Natural Language Processing and Quant Jobs?

[Jim]

I'm a working software engineer thinking about going for a PhD. I'm not "gunning" for a quant job, and am simply interested in Natural Language Processing and A.I. in general. But I just want to see what my options are if I ever decide to pursue an advanced degree.

Seeing how the quant job landscape is constantly changing, do you see any room for NLP/Comp Ling. people finding jobs in finance?

 

[Alejandro Canete]

Hi dddddd,

NLP/Comp Ling is computationally intensive so you have to learn to use clusters and work with 100 gigs data files on a daily basis (thats useful in the HFT/stat arb space) and depending on the approach that you follow (i.e. probabilistic vs machine learning vs structural) you might be able (in the first case) to transfer some skills to financial modeling, but (IMHO) the core skills in both spaces are fairly different.

Hope it helps.

 

[vincegata]

@acanete, what is "structured" approach?

thx

 

[Byron]

Lots of trading firms use AI - I know of companies which use GP or SVM to shortlist and optimize variables. (I'm an algorithmic trader in a firm in India, and have an undergrad in CS with my internship in AI and thesis/paper in NLP). There are even firms who trade on the news - use semantic parsing on news feeds to generate signals (there was some article recently about Berkshire Hathaway's shares rising when Anne Hathaway is mentioned in the news, check it out). My guess is that it will depend on how well you can sell your qualifications once you graduate, there will definitely be takers.

Personally, I think an AI/NLP background is helpful in this field because it trains you to translate intuitive human behavior (like say WSD) to formal, computational/statistical constructs. Sort of an abstract notion, but often useful in financial modeling.

 

[Alejandro Canete]

@acanete, what is "structured" approach?

thx

When you use heavy prior knowledge (syntactic model + a word ontology) to solve a problem instead of only data.

For example if you want to do machine translation between spanish and english you can do it by either

A) Training an optimal n-gram bayesian posterior discriminator using a paired set of texts (in both languages) (this is a probabilistic approach)

or,

B) You can use a syntactic model (tagger + word ontology + disambiguation rule) + a spanish/english dictionary (yep.. an actual dictionary :P) to do a symbolic type of translation. (this is a structural approach)

This is of course an extreme example, practical applications are usually in between.

 

[Neuron]

Lots of trading firms use AI - I know of companies which use GP or SVM to shortlist and optimize variables. (I'm an algorithmic trader in a firm in India, and have an undergrad in CS with my internship in AI and thesis/paper in NLP). There are even firms who trade on the news - use semantic parsing on news feeds to generate signals (there was some article recently about Berkshire Hathaway's shares rising when Anne Hathaway is mentioned in the news, check it out). My guess is that it will depend on how well you can sell your qualifications once you graduate, there will definitely be takers.

Personally, I think an AI/NLP background is helpful in this field because it trains you to translate intuitive human behavior (like say WSD) to formal, computational/statistical constructs. Sort of an abstract notion, but often useful in financial modeling.

Where can I find information on technology using semantic parsing on news feeds to generate signals? I am alway interested in trading technology outside of the world of general stat arb. or typical HFT methods