET quickstart tutorials at http://quickstarts.asp.net/QuickStartv20/default.aspx, C-SharpCorner is good, too, if you don't mind C#.
If you don't object to paying a little money, Lynda.com is a decent place. They have OK tutorials on all kinds of things, not just programming, and I got a decent grounding in Javascript using one of their tutorials. They are adding new things all the time, so if they don't have something on Hibernate now, they may later on. I think their basic rate is $25 per month, but you can just pay for one month and then soak up as many courses as you can find time for.
Asking a more specific question will get your a more specific answer here. When I want to read up on something I usually head to Wikipedia and then Google.
The truth is none of us have the time to read everything we'd like to. So I let someone else do it for me!
The way I solve this is by speed-reading the web - aka. subscribing and reading to other peoples blogs.
Everytime I come across something I'm not familiar with I google it.